Rob Csongor – Vice President-Investor Relations
Jen-Hsun Huang – Co-Founder, President and Chief Executive Officer
Jeffrey D. Fisher – Senior Vice President-GeForce Business Unit
Philip J. Carmack – Senior Vice President-Mobile Business Unit
Steve Allpress – Vice President-Mobile Communications Software
Daniel F. Vivoli – Senior Vice President
Sumit Gupta – General Manager-Tesla GPU Accelerated Computing
Karen Burns – Interim Chief Financial Officer
Shawn R. Webster – Macquarie Capital Inc.
Dean Grummels – Stifel, Nicolaus & Co., Inc.
JoAnne Feeney – Longbow Research LLC
NVIDIA Corporation (NVDA) Annual Investor Day Conference Call April 11, 2013 12:00 PM ET
Good morning guys, good morning. For those of you who don’t know me, my name is Rob Csongor. I run Investor Relations here at NVIDIA, and I’d like to welcome you to our Annual Investor Day. As you guys know there is a lot of exiting things happening at NVIDIA, and I want to spend just as little time as possible on stage but just give you an idea of what we’re going to be talking about today. There is so much stuff, so many things that we can talk to you, and based on the feedback that you guys have been giving us, we try to focus the material today on the things that you want to hear as well as for us to share the things that we think are important for you guys to understand about our business. So we try to walk that fine line and making sure that you get as many of your questions answered, as well as get insight into what we’re doing.
Okay, so let me just give you a flavor of what we’re going to talk about today. Of course please note our Safe Harbor. What we have in terms of the agenda is, first of all Jen-Hsun is going to talk you about the business, right. And he will run through a number of the critical things of course that matter to you guys. We are going to give you an update on our GPU business. And we have Jeff Fisher and those of you who, I think you are familiar with Jeff, Jeff has been at NVIDEA for a very long time, and he is one of our regional employees. I think I’ll focus something of course very important to you guys is our Tegra business. And presenting to you will be our Senior Vice President of the Tegra business. He started the Tegra business, Phil Carmack and he is going to run through all the material on Tegra.
We also have with us Steve Allpress, who is the Chief Technology Officer, behind the brains of our DXP processor, the heart of our modem. All right, so we are going to make sure that we cover a lot of this material for you. We are probably going to take a break there for lunch; it will be an interactive lunch. We will take about half hour, 45 minutes executives will sit with you, and you can some interaction but we want to get quickly back into it in the afternoon. We are going to give you an update on GTC, and for those of you, who are, I think increasingly you guys are aware of the importance of compute and how it’s not just about Tesla, but it’s important now increasingly across all of our businesses.
So we think it’s really important for you guys to understand what is happening at GTC, because that is where you see a lot of the early indicators of what is going to be driving our business. We are going to follow that naturally with an update on Tesla. So Sumit Gupta, Dr. Sumit Gupta is the GM of the Tesla business. There is a lot of exciting things happening there. We will follow that with financials, and we want to make sure that we have plenty of room left over at the end for Q&A. So in case we schedule this end, we’re not trying to run this until all day we purposefully try to focus in material and leave time at the end. So that you guys can ask questions, get Q&A. We’ll get all the executives back on stage. And then we will have a lot of time for your Q&A.
Okay, so with that, I am going to get off the stage, at this point please let me introduce the CEO and Co-Founder of NVIDIA, Jen-Hsun Huang.
Thank you, great to see all of you guys. This is one of my favorite events all year. We got a lot to talk about today, and so let's get right to it. Before I start and talk about our company strategies and how we are investing our resources, let me first frame what's happening and the conditions surrounding us at the moment, and I think if I just show you in the next slide that will be relatively clear almost right away.
The PC industry is down, business conditions are rather muted in the first half. The PC has declined more this last quarter than any time in the last 20 years, in fact more in the last quarter since IDC started tracking the PC industry in 1994. Obviously the computer industry is changing. The PC still represents a large part of our revenues and so impacts our business.
The second part, the second characteristic or condition is Tegra; a lot of you have questions about what's going on with Tegra? Where do you think the business is going to be like this year? And we made some decisions last year that are really important to consider. There were two things that we wanted to do last year, that we thought it was really important for our company to pull in, two investments that were really vital. One investment was our integrated modem. The LTE marketplace is growing faster than frankly any transition has happened in wireless.
And it was important for us to engage that marketplace as quickly as possible. We have all the necessary assets to be a big player in this space. And so we decided that we would sacrifice the schedule of Tegra 4 by about a quarter so that we can pull in Tegra 4i by more than two. We have production silicon now of Tegra 4i. That was a good decision. By all measures although it does sacrifice a quarter or so of revenues for Tegra 4, where we held position to win some businesses, we were very well positioned still to win a whole bunch of them, but net of it is that we were able to pull in Tegra 4i by more than two quarters.
Phil later on in his conversation with you, he is going to talk to you about the implication of Tegra 4i and how it is positioned in the marketplace, and why we are so excited about Tegra 4i, and what enabled it to achieve the price performance that it achieves, which is at the core of the Icera modem architecture.
The second thing we wanted to pull in is something I will show you a little bit later on, it has to do with the fact that our crown jewels. Our GPU, the highest performance GPU's in the world can make a real contribution in mobile. It can make a real contribution in Android, it can make a real contribution in these new digital devices, and we wanted to do what we could to bring in our crown jewels, mobilize it, shrink it into a form factor that allows us to bring it to the marketplace during this very important transition to DX11. I'm going to talk to you about that a little bit later on, but the net impact of it is that Tegra 4 relative to its previous generations of Tegra 2 and Tegra 3 is off by about a quarter.
Now when you look at our shipments, you could see the peaks, each one of the peaks are about a year apart, and the reason for that is because those are the product transitions. Tegra 2 was the world's first dual-core processor, generating enormous amount of enthusiasm and excitement, shipped for about a year because it really only served one segment, and then we were off to Tegra 3, which was the world's first quad-core processor, shipped for about a year, and each time it leaves a longer tail. Each time Tegra 2 leaves a longer tail, Tegra 3 leaves a longer tail, but nonetheless because we are essentially a one chip product line at the moment. These peaks and valleys happen once a year.
Now long time ago for those who have been watching us over the years, we have the same challenge in PC graphics, we had one chip and because of the inspiration, the innovation, the creativity of the Company that one chip was always incredibly successful, and we would ramp up and then ramp down, ramp up and then ramp down. It wasn’t until we institutionalized the methodology that many analysts characterized as three chips, two seasons, and a couple of analysts in the room, I think quite remember that quite well. Three chips two seasons make it possible for us to address multiple segments in the marketplace and continue to grow as we introduce new capabilities into the marketplace.
We decided last year that we would make that one quarter sacrifice so that we could have a multi-chip family, so that we could have a multi-chip family, so we could address multiple segments in the marketplace at the same time, and that we would essentially bring this three chips two seasons to the marketplace in the mobile marketplace, so that we could start to build on our franchise business, and hopefully reduce this volatility in our business. Those were the decisions that we made last year, I’m incredibly happy that we did that. We are better positioned today than ever, but there is that lumpiness, second thing.
Third thing, we increased our OpEx this year, because we are investing into some very important opportunities, once in a life time opportunities. I’m going to dedicate vast majority of my conversation with you guys today. Understanding what opportunities that we see our strategies and why these investments are so timely must happen now once in a life time opportunity for all of us.
So the net of it all is that the first half is rather muted for a couple of reasons. But we are confident that as we go into the second half the growth will return. Net-net one of the questions that many of you have is how do we feel about the Tegra business? Net-net we think that although we've been growing very quickly in Tegra, this year we’ll be about flat. This year Tegra will be about flat, because we excluded about a quarter of Tegra 4 shipments out of the business, but net-net it will be about flat.
Okay? So that captures the current condition, now I want to switch gears and talk about the future. NVIDIA is the world leader in visual computing. We are in the graphics business. We are singularly focused on the graphics business. We're passionate about the graphics business. We are expert in the graphics business. We've contributed more to modern computer graphics than just about any company in the world. From the invention of the GPU, the invention of the programmable shader, fundamental investments and fundamental inventions have led to about 5000 fans around the world.
The look of modern graphics and then recently several years ago inventing a technology called CUDA that bought GPU computing to the world, making it possible for us not only to create beautiful graphics, but to take graphics really to a completely new plane a technology called computational graphics, and the visual effects have really, really changed, and we'll show you some of that today.
We are the world's only visual computing company, dedicated visual computing company. We’re end-to-end visual computing from the fundamental inventions of technologies to the processors that serve PCs, workstations, high-performance computing to mobile devices, modules that are incorporated into various computing devices to now systems we call a Visual Computing Appliance. The Visual Computing Appliance is a recent invention that makes it possible for us to put computer graphics on the network. It no longer has to be inside the box. Computer graphics can now leave the box. The benefit of leaving the box of course is that it doesn't matter anymore what operating system your box is, it doesn't matter how small your box is, it doesn't even matter whether you have a box. We are no longer constrained. Our growth is no longer constrained by the limitation, the physical limitations of the devices themselves.
Through our inventions, we have invented an end-to-end visual computing offering from technology all the way to systems, serving OEMs, serving gamers, all around the world. We have three target end markets; one of which is video games. Why is video game so important? Well, the reason for that is as it turns out video games is one of the most technically challenging applications in the world. It has to be beautiful. It has to be high-performance, meaning it has to be real-time from the moment that you press a button to the moment all of the processing needs to be completed. It has to be done in tens of milliseconds. Best effort is not enough, we have to get it done on time. We use a sense of interactivity. It is scientifically very challenging and the reason for that is because they are trying to simulate reality, they are trying to suspend you in an alternate reality. And if it doesn't feel real, look real, behave real, it's just not as fun.
Video games also has another characteristic that's quite unique. Of all of the challenging and heavyweight applications, few are as high volume. This market is absolutely enormous. There was a time 20 years ago when we used to have to talk about gamers, the notion of gamers. Today, I don't really know anybody know any children who is not a gamer. It's a little bit like saying, are you a movier, are you a musicer, are you a booker, are you a gamer? Yes, we play games. It's fun. It's beautiful. Gaming is a large market for us, important market for us.
Enterprise computing, the creating of the games, the designing of cars, the making of tennis shoes, the designing of planes, trains, automobiles, making movies, creating videos, translating creativity, and translating imagination to experiences is the work of enterprise computing today. Making it possible, using computer graphics to help us connect with information, using computer graphics to help us connect with imagination that is the work of workstations today, that is the work of enterprise computing today. We served every computing company in the world. Our workstation business is the standard of workstations. I don’t know any car companies or plane companies or tennis shoes companies for that matter, who don’t use Quadro at its core to design their products.
In 3D Graphics, the visualization if design is of importance to you, it is likely the case that NVIDIA graphics is at its core, and the reason for that has to do with the technology, the capabilities of our technology, but the compatibility that we have with the entire body of software work all over the world, decades now of work products have been build on top of our GPUs, on top of our processors, on top of our platform. That compatibility that assured compatibility that you can bring up the Boeing 777 database 20 years from now, on a Quadro is the reason why you rely on NVIDIA’s graphics.
Enterprise computing, third devices; volume matters, the reason why we disrupted the companies we have disrupted in the past is because of volume. Volume translates to velocity, volume translate to innovation velocity, revenue velocity translates to even more energy in your business, volume matters. The PC was the high volume business of our time, 20 years ago when we first start the company. The PC was to us at the time, we were the last PC technology company to come in to the world.
I know it's surprising, we’re the only one standing today, but yet we were the last one to show and the reason for that is because 3D graphics is the heaviest of all the technologies and it takes the longest for it to be possible on any platform because it’s just so heavyweight.
Today the high-volume devices, mobile, tablets, cars, TVs all of these visual devices add up to some 3 billion units in a couple of years, 3 billion units a year, billions and billions of these devices are being made. We don't serve all of them just as we don't serve all of the PC industry, but for us to be able to serve 20% of the PC industry and the thing that we can serve a large percentage of these devices make perfect sense to me.
Some of these devices, not all of them are going to want the world's best computer graphics. Some of these devices, not all of them are going to want the visual computing capabilities that we offer. We can make a contribution here. These are our three end markets.
Now, before I go into our strategies and address the niche markets, I think it's instructive to look at what's happened in our industry in the last 20 years. We were founded in 1993. In 1993, the PC – the idea of a home PC was just still developing, CD-ROMs haven't really become mainstream and surely haven’t become integrated into PCs. The notion of an LCD display doesn't exist. There was the CRT. The resolution at the time 10x7 and it required a really high-end work station to be 12x10, and if you go absolutely ridiculously crazy, it will be 16x12, but they were all CRTs. The size of CRTs constrained the size of our business opportunity because the physicality of it doesn’t allow it to scale, eventually LCD displays showed up and our growth took off.
A lot has changed since then and you guys know the changes that are happening but when you look at it on a scale like this and how quickly it happened, there was peace in the industry for 17 out of that 20 years and then three years ago everything changed. It doesn't matter what your market position was, it doesn't matter whether you were characterized to be so strong in your market that somebody even described you as a monopoly. You were disrupted by some fundamental computing forces and it happened literally overnight. That's why there are so many companies, caught ill prepared without a strategy, without a position to be able to address this new future. That's why it’s so important for us to not let our foot off the gas right now. The opportunities are enormous, as well as the threats.
Now the computing architecture of today radically different than 1993, but what's amazing is the opportunity. If we use processors, the white line is basically the total number of processors shift. Each year, more ARM processors are shipped than all of the x86 processors shipped since the beginning of mankind, and apparently it's still growing and there is every evidence that it absolutely makes sense, that it rather keep growing.
Although things have changed so much, the opportunities are also absolutely anonymous. We are well prepared to deal with this new future. This new future has PC still utilized by many to create content for the heaviest weighted applications, for productivity, for design, for play, but surely for consumer average computing the PC is going to get supplanted by whole bunch of other computing devices that are cropping up everywhere. We are going to be surrounded by computing devices.
Here lies the opportunity, and so today we will like to spend all of our time with you talking about growing beyond the PC. We still have enormous confidence in the future of the PCs. I am going to still going to talk to you little bit about the work that we're doing in PCs, the work that we’re doing in the core part of computing, and why it’s still so exciting, and why there is early indication that growth is going to continue. But there is some characteristics here we like to talk about today, elements of our strategy.
Visual computing is more important than ever. All of these computing devices are connected to high definition displays. The display is the computer. And if you are not sure about that look at what people are touching these days. They are touching displays, and those displays are going to become more and more beautiful over time. There are 3 billion of these high definition displays connected to devices. If the number of those – if the percentage of those devices that we serve is consistent with the number of PCs that we serve, it’s a very large market opportunity for us.
Even at its core GPU’s are still growing. Although the PC industry has been flat, although our closest competitor has been effectively flat, our business has grown at 12% CAGR. The reason for that is because of the work that we did with CUDA, the work that we do with GPU computing. We talked to you quarter-after-quarter, year-after-year about GPU computing. I appreciate the graciousness that you show us and listen to all those stories about CUDA, and why GPU computing is so important.
There is every evidence that we’re at the tipping point of GPU computing, that the GPU itself has morphed from a graphics processor to a general purpose parallel processor. We’ve extended our GPU into the network, we call it GRID. GRID to us is a little bit like a router, it’s a little bit like a storage appliance. It’s a visual computing appliance, and it was made possible by several inventions; one, remoting, to disassociate graphics from the computing device. Two, virtualization, to make it possible for multiple users, multiple tenants to use one graphics processor at one time and all of the load balancing software and system software that's necessary, so that we can all share a performance critical appliance and yet have good experience from it, good quality of service.
Tegra; people call Tegra mobile processor, I think that's fine. To me Tegra is just a computer. Tegra is not about the future and various feature in smartphones only. Smartphones are very important to us. It's a very high-volume device. And volume as I’ve told you, matters. Tegra is the future of computing. Someday, I am not sure whether it’s seven years or eight years or nine or 15, but someday every chip we make in the company will be a Tegra.
Tegra’s will become better and better and higher and higher performance on the one hand, on the other hand all of our GPUs will incorporate some processors in it. This is not about us growing into mobile devices only. This is about us inventing the future of computing altogether for ourselves, for the industry. Tegra is an essential future of our company. It is much, much more than smartphones. It is everything computing to us.
And then lastly, I recognized that we are investing, I believe we are investing smartly, I’m going to show you some of those numbers. It’s because of these investments that made it possible for us to grow at 20% CAGR when the rest of the industry is basically flat. And it’s because of these investments that made it possible for us to feel a sense of confidence about our future, and our cash generation capability that we instituted our first dividend, Q4 of last year.
And so let me now jump into these discussions in more detail. Let me illustrate to you our strategy. As you know we have three basic strategies; one, continue to investment in the future of GPUs, advance the GPU, make the GPU more useful, make it more versatile, expand the application reach, new applications go into new markets, make the GPU more programable, CUDA, GPGPU, all of it contributes to that.
The second part of our strategy, the second strategy is to take GPU out of the box and put it on the network grid. The third, to leverage our visual computing expertise into devices. Okay, basically three strategies; CUDA, GRID and Tegra, maybe the way to summarize that.
Well, as I illustrate – as I think of us illustrating this, one way to start is just conventional wisdom. Conventional wisdom and this is a question that people have asked us year in and year out, year in and year out, well probably more frequent than that. Day in and day out, day in and day out, is when will the GPU be integrated into the CPU? Well, that conversation started in 1994, in 1994 it started. And we’ve been having our conversation every single moment since then. The answer is, if we did nothing, if we didn’t invent the GPU, if we didn't invent the programmable shader, we didn't invent CUDA, we didn’t even GPGPU, we didn't do all that. If we didn’t invent DX11, then the answer is long ago. But there is so much headroom in innovation and visual computing and there is still so much headroom as we stand here. That the GPU has become so complex that the question now is, do you integrate the CPU. And the answer is yes, of course. The GPU integrated CPU turned into Tegra.
But nonetheless, conventional wisdom would say, that eventually the GPU will be integrated. We continue to innovate, we continue to see growth, fuel demand because people like new experiences, new experiences feel demand, because the GPU became more programmable, the programmability made it possible to accelerate more applications, more applications increased our market reach and field demand. That was responsible for a 12% CAGR.
And then, we invented this idea called Grid, like a net created the world's first of its kind visual computing appliance like a net app storage appliance like a router, networking appliance both streams by the way used to be inside a computer, you guys might remember this. But the world's first router was the Sun 3/280, it was a file server because of course it could route, it could take jet protocol and translate it to Apple Talk. Of course it could take Apple Talk translated to Ethernet and the reason for TCP/IP and the reason for that is because it’s just software. We created the world's first visual computing appliance, putting it out of the box. When you put it out of the box it creates brand new opportunities and I'll show you some of that today.
But nonetheless, as much as many markets need great GPUs, we have to recognize that long-term as we continue to go, at some point maybe the low-end part of the market, the basic PC marketplace don't need the capabilities that the high-end does. The average users find their PCs to be good enough, in which case integration becomes a more valuable thing for that user than capability. You extent that one step further, it’s the reason why tablets are cannibalizing PCs. Now you guys have heard me say this before.
When somebody says the PC is becoming increasingly good enough and therefore GPU have to get integrated into the CPU, you have to realize the next shoot or drop is that the PC has become good enough rather be replaced by something else altogether? A tablet, integration doesn't just happen to the GPU integration happens to everything, and so our strategy in recognizing us.
Let’s get a head start on this feature. If there is anyone who is going to go disrupt the PC it ought to be us. We should take the opportunity as early as we could and as hard as we could to go build this future where their devices, where people enjoy the multimedia capability of it, it’s not designed for computing, it’s really design as a computer for enjoying content, to enjoy this device wants to have great graphics, great multimedia, great interactivity, great content all of the things that we’re good at.
The things that we’re good at. And there are things that will become increasingly important in this marketplace. This marketplace has turned out to be real, we talked several years ago, couple of years ago about the importance of Android tablets, it was questionable at the time, the future of the Android tablet but you guys know that I never gave up on it. Because it was too obvious that it was going to happen.
All of our digital content was going to be in the crowd and you’re going to enjoy that on your phone when you’re mobile, you can enjoy your tablet, when you sit home at on your touch. And so our growth strategy now basically has three elements, continue to expand the reach of digital computing make it more capable, make it more programmable, make it more usable, making more beautiful all of those things.
Leverage that capability on to the network with GRID and leverage that investment, leverage that asset into devices through Tegra, three basic strategies and we hope that with these three basic strategies it creates a large growth opportunity for us.
Now the question is, how do we invest to do these things? They seem like three things, but to me it’s an extension of one thing. It’s all about computer graphics. It stems from the GPU. It stems from our knowledge about computer graphics. It stems from our knowledge of our visual computing. So how do we create an organization thus able to create these things and compete with giants? And so the way that we see our investment model, this is basically a mixture between the company’s R&D architecture and how we invest.
We think of ourselves as having a core R&D platform. It’s basically the GPU. It is the beginning of almost everything we do. Just as for Cisco, the beginning of everything they do have something to do with networking. Just for AMC, everything they do have something to do with storage. For us, everything we do have something to do with computer graphics. The GPU has had its core. There are several elements of investment in GPUs. This is how a GPU is made, there is the architecture part, there is the design part, there is software part, the APIs, the OpenGLs, the DX11to OpenGL ES’, the OpenCVs, all the different operating systems from Windows RT to Windows to iOS to the work we do from Mac OS to Android, all of that is embodied in software.
We don’t want to repeat that work for each one of our businesses, you can’t, it’s too much and it’s a huge asset, an enormous body of work that is so valuable to our company. We leverage it even when we put the GPU on the network. We leverage it into GRID. That’s why all of the applications just run on GRID. We leverage it into our Android devices. And that’s the reason why we were partnered with Google on the first Honeycomb that required the first OpenGeo implementation for – hardware OpenGeo implementation for Android. It’s the reason why Microsoft selected us to work with because it required a fundamental investment in DirectX. We have the system software expertise to help companies build an entire stack.
VLSI. VLSI, the implementation of these GPU based processors, as you know, Titan or GK110 is the worlds’ largest processor. It's physically the largest processor in the world. More people worked on it than any single chip ever. The implementation of that VLSI is very, very extensive.
Content technology. We create new capabilities, we create new uses and the way to do that is to inspire content developers. We have to inspire them, we have to provide them technology, we have to provide them tools and we have to provide them support. So that the Adobes of the world, so that Autodesks of the world, so that SolidWorks and Dassault of the world. So that every game developer in the world can take advantage of the latest state-of-the-art ideas in computer graphics and implemented into their applications. Content technology makes that possible for us to do that.
Process R&D. As you know, these geometries are getting smaller and smaller and now we’re are talking about FinFed we‘re working on 12 and there is ‘16, ’14, ‘12] and of course behind. These geometries, the process technology, the alignment between our design and our ideas and what’s many factorable, requires a great deal of science. We have a process R&D team. There is no sense doing that over again. That entire platform represents about $900 million worth of R&D.
The two fundamental new things that we’ve been working on the last several years GRID and Tegra are incremental investments on top of that. GRID leverages so much of our core work in computing because it’s at the core fundamentally a workstation, a fundamentally high-performance computing server just extended onto the network. The invention took a long time, but the investment levels not significant and yet the market opportunity is. Tegra, building new processors, investing in new software stacks, Android for example, about $300 million a year.
Now those are incremental investments, $300 million a year. At the accounting level we could argue about how we should account for it. Some companies like to think about just contribution margins, and so the business incrementally cost $300 million a year. Now we would have spin that off as a company, it would cost us $1.1 billion a year to build it, maybe something less than that, but something a lot more than $300 million. So it makes sense for NVIDIA to build that in-house. Because it’s an incremental $300 million, if we did about $600 million of revenues, we’re breakeven.
So what that basically says is Tegra is running around $750 million a quarter – a year, [I am already] living into next year.
Printer, $750 million a year, $300 million of investment, 50% gross margins. This says that Tegra is break-even. Not every quarter, because some quarters are a little spottier than others. On an annual basis, Tegra is at already at break-even levels.
So these aren’t science projects anymore, they are product lines, and they are contributing to our company. The more importantly they are contributing to our growth. I’m going to talk about SHIELD a little bit later, and why it’s so essential to our strategy.
Now when you look at this investment, basically about $1.2 billion, you would never imagine that a $300 million investment can go toe-to-toe and be the reference, be the other company that these giants talk about. And the reason for that is because, obviously, we invest in incremental $300 million, but because we’re smart about our investment, it looks like a giant built this.
We are leveraging all of our GPU assets, which is 20 years in the making, the best in the world. And with an incremental $300 million we can redirect it and reapply it into all of these new computing devices. It makes it look like we are huge even though we only invested in incremental $300 million. And that’s really the whole point of it, leveraged investment. Leveraged investments allows us to grow faster than we could naturally otherwise.
Now when you look at the results, I think results are speak for themselves. Over the course of last three years, we’ve grown on a CAGR about 20%. Almost every single business is growing. And it was kind of interesting that when you are vibrant and you are strategically relevant to your customers. Every single PC OEM today has a desire to expand beyond the PC. Every single PC OEM has a desire to expand beyond the PC. When you are strategically relevant to them in one area, it helps your business even in the commodity areas.
For example, GeForce OEM, GeForce OEM is basically a basic PC and it’s growing at 12% CAGR. Now at 12% CAGR, some of it is market share, some of it is just because we are just a more important customer for people to work with, more important supplier, or more important partner for people to work with because there are so many other areas where the technology that we create are essential to their business. Essential to their growth, they would like to partner with us. And so as you can see we are growing across the board. Tegra is about 109% CAGR, Tesla 68% CAGR, Quadro is 15%, gaming is 9%, okay, growing across the board.
But what's really exciting to us is that we have traction now across these three areas; GPU, core GPU, GRID and Tegra. But the exciting thing is the market opportunity ahead of us. The TAMs are substantial. And then throughout the rest of the day and in some of my parts, I’ll talk to you about how we see the TAM, how does it come about and why does that make sense. But we are looking at now from our incremental $320 million of investment out of our approximately $1.6 billion of spend this year. That $320 million of investment is resulting in a TAM growth of about $20 billion.
So, now let me talk about each one of the three GPUs, GRID, and Tegra. GPU, the fundamental invention some six years ago, some six years ago with the introduction of G80 that GPU really revolutionized how we thought about computing altogether. Using a heterogeneous approach or some people say the right tool for the right job, CPUs are really good at single threaded applications. IBM calls it a strong thread.
Sequential applications that you can accelerate expect to run it faster. There are some things in life that are simply sequential or very difficult to make parallelize. And there are some things in life that is very easy to parallelize, very little dependencies. The two processors are complex in its own right in very different ways, whereas the CPU has one or two threads per core, we have thousands of threads, work threads, threads of work, lines of work running throughout our chip, thousands of them all at one time and we have to keep it coherent.
Very, very different type of work. The two type of processors come together to do something very, very obvious in computer graphics. Computer graphics is just a piece of software by the way. It’s no different than Excel or no different than a fluid simulation, it’s just computer graphics, just software. That software runs in parallel, some of it is sequential, lots of it is parallel. But that CUDA made it possible for us to expand the applicability of GPUs to far beyond computer graphics to fluid simulations to mechanical simulation to mechanical simulations to particle simulations, molecular dynamic simulations, CT reconstructions. There was just an article I read in Wall Street Journal about how every imaging company, diagnostic imaging company is trying to reduce the dose of CT and the reason for that is because they are herding many people more than appearing.
While every one of them is in the process of reducing their dose by a factor of 20, and at the core of that capability is CUDA, they’re CUDA customers. Imaging is $20 million business for us each year, a [prime run] for the rest of time. And it’s such gratifying news to know that the work that we do make a difference, in that one little area.
We are seeing – starting to see all kinds of use of instrumentation now whether if you can imagine using that for CT, you could imagine using that for machine visual inspections. So that you could do at line speed instead of only visually inspect one out of every ten, you can now visually inspect every single one.
CUDA has really revolutionized computing. And you can see the results. We have been growing overall GPUs 12% CAGR. The industry our closest competitor both effectively flat.
Our market share has gone up a little bit, but more importantly, our gross margins has gone up a lot, because the contribution we’re making to the system is greater and greater and greater over time. The CUDA investment is approximately somewhere between $60 million to $80 million a year, it’s not insignificant. We made a conscious and a quantifiable investment in GPU computing. And you guys can tell because I talk about it every single time in front of you. And there is every evidence that GPU computing is revolutionizing our industry and the growth is reflective of it. It’s revolutionizing the industry on a whole lot of different points.
I mentioned earlier, if not for the invention of the GPU, if not invention of shaders, programmable shaders, if not for the invention of CUDA, computer graphic won’t continue to become more rich and if the look and feel of computer graphics continued to be more rich, why would people buy the new computing device? Why would people buy up? Why would people buy our best stuff? Now we’re going to show you guys some stuff today that is just utterly mind blowing. We went from cartoonish type looks to now completely cinematic looks. You can’t tell the difference hardly anymore between the work that we do in real time versus the work that is done offline in cinema.
A whole new level in computer graphics and all of that would just fuel our growth. What drives our growth in computer graphics is really very simple. The breakthrough technologies that we introduce every now and then that causes everybody to want the next great thing. No different than from DVD to Blu-ray and then hopefully from Blu-ray to 4K. It is so visually stunning that people want the next thing. Second thing, of course, is the great titles. You need the next Iron Man to fuel people going to theatres. And then of course more and more people who can enjoy video games on more and more platforms. Those are the three fundamental drivers of our business.
We start by creating the next breakthrough in visual richness and get people all excited about it. We don’t just make it beautiful now, we don’t just make it fast, we simulate, if not just art anymore, the line between art and science is becoming rather blurry and imagery are really quite stunning. Breakthroughs in computer graphics on so many different dimensions as a result of the invention of GPGPU, as a result of the invention of CUDA.
Scientific computing we talked about and I don’t think you can deny (inaudible) these are early adopters. This is what in our company we call EIOFS, early indicators of future success. This is the top 500, the number of systems that have adopted GPUs for scientific computing around the world is growing at a pretty remarkable rate. Another way of thinking about this is what some people call the tipping point. All of a sudden one day, the day before you thought, well, it’s might not be for me, the day before that you thought this ain't going to work. And then all of a sudden, one day at the tipping point you realize everybody else is doing it, and it makes perfect sense, the tipping point. You could argue, we’re at the tipping point, we’ve had a very genius computing.
What’s not just for computer graphics is not just for scientific computing anymore, it’s for our mainstream, without us doing anything. Engineers all around the world are using it for big data now. Shazam, they did all themselves, brought some graphics cards because every NVIDIA graphics card is compatible, downloaded the SDK, and now they are doing 10 million queries a day against 27 million entry library of content. Whenever you do a Shazam, whenever you do a tag, within a few seconds it has to comeback with a query.
eBay is trying to do the same with Cortexica. Now with imaging because eBay has all kinds of stuff on it and it’s tough to characterize exactly what it is, but you can take a picture and say, give me a whole bunch and that looks like this one. The inverse search, computer vision search, search the way our mind can do it in a way that search engines cannot, you finger print it. This is essentially the world’s fastest fingerprint search engine. You take a picture or you take a photograph, it creates 500 point fingerprint and it searches against a whole library of other images, 500 fingerprints each and responds back to you in seconds.
Tweeter; there are 500 million tweets, I think it’s growing at 50% a year. And one day somebody discovered a great thing, those guys at Salesforce.com are clever. When it would be great if I could use that social energy and discover the sentiment of a brand, and so there are all these search terms, they call them expressions, they have about a million expressions and all these 500 million tweets are flying by and they are filtering it against this 1 million expression filter. And they have to respond in minutes, it took them hours before and when they ported it to GPUs when they use CUDA instead, now they respond in minutes. They are seeing an exponential growth here. The number of customers who like to know the sentiments of the brand around the world is growing, the number of expressions are asking them to filter is growing, and the number of tweets are growing.
GPU computing is really making great progress. GRID, what is GRID? Think of GRID as a graphics card that contains a whole bunch of other graphic cards. It’s a system of graphics cards. What we’ve invented is a technology that makes it possible for you to put it on the network, to disassociate it from the basic platform. So I can give you awesome graphics even if you are using a MacBook Air, I could give you awesome graphics even if you are using a netbook. I can give you awesome graphics even if you are using an iPad. It doesn’t matter what operating system you use, it doesn’t matter what computing device you have, it doesn’t matter hardly even if how old it is, so long as you have H.264, which almost everyone does.
We download a little tiny client, a receiver, and [all of a sudden], this wealth of capabilities available to you. Now just to demonstrate it to you, we’re going to show you now GRID running, but before we show it to you, I need you guys to now suspend yourself and realize that what you’re seeing is on the network. I know it looks like it’s right there on the computer, it’s going to be right here on the computer, but each one of those applications are actually running on the network. It’s like Netflix, except it’s interactive, meaning that there is no buffering involved. The moment you touch a key, it sends it off to that server. We immediately process it and send the result back and it does it so fast and it does it in a compressed way that the network capacity is a non-issue and the interactivity is really good. And you feel like the application is right there on your computer whatever happens to be. Ian, take it away.
Unidentified Company Representative
Sure. Thanks, Jen-Hsun. So let me describe what you’re looking at, which, I am sure, you’ll recognize as a Mac desktop and let me just go ahead and launch three applications each of these. So each of these applications, as Jen-Hsun mentioned, are running on a GRID VCA effectively on the network, and each of these is like having a separate workstation and they are delivering the application to this Mac desktop. So I can go select, for example, this application, which some of you may recognize as 3D Studio Max from Autodesk. But the key thing about this application is that it doesn’t run on a Mac, but we are running on a Mac, as you can see. So we can deliver that visually rich content that someone would use to create modeling and animation and allow that person to now work from a Mac or an iPad or another device.
If I flip to another application, let me show you this one, this is Adobe's Premiere Pro and as you can see it's a fully interactive session. I can stop and rewind the movie and scrub around and do various different things. I can move it, but again the power of the GPU that Jen-Hsun was mentioning earlier, I can actually just go and show you what we can do in real-time and this is some of the value that people doing movie, editing on movie loops, I can achieve real-time go and apply a blur effect which as you can see is blurring that movie as it's playing. If I can have a GPU, it would stop the movie, apply the blur effect, have to post process it and then let you see the movie again. So this interactivity of being able to look at the effect of what I want to do is a key element of GPU’s.
Let me switch again to another application, this is actually SolidWorks, some of you may know, this is the world's most popular professional CAD application and as you can see the graphics are visually stunning. Again the GPU’s this is sort of state of the art and what we can do, what I just did there was to turn on something called ambient occlusion. And it's kind of obvious when you see it, but without ambient occlusion, what you noticed is the that there is a light source in the scene and that's making a highlight if you can see the cursor on the screen, it's making a highlight on certain parts of the object, but you would think that there would be a shadow behind that features of the object, but without ambient occlusion, you can't actually see those shadows, but with ambient occlusion it puts those on to the object itself.
Some people call it self shadowing, an object shadows itself. Okay self shadowing. This is using a technology called the tin shader. We invented this idea called the programmable shader, we can actually go into every single pixel and add a special effect if you will to it, this is really a tin shader. Imagine you created this design and you want to create a manual for it. And most manuals, mechanical manuals looks like cartoons, okay and so this automatically does it instead of people do it by hand. We call it a toon shader and now this is a toon shader that actually has reflections, notice underneath that you see that toon shader with reflections and gosh self shadowing. And all of that runs on a Mac and yet SolidWorks does not run on a Mac. Now there’s a couple of things that’s really interesting here.
He’s got three workstations, now if you’ve heard, if you were a workstation user, you would have one workstation now, all of a sudden you got these three workstations, you can have more, you get as many as you like. And so now the resource is on the network, if you’re the only person that needs it because workload doesn’t always balance across a company and the person who needs it most can get more of it, the resources are flexible okay. Huang?
Terrific thank you. Did you have – was there anything else?
Unidentified Company Representative
No, that’s it.
Okay. That’s terrific. Thank you very, very much. This week at NAB, we are GTC we launched VCA, introduced it to world. This week at NAB we’re meeting with all the partners. Already we’re signing up some of the most important partners for taking VCA to market. VCA is designed for companies that appreciate an appliance, now if you have workstations and racks and racks of them, if you incorporate NVIDIA Quadro in a remote workstation, and you have an IT department, you’re already served. However, if you’re a small, medium business and about 50% of the world’s users of Adobe, Autodesk, SolidWorks are small-medium businesses, you would really appreciate, it implies because your IT department is just not very large. They are served by all of these VAR’s around the world. As you know, Adobe takes the price of the VAR’s and Autodesk takes the price to VAR’s and SolidWorks takes the price to VAR’s. This is the number one VAR in the world for SolidWorks and they’ve been playing with the VCA and really doing real work on it every single day, and they are just so excited about it. It’s easy to use, it’s powerful in its flexibility, powerful in it’s access, powerful in how simple it is, and they really like it because it’s going to be simple to sell, the applications just work.
Now, the reason why GRID is so important to enterprise computing, the whole idea of GRID is so important because all of our enterprises are changing in real-time. All of the fundamental changes that you were seeing earlier in the slides is manifested right here in enterprise. My company, your company, they are all exactly the same, we’re impacted by the same issues in computing today. People want to bring their own devices. It’s a little bit like getting a company laptop is a little bit like giving a company car. It really is. By the time that you joined a company, when I first joined a company, I didn’t have a computer. I went to work and they gave me a computer. There are no – we don’t hire anybody who has never seen a computer before anymore. Everybody we hire already has a computer. You know what they would like, they would like you not to replace their beautiful computer that they selected themselves with your grey and black box. They would like to just use their computer.
All of your workforces are mobile now, you can afford to have just the capabilities when you’re sitting here. You want to have your capabilities wherever you go. The world is heterogenous now. There is no singular operating system that companies need to support. And so as a result it makes perfect sense to disassociate the computing from the data, to disassociate graphics from the device, from the computing, and because we’re moving around so much data, if you could centralize if you will put under network, digital computing you don’t have to move the data to everybody’s computer. You just do the data right there, you keep the data right there where the VCA appliance is, and you simply stream out to all of the end users. The net result of it, which is a very small amount of data relatively. If you can stream Netflix, if you can stream YouTube, you can surely stream the output of VCA, okay? The fundamental reason why GRID makes so much sense.
The way to think about GRID, I think of it as an appliance, we’ve taken just like they’ve taken storage out of the box. Just like they have taken networking out of the box. We’ve taken computer graphics out of the box. That’s essentially what we’ve done. We have to invent new technologies to make that possible, but that’s essentially what we have done. GRID is about $10 billion opportunity for us. Gaming, thinking about Netflix per games, there is 700 million cable subscribers, some percentage of those would subscribe to a game, some percentage of those would be using it concurrently, and based on that concurrency that’s the size of the server you need to build.
The quality of service matters. The business model matters here. And so this will take a little longer time. We’re engaged with all of the major players in the ecosystem. We’ve invented technology that helps them – makes possible for them to engage their customers and there is 25 trials around the world.
Eventually, this could be a very large business. Eventually, I can kind of imagine everybody’s cable channel; cable provider offers them games that they can play just as they could enjoy movies and video. VCA about a $3 billion opportunity. 10 million creative designers around the world using Adode or Autodesk or SolidWorks or Dassault, about 50% of them in small, medium businesses.
We launched that at GTC. We’re certifying ISVs, because Quadro is the most compatible platform in the world, NVIDIA software stack is the most compatible platform in the world, it’s all going to get certified. It’s just going to work. Now we’re signing up (inaudible) take VCA to market.
Enterprise VDI, 700 million enterprise workers, none of them today, very few of them are customers. However, if you aggregated their capability into the server, all of a sudden you need a whole lot of graphics capability. Although each and every person doesn't need very much, when you aggregated a hundred of them into one, then you need a lot.
What used to be a non-business opportunity for us turned into a business opportunity because of VDI. 160 million of those are knowledge workers, Citrix gave them remote capabilities, Citrix gave them accessibility capabilities. But Citrix deprived them, as a result, for many of them the performance capability they need.
VDI is a real growth opportunity. Dell, HP, IBM, Cisco are already certified partners of ours and we are doing some 75 trials around the world growing every single day. GRID is about a $10 billion opportunity.
When we talk about Tegra, the most important thing to me, smartphones is important, but the most important thing to me is Android. You guys know this very well, aside from the iPhone, the rest of the world really didn't ignite into this incredible growth until Android took off. Android has two characteristics about it that's really amazing. It's very, very unique. One, it is really the embodiment of the spirit of Linux. It's open. It's the freedom OS. It's the OS of the people. It embodies the spirit of Linux. Well it embodies the spirit of Linux, but in a guided sort of way, so it's not too chaotic because as you know, the markets they could be very chaotic. And so android is a bit like a guided Linux.
But the other capability that Android provides is, Android has all the services that Google provides, the valuable services. The day that you sign on to the Android, you have Chrome, you have YouTube and you have search, you have maps, you have all those Google Play apps, it's a rich ecosystem. It's a useful platform. The day you turn on an Android device, it's useful to you. Something that Linux was not able to do. And so on the one hand Linux, Android because it’s open, it attracted the energy of every single company in the world. There is not one company that I know today who has not at least at some level either working feverishly on Android or working on Android. Every computer company in the world, every device company in the world, I mean we’re talking car companies, we’re talking dishwasher companies and refrigerator companies and TV companies and computer companies, smartphone companies and tablet companies all over the world.
The united energy investment level around Android is what’s making it causing it to grow exponentially. It’s the same thing that happened back in the good old days around Windows. The number of people investing around Windows exceeded any other computing platform alone. The number of people investing around Android today exceeds any other platform by some very significant amount, because everybody feel they can build something with it. Android started out of the smartphone, but you guys knew I always cared that eventually all of that content that you owned from your smartphone was going to be available on the tablet, you take that logic, you extend it, it will be a clamshell, they will be on the TV.
And it’s exactly the reason why we felt somebody had to go build a gaming device around Android, so that you can enjoy all the content that you’ve aggregated. You’re going to collect your digital assets not for this device, but for your life. If you’re an Android user like I am, I am going to be using Android 15 years from now, not just the more books, more photographs, more music, more games. And I would like to be able to enjoy it on all kinds of different devices, it depends on what I’m enjoying if I’m enjoying magazine for the tablet.
If I want to do a little bit of work, work on a spreadsheet, or pull out my clamshell, you just have to turn it on, all the contents are there, that’s all connected, you can store it right into the cloud. So it's going to be available for everything else. It makes perfect sense that this computing platform or ecosystem is going to grow and grow nicely. That's what Tegra is all about. Tegra is more than mobile computing. Tegra is about the future of computing. You could see how we think about Tegra, Android Tegra, Linux Tegra, all kinds of computing devices.
Now, I'm going to focus – Phil is going to talk to you about many of the features of Tegra. I'm going to focus on just one thing. I'm going to focus on graphics and games. Now when I say one thing, and I said graphics and games that seems like two things. Well, it turns out that graphics and games, best graphics and best games go hand-in-hand.
The UI is a problem that's going to get solved. It will be plenty good enough. Everybody's UI is going to work great, everybody’s video is going to work great, everybody book flipping is going to work great. The one thing that's going to keep getting better and better and better and better and better, never get good enough is video games. The reason why NVIDIA's graphics is the best in the world is because we play video games great. When I started the conversation with you today the video games as it turns out is really very hard. I'm going to show you some examples today of how video games are just visually incredibly stunning and how it's about to go to the next level.
What turns out, video games is quite popular. This is an independent work done by they guys called Business Insider, and they said that tablets especially are gaming devices, about a third, about 40% of the people who use their smartphones, they spend about 40% in time playing games but about two thirds of the time spent on the tablets playing games, kind of make some sense, games are fun. If you ask now the game developers and this is a survey done at the Game Developer Conference, GDC happens every year, big deal.
Ask the game developers what they are spending their time doing. Well it makes perfect sense, where they are spending their time? 55% of them are building games for mobile devices, but yet only 13% of them are building games for game consoles. And the reason for that is pretty obvious, there is 1 billion gaming mobile devices that are capable of playing games, at any given point in time. There is going to be few million game consoles.
You could also see that the PC is way up there, and the reason for that is free play. The PC is an open platform, free-to-play or in game economics, in game business transactions, in game purchases, premium games, PC is the perfect platform for that, it’s open. That stands to reason why mobile. So mobile devices, mobile gaming it’s a real focus, consumers enjoy using it, and game developers are really focused on it. Well, just as we created GeForce, a graphics card that game developers could target and focus on, we had to develop a graphics card for Tegra, a graphics card that everybody can focus on and say, this is what GeForce is or this is what Tegra is, and I could develop games for it, and if I develop games for it, it will work on all of the other Tegra devices.
We gave game developers that idea some 18 years ago, it revolutionized our business. GeForce is the platform that game developers develop on. So of course it would be the platform that it works best on makes perfect sense. It was developed on GeForce, it would work on GeForce. We need all of the game developers around the world, there are so many of them, they have a singular platform to go develop on.
A singular platform we call SHIELD and that’s why Project SHIELD was created. On the Extremes, I see SHIELD and it’s opportunity, on the one side it’s a device for the most enthusiast of Android gamers, the enthusiasts, you just love playing Android games. Two on the other extreme because mobile gaming, because open platforms, because free-to-play these characteristics, these fundamental trends it’s not a fad, it’s the future.
SHIELD becomes the ultimate Android gaming device. So from one extreme of enthusiast to the other extreme it goes Mainstream. I’m not exactly sure where it is, we’re going to find out and we’re going to start this journey with you guys together.
We’re going into production soon and hopefully by the end of Q2, we will put it on the shelf. The strategy of SHIELD is summarized simply here. Now that we have a gaming platform that people can focus on, they can build great games for Tegra. Those great games are deposited in a store called TegraZone. TegraZone is a curated store for free Android devices, free Android games, and some are priced modestly, some are premium end game purchases, but TegraZone is curated. Every game that goes into SHIELD ends up in TegraZone and every single Tegra device is compatible with TegraZone. And as a result, every device that uses Tegra has the benefit of this capability. Over time, the compounded effect of all of this will just make Tegra more and more and more valuable.
The industry is changing dramatically. We invented programmable shaders 12 years ago. And today, it defines the state-of-the-art for mobile devices, tablets and phones, programmable shaders, ES 2.0, Programmable Pixel Shaders. About five years ago, six years ago, we invented the idea of GPU computing, GPGPU, invented this new idea called geometry shading. No longer just writing a program for a pixel, we are now writing programs for geometries, for objects. And there is so many things that we do now as a result of the geometry shader, and we’ll show you many of it today.
The complexity difference between the two are simply incomparable. The complexity difference is so great. One of them we used for basic graphics, the other one we now use to power our nation’s fastest supercomputer. One of them is a, if you will, a microcontroller that controls small programs for pixels. the other one is the general purpose processor where architectural efficiency matters everything, a big discontinuity.
finally this year, the game consoles are moving to DX11 finally this year. the overwhelming amount of content that’s about to come for DX11 is pretty significant and we’ve been racing, we’ve been racing to mobilize, to shrink.
The GPU that goes into our nation’s fastest supercomputer into a little, little, tiny thing, and we think the payback is going to be huge, and we’re still excited about it. We decided to re-prioritize it along with Tegra 4i for a massive pulling. we’ve challenged the engineering team to bring us DX11 for a Tegra form factor as soon as possible.
Today, we’re going to show you some of that. This is the platform that they brought it upon. this is not a product, this is a bring-up system; the smallest GPU NVIDIA has ever built that is state-of-the-art. this is fully compatible with our nation’s fastest supercomputer; there were just run very, very slow. But it does it in just hundreds of milliwatts, instead of thousands of kilowatts.
This is Kepler mobile. and what this allows us to do is, if we can go back to the slides please. What this allows us to do. This graphics for sure is tiny. This is the tiniest graphics cards we have ever made. This allows us to create imagery that is really stunning. This is the state-of-the-art of computer graphics. We want to make it possible on every single tablet, because the content developers are working on it right now, for next generation game consoles. For the first time the PC, the next generation game console and now mobile devices will use exactly the same technology DX11 and OpenGL 4.3.
We have an opportunity to make that happen, to intercept this big Tsunami wave. We needed to get a development system into the hands of all our developers as quickly as possible. We would assure you some of these imagery running in real time in just a moment, but here is the challenge.
Fermi was very capable, but also rather high. Kepler’s huge innovation was to shift that computing efficiency, which is measured in purple white over by about a decade a factor 10, reducing it into something that is now the best GPU in the world. However, it’s still a 100 watt, 50 watts, tens of watts, and we needed to shrink down to hundreds of milliwatts. And so we challenged the engineering team to create Kepler.M which is right here. Kepler.M, so that we can bring that capability down into the form factor necessary for TV set-top boxes, little pods to tablets to clamshells without fans to smartphones. And this is the world’s first event.
Now when you measure that, and we'll show you some examples next time of measuring that against the competition, the difference is really quite staggering. It makes perfect sense. This is our fourth generation DX11 class GPU. This is our fourth generation GPGPU, and the rest of the world hasn't built the first one yet.
We want to hurry up and bring that to the world and get four generations ahead. We want as soon as possible to get multiple years ahead of that competition and this was such a great endeavor, we thought it was worth the sacrifice, this is the second of the strategic decisions we made and I'm so incredibly excited by it.
So now I'm going to show to you, what we have done here is this, the next generation chip we're building is called Logan. What I've done is I've taken this put into a PC, this is a mobile GPU by the way. This is meant to be an Android device. We've taken this and we stuck it into a PC, literally just plugged it in. When we plug it in all the software just worked, OpenGL worked, Windows works, everything works, and we took the CPU, we took the lowest, we took a CPU that we could clock down to effectively the same performance, effectively spec in of Logan.
Okay? So what we want to do is, we want to model for you what Logan will do, but before we do that let me show you what state-of-the-art mobile graphics look like, just to remind you. Lars, can we do that real quick? Lars, not this one, now yeah okay all right.
Okay. So what you’re looking at here is the state-of-the-art and tablet graphics this happens to be quite a high-end tablet, this is a latest generation iPad, 40% of the A6X, 40% of the silicon is GPUs, a whole bunch of GPUs in here.
Okay. So this is an iPad pretty good looking graphics, all right, why don’t we now change it over to Kepler.
What do you guys think? A little bit different. Okay. So guys dynamic lighting of course particle systems, the rich geometry, the high dynamic range lighting, all the shadows, the active shadows now it is instructed to go back. Now you look at this and let’s go back one more time please.
This is mobile Kepler and this is a state-of-the-art today’s graphics. This is basically vintage. I would say vintage 1999, and Vintage would be a good word for this. Very beautiful, very fun apparently two thirds of the people who are using tablets enjoy this and we love to be able to bring them to next level.
Okay. All right. Thank you very much. Thanks Lars, good job. Just back to the slides real quick, we are now seeding developers, we are seeding developers with this new platform, it’s called KAYLA. KAYLA is the world’s first arm Tegra with DX11 Kepler, years ahead of the competition. And now we are developing content for it, and then surely soon I hope and Logan will be able to benefit from all the content that’s being created for this and go-to-market with really exciting content.
Okay, the Tegra TAM, let’s talk about the TAM. We’re going to come back and talk to you about Tegra but the way we see our market opportunity for Tegra is number 1 Android devices. Android devices as phones, tablets, you’re going to see TVs, you’re going to see PCs, you’re going to see gaming devices. It’s growing incredibly fast, as you know tablets from literally nowhere where we thought Android tablets made no sense to now in a few years. There will be half a billion tablets sold, larger than PC’s. Says something about the fact that tablets are becoming or mainstream computing is now possible in most people’s tablets.
LTE, the market is quite large, is going to double more than double by 2016 from this year about 2.5 times about 450 million LTE devices. WinRT, the way we think about WinRT is this, in a few more years time, call it two, call it three, the difference between a PC and a WinRT PC is indistinguishable. It makes no difference whatsoever, it’s just a PC. If they saw one problem, it will be exactly the same. That one problem of course is Outlook. If Outlook was available on both of these devices WinRT, Windows 8, it doesn’t really matter, it’s just the PC. We are going to be buying applications from the App Store anyways, backwards comparability most of us don’t have those backwards comparability issues with the exception of Office, which includes Outlook. And so I’m holding my hopes up, and hopefully that gets result. And one of these days therefore the WinRT opportunity is as big as the PC opportunity.
Automotive; we are already designed into seven different car companies and 34 models of cars, and they are in the process of either going to production or finishing. We have about $2 billion worth of design wins. It takes seven years from the moment that you engage that market to the moment that you are shipping the first one, we now have $2 billion worth of design wins un-shipped. We were based on our projections, looks like we are going to peak with the current design wins in FY16 at about $450 million a year. Auto alone for Tegra is a real business.
Now if it makes sense for your phone to have such great technology, it makes no sense that when you enter your car that your car is an analog radio. Everybody now knows that having a connected car increases the ARPU for the car companies. The customer satisfaction is higher, the number of applications and services they could deploy to that end-user is higher and everybody is learning from the Tesla electric car, which has two Tegra’s inside by the way. Okay, so the Tegra TAM quite large.
And that kind of summarizes my talk with you. We talked about really several things, one, we talked about the fact that our investment in GPU computing has made it possible for the GPU to continue to grow. It’s growing at 12% CAGR. Number two; we’ve taken that GPU and disassociated it, disconnected it, raised it out of the box through remoting and virtualization inventions and created this new category called GRID.
Now you can have a virtual, a visual computing appliance, like a network appliance, like a router easy for you to connect into your workgroup. And then three; recognizing that in the future Android is going to be more important than ever that computing devices are going to be literally everywhere. So important, so disruptive that it may come back and disrupt the PC all together. It is important for us to be part of that disruption. Both because it is essential for us to protect our business, but also so that we can grow into this enormous opportunity. We can make a real contribution.
What makes all those three things exactly the same to us is visual computing. We leverage one singular capability, one that we are world class at, passionate at, and one we are the world’s best thinkers in computer graphics come work at it. Three opportunities all together representing about $26 billion worth of opportunity today. We also saw the way we invest highly leveraged and in a basic investment platform, core platform about $900 million with an incremental $300 million or so that makes it possible for us to engage this $20 billion worth of additional market opportunity.
Over the last four years, three years, we’ve grown at a CAGR of about 20% and recently, we committed to a dividend for the first time in our company. Today, I’m delighted to announce that we’re going to take – we’ve been evaluating our capital allocation policy and because of the developments and the traction and the progress of these strategies, we’re confident in our future cash generation capability. And the Board of Directors has approved that we’ll return $1.2 billion in total through fourth quarter of this year. And so let me tell you how it’s all broken down.
We’ve already announced a $0.075 per share dividend. That represents approximately $50 million a quarter. We started that fourth quarter last year. In addition to that, we’ve been repurchasing about $100 million worth of stock. We did that Q4 of last year. What we’re going to do today is we’re announcing a commitment to repurchase $750 million and in addition to the dividend that we had already previously announced, all of it in combination through the end of this fiscal year represents in these five quarters, $1.2 billion.
And so that’s that. That’s my presentation, NVIDIA, the visual computing company. Over the last several years, you’ve heard a variety of announcements from us. We’ve been in our mind building this out, an end-to-end visual computing company, the world's most important computer graphics company. We want to be able to serve computer graphics, visual computing opportunities from fundamental invention through processors, through modules that people could include in their systems, to systems of our own, end-to-end visual computing company servicing – serving a $26 billion opportunity that we look at today.
Thank you very much for your attention and we'll be around all day. All right, thanks.
Okay, thank you, Jen-Hsun. Let's move to the next part of the agenda. I'd like to introduce Jeff Fisher, he is our Senior Vice President of GPU, all right?
Jeffrey D. Fisher
GeForcea and GPUs
All right. Thanks a lot.
Jeffrey D. Fisher
Thanks, Rob. I've met some of you last night, some of you over the last several years, I've been at NVIDIA for roughly about 19 years now in some capacity always working on the GeForce Business, I think it's kind of hard for people to remember my title or what specifically I do because I've been involved in various different aspects of this business. But I am pleased to be here today and give you an update on the GeForce business.
Last year we introduced the world to – the PC World to Kepler. Kepler set a new bar for performance, power and ultimately acoustics. And across the globe it got great rave reviews. I’d hope that that was one of my demos. Everybody get up and dance, sorry about this. Am I supposed to be doing some thing up here? Thank you. I am not that good of a dancer or entertainer or singer. You have to do that on your own.
Last year, we rolled out Kepler. It was an amazing product. It got rave reviews and it’s really drove a lot of our growth last year. By the end of the year last year, Kepler had been – was offered from $99 up, but we weren’t done with that. On February, we launched a brand new Kepler device, [the crown] towards the GeForce GTX family, GeForce GTX Titan. As the world’s fastest GPU, Titan set a new performance standard for PC gaming. It also set a new acoustic standard for a high-end GPU. Titan is priced at $1000. It’s available globally and every unit that’s hit the shelf so far has sold out. The demand for Titan is incredibly high.
The perf-per-watt of Kepler that we talked about last year and throughout all of the presentations on Kepler, perf-per-watt really shines in a mobile environment, in the notebook environment. This year we’re rolling out our second generation Kepler notebook devices. The GeForce 700 M series the perf-per-watt, the performance of GeForce 700 M leaves us well position, we believe, relative to Intel integrated graphics as well as the competition.
The device you see here is a new Asus Ultrabook, touch based Ultrabook featuring GeForce 700M graphics. In addition, several years ago we rolled out a feature called NVIDIA Optimus. Optimus is designed to enable GPUs to go in low TDP in thin-and-light notebooks. Optimus effectively scales up the performance of the GPU when you launch a demanding app calling on the GPU. When you are in productivity mode, running Outlook or Excel or other non-demanding apps, the GPU effectively goes to sleep. This allows you to have the best of both worlds, highest performance when you need it and longest battery life when you are on airplane doing productivity apps.
Optimus is now a standard feature in GeForce notebooks, roughly 99% of notebooks with GeForce feature Optimus. Based on the perf-per-watt advantages of Kepler, our market share gains late last year in notebook are differentiated features like Optimus. I believe this year is going to be a good year for notebook for GeForce. We should be able to outperform the overall notebook market this year with GeForce GPUs.
But I want to talk about PC gaming today. PC gaming market is very robust. Historically AAA games migrated from console to PC. PC gamers were basically able – taking what they can get off of console. But with the growth of PC performance, the games and PC performance, the PC as an open platform, the PC as upgradable, the migration of DX8 to DX9 to DX10 to DX11 offering new visual capabilities for the PC, the PC has become the gaming platform of choice. Today AAA titles are launched, virtually all of them are launched concurrently on PC and console. In fact the top MMO, the top first person shooter, the top real-time strategy game, the top free-to-play game are all played on PCs. Steam has roughly 2000 PC games available for sale for PC. The best console has roughly half that, about 1000 games available on Xbox 360. PC is clearly the gaming platform of choice and it's growing.
Last year was a good year for PC titles. There were a number of AAA games that came out that [pushed out] and really drove the PC market, both software and hardware. This year is going to be no exception. In fact, with the tailwind of the consoles, of the new console refresh later this year, I would expect this year to be even a better year for PC games. There is a parade of AAA content coming out for both PC and console, including new Battle Field, Assassin's Creed, Far Cry, we've already seen Crysis 3, we've seen BioShock. These are all games that are not just AAA titles, but they push GPU tech and thus helps sales of GPUs.
One of the titles we have shown here, Metro Last Light is a game that really pulls together all the tech. Metro Last Light is going to ship soon in the next few months. But you get to see it here today as one of the first folks being able to experience Metro, it’s being demoed on this station over here during the break. Metro brings together all the best of visual effects. It has got PhysX particles, PhysX disruption, PhysX turbulence, you’ll see ambient occlusion, you’ll see HDR lighting, you’ll see depth of field, motion blur, everything pulled together on this 4K display playing Metro Last Light. Take a look at what the current state of PC gaming is over on the right. But Metro Last Light it is just one example of the games that are going to be coming out in the next 12 months really driving, driving this year.
Also I expect over 50 games to launch this year that will be pushing GeForce GPU tech. In total sales that would represent about 70 million units sold over the life time for these games. That’s compared to about 30 last year. But what’s really exciting about PC market is free-to- play. Free-to-play is driving a new business model for PC gaming, Free-to-play really came about as a result of piracy. It was a way to get people in the games and get them then to decide to upgrade and in game [pay per] in game items. But it’s really taken over as a brand new business model and has had explosive growth especially in some of the emerging markets Eastern Europe and China and Asia, but it is also really grabbing hold in Western Europe and U.S.
The number one video game in the world is a free-to-play game, League of Legends. League of Legends has roughly 3 million concurrent players at one time, all online playing League of Legends. There is roughly 1 billion hours of gameplay a month in League of Legends. Halo, one of the top titles for Xbox 360, since its launch in 2004 has had 2 billion hours of gameplay. League of Legends is 1 billion a month and growing. This type of dynamic is really driving free-to-play game.
Here is hundreds of free-to-play titles now available and that number is growing. As more gamers come to free-to-play gaming, we have 200 million and growing free-to-play gamers out there. As more gamers come and as more money comes, so comes tech. Developers want an opportunity to differentiate their titles from other titles. And as with AAA games that is now coming in the free-to play-space. And we expect that to accelerate.
Some examples are there of that World of Tanks. Is there anybody in the room who has played World of Tanks? Oh, sweet. Phil, you don’t count. You are an employee. World of Tanks is a game. Enter World of Tanks, it’s a ton of fun. You can play it on many virtually any PC. But if you want to turn on the terrain, if you want to turn on the smoke, if you want to turn on the explosions, if you want the visual experience, that immersive experience you need GeForce GTX class product to turn on, not everything, but most things, most of the visual cues in World of Tanks. And I don’t know what that you few players out there, but I happen to like the hide in the bushes. You can live a little longer and you can snipe a bunch of other tanks and you really want that bush to look like bush, not just a square block that you’re sitting inside of. So I’m happy I’ve got a GeForce GTX in my PC at home.
World of Tanks has got 40 million, 50 million users. It’s a huge game, it’s very profitable and there is new versions of the World of War gaming is going to be coming out with over the next several years. PlanetSide 2; PlanetSide 2 is a relatively new title, features GPU PhysX, particle effects, explosion, smoke. For more realism, it’s a first person shooter, multiplayer game, free-to-play. Hawken, Hawken is a game that’s in open beta right now. This is really cool. Hawken and we’re going to be demoing over here on another 4K display. You’ve got to check it out.
We’ve shown you guys destruction over the last several years. Every time we get together, we blow something up on stage. I’m not blowing anything up on stage today. But we’re blowing it up over there. And Hawken is the first demo, the first game demo that actually includes destruction in the gameplay. You’re not blowing things up for the sake of seeing them blow up, but you blow them up to actually change the playing field.
You can bust through a wall and create a new pathway in between rooms. You’ve not seen that in a game before. You can explode – blow a hole in the floor and jump through the floor and create a new pathway down to a lower level. These are the new types of effects that are possible in not just any game, a AAA game, but now you’re seeing them, you’re seeing the tech coming to AAA. So make sure you check out Hawken as well.
With the progression of our GPU roadmap, more and more horsepower being brought to bear on the PC gaming market. We are raising the ceiling on what’s possible in terms of visual effects. That’s really important. It isn’t just the GPU that is raising a ceiling, but it takes a dedicated team, a passionate team to work with game developers and invent new ways to bring visual realism to games. You need both a roadmap that’s driving tech and you need content that’s driving visual realism to keep this market robust.
We have dev tech team inside of NVIDIA that is several hundred engineers dispersed globally. Some of the best visual engineers in the world whose mission in life is to develop algorithms, develop libraries, develop effects for next generation games to continue to push visual realism. These effects cover a number of different visual areas such as hair, destruction, smoke, particles, clothing, ambient occlusion, different lighting effects, TXAA, cinematic anti-aliasing, overrun this system back here. We are going to be show – we are showing water, it’s a library of water, realistic mind blowing water that is a self contained library. You can drop it in to any game that where you want to show water effects. And take a look at the current state of water works in the back when you have a break.
One of the toughest challenges, however, in real time game rendering is faces. To render a face and really suspend belief, make it look like it's a real person, you have to recreate not just the lighting effects, the eyes, how the light moves through the different parts of the face, the curvature of the face, the gestures, everything has got to look perfect as to mimic reality to pull you into the game. That’s why when you see games the people generally, the heroes are generally masked iron clad characters. They don't look like real people. The villins, now they are the monsters, but you don't see realistic faces and human bodies inside of most games and the real reason for that is because it's really hard to suspend belief.
We’ve have been working with Dr. Paul Debevec, who is the Associate Director of Graphics at the Institute For Creative Technology at USC to develop this demo. Dr. Debevec's team has created a motion capture technology called Light Stage. Light Stage effectively can capture all of the gestures and the motion of the human face, how it's lit and integrated into a 3D model.
Ladies and gentlemen this is Ira, as a result of that effort. Ira, we've got a room of investors here today that want to here the NVIDIA story, yes. Some of them are sell-side, but many of them are buy-side, you got it. So let's take a look at some of the effects, the face work effects that bring Ira to life.
Let’s go the wireframe. So this is Ira facing and one of the important aspects of how his face is constructed through the wireframe or the curves in the face. When you zoom in on different characters and games often you will see kind of a blocky or a triangular silhouette. As you zoom in the Ira you will see more complexity, more geometry is automatically added to smooth out every aspect to this face. This is called tessellation. Zoom in you get more complexity, you get more details, zoom out and it’s not as necessary. The GPU does this automatically through tessellation. So that’s the outline, the outline of the face.
Now let’s add some skin Ira skin back on. Skin has a unique reflective property that is unlike anything else. Light is absorbed by skin and then reflected in a defused way and that’s what makes skin look real. It’s called subsurface scattering.
Let’s take a look at what Ira looks with sub surface scattering off. It’s looking a little pasty and a little concrete like light which is bouncing off its face it’s not something that you would accept as real. Let’s turn it back on and take a look. Take a look at Ira’s ears for example, you can now see the light diffusing through his ears. The skin is now has a soft feel to it because of sub surface scattering. Let’s take a look at Ira’s eyes. Ira’s eyes in this demo are fully ray traced. Ray phrasing creates a almost exact model of how lighting would bounce off of every element inside of an object. Eyes are really important to get that reflective quality. The light rays bounce off, go through the cornea and bounce of the iris in this demo. You can almost see into the eye, you see the specular lighting, you see the moistness in the eyes. That’s all due to ray tracing inside Ira’s eyes, let’s back away a bit.
So Ira’s, based on Jen-Hsun key note, what do you think about running out and buying a million in video shares.
Okay. Ladies and gentlemen, that’s Ira face works.
Okay, we talked about PC Gaming, we talked about Visual effects. Every year we work with – over the last three years say, we’ve worked with Epic. Epic is one of the top producers of game engines and games in the PC industry. So last three years, we’ve worked with Epic to put together a state of – what’s possible demo for the game developer conference.
Two years ago, it was a demo called Samaritan. Last year it was Elemental. This year, there is a brand new demo that they have pulled together called Infiltrator. An Infiltrator did not disappoint. Let’s take a look at Infiltrator.
So Infiltrator pulls together all the effects we’ve been talking about into a single demo. Really raising the bar for what’s possible for everybody else. And as I mentioned, once again this year they have done that. Jim, let’s go to wireframe mode.
One of the most interesting things about Infiltrator is when you’re watching the video, when you’re watching the game, when you’re watching the demo; you really think that you’re seeing a moving running. A movie that could take days or weeks to render frames, could take months to put together, but this in fact as you can see is not a movie, this is a real time 3D demo. Because of wireframe, the complexity inside of Infiltrator this is up to 16 million triangles per frame. 16 million triangles per frame, the top end game this year will be roughly 1 million to 2 million triangles per frame, so the future of gaming will continue to push tack.
Let’s paint up Infiltrator, add some diffuse textures to it. Several years ago this was roughly the state of PC gaming, way back in the day, when I was younger. Let’s run a few frames here, just a few textures, but with computational graphics as Jen-Hsun had talked about earlier, we can now fully light this environment and add realism to it, particles high dynamic range lighting, and now run this demo in real-time. Jen?
Once you take a moment and talk about some of the different elements of Infiltrator that are pulled together into this entire seen.
Unidentified Company Representative
Thanks, Jeff. So as Jeff said, this is all running in real-time, and we just want to highlight the complexity, and what goes into actually generating a scene like this. So we’ve seen the sensors the final rendering and then all of these boxes that you see around the edges are basically various components, they go into generating those final pixel rates. So these are basically, it’s highlighting a technic called deferred lighting, deferred shading, and what that is, all these different components including that underling diffuse color which is basically the color of the object, the specular color shiny it is, and also surface parameter like the normal which is the orientation of the surface of the depth which is used to compute the effect like depth of field, in focus and out focus, all those are being computed in real time for every pixel and then composite it using complex mathematical formulations of lighting and lens effects and things like that to give you that final rendering in the centre, so per pixel and Infiltrator we were computing about 6000 to 7000 operations per pixel at (inaudible) that’s about 15 billion, almost 15 billion operations per frame and at the frame rate this is running at that’s over 500 billion operations per second.
Unidentified Company Representative
Okay. That’s great Jim. Right there is 60 million triangles that setting. Let me point out Jim has been NVIDIA employee for about 10 years, he leads the team that has been working with Epic in all these demos and he has done a fabulous job of delivering this to market. Thanks a lot. So let me also point out this demo is being run on a single Titan in the back, the face demo, the Ira demo, that demo requires about 8,000 operations per pixel in HD resolution at 60 frames per second like you’re seeing today. That’s roughly five teraflop per second and that demo is being run on two Titans in the back. So there is still a ton of headroom to get to anything close to reality in the PC gaming market which is really exciting for us both on a content side and on the GPU side to keep driving the state of visual realism inside of PC gaming.
So let me talk a little bit about the business. Our roadmap fuels our business, compelling content also fuels our business. We’ve talked about the growth in PC gaming market, we’ve talked about the new dynamics and free-to-play. We’ve talked about tech, and the continuing evolution of tech pushing GPU sales.
Jen-Hsun had shown earlier how we are working to split out our GeForce business between gamer and OEM. The GeForce gamer business has grown 9% to 10% annually over the last several years. On the strength of the content, the strength of our roadmap, I expect growth for the years to come. We should be able to outperform the PC market, and GeForce gaming has also highly leveraged into the economics of our business.
OEM represents roughly 80% of the units. Market share driven by OEM units. However, it’s less than 50% of revenue. GeForce gaming drives over 50% of the revenue. And GeForce gaming drives almost 75% of the product margin inside of the business. So while there is significant headwinds in the PC market overall, we believe the PC Gaming market is robust. We think our roadmap, we have a leadership roadmap. We know there is compelling context coming. And this business is highly levered inside of the G4S business. So this breakout I think is really the time that we are trying to show you the leverage inside of the G4S business and the leverage for gaming in our business overall.
Robust PC gaming market, strong leadership, roadmap, passion to continue to driving tech inside of games and a commitment to do so. Personally after 19 years, I am still as excited as ever about the GeForce business and certainly the PC gaming market and appreciate your attention today. Thanks so much.
Unidentified Company Representative
Okay. Hello, all right, thanks a lot of Jeff Fisher. Let’s move on to Tegra at this time. I’d like to introduce our Senior Vice President for the Tegra mobile business unit, Mr. Phil Carmack.
Philip J. Carmack
Can you hear me, great? We are not allowed to call him Fish, is that a new rule or something. After 19 years, I thought we can call many thing we want. Well, like Rob mentioned, I am Phil Carmack, I am delighted to be here with you today and talk a little bit about Tegra. A lot of exciting things going on, Jen-Hsun did a good job of introducing where we are. And I think having you see about the PC graphics business or about the GeForce business prior to me is perfect, because you get a sense of where we are going into the future as Jen-Hsun mentioned some of the exciting things we have coming. But let me start by talking about what we’re doing this year and some of the important things that I like to share with you.
First of all, I think Jen-Hsun framed this pretty well, but the Tegra business is substantial. $723 million last year of total revenue, good margins, represented a 109% combined annual growth rate over the previous three years. Of course, we needed to be driving toward a much bigger market. I think everybody understands that and we know what we needed to do to get there. And you can see in the decisions we are making, the focus and vision we have for where we are going to go with Tegra and how relevant and important it is to the industry as a whole and particularly through NVIDIA us making that investment to be able to be at a point where we can deliver a family of products, the portfolio of products together was a very important strategic decision we made based on some of the things we saw that’s how our business was working.
And I will talk a little bit about that cyclicality that Jen-Hsun pointed out earlier in a little more depth. So Tegra 4 is a family now with Tegra 4, Tegra 4i and a standalone i500 modem. Also visual computing matters, visual computing matters in mobile, not just in the markets where we have these ultra-high-end graphics solutions like the one tera-ops needed to render iray in a high end machine today, there is every expectation that we need to be able to render iray some day in a mobile device of course. We expect that capability – every capability that's ever been built for prior generation computing appliances or machines has been brought to mobile eventually and it really is a question of how fast you can get there, not whether people want it and when it comes to gaming, 76% of android app revenue is in the most recent report is games.
Furthermore there is much more significant and broad uses for computing in mobile like high dynamic range photography. This is a phenomena that's gone from zero to 106 million downloads, HDR photography applications on mobile, clearly something people want but requires competition, visual competition. And the last thing that I'm going to explain to you today, with help from my friend Stephen Alvarez, the CTO of our modem team is about our new LTE modems, we have two of them.
A year ago, we were still telling you we're working on it, now we have implemented in two different ways, one in a integrated single chip SSC that is highly efficient because it has the modem and the processor and everything else all in a single chip and then the higher end version which has a stand-alone modem and a stand-alone application processor together, Tegra 4 and i500. And they're working and they're making good progress and the certification is – well there is a lot of work to get the certification but we know what we need to do, we've been there before and we'll explain to you how we're going to get there this time and how long it's going to take.
When we think about the $10 billion business that we’re going after starting from the $750 million business today, where does that come from? Well maybe it would be easy for me to start by saying where it doesn’t come from? We talk about this $10 billion TAM out in the future 2016, 2017 that’s available for us, we call this the Tegra TAM, we’re not calling this the overall TAM. If I told you the smartphone and tablet SoC TAM that number would be a lot bigger wouldn’t it.
This represents the non-integrated players, the non-vertical integrated players like this takes out, this excludes Apple, this excludes Samsung in the TAM for both tablets and smartphones in the future. It assumes the growth that the analysts are predicting. It doesn’t assume that we get into the $100 smartphone segment either or sometimes called the Value segment, the extreme low end.
What it does assume is that because we brought online, Tegra 4i, we are able to begin to move into the more mainstream part of the smartphone industry which is about three times the size of the Premium segment we’ve been playing in and together those two things mostly for Android add up to about $6.5 billion of that and the tablet market is projected to grow to about $3.5 billion again not including Samsung or Apple.
A couple of really interesting things are happening to make this important and accessible to us, and it’s both the things we are doing and the things that are happening in the industry. One of the things that many of you’ve seen is the advent of these white box tablets for example especially in China where non-branded tablets are being sold at a relatively low price. They have been growing and then they are starting to tail off and part of the reason they’re tailing off is because of the experience, struggling to deliver everything you want. Meanwhile, there is these mildly subsidized tablets like Amazon’s Kindle or Nexus 7 from Google which we’re in, all of those things together are making us, so people expect a lot more for their money.
And white box doesn’t mean that people just want something with a number that says 159 on it. What it means is that people are looking for value and so they’re looking to these alternatives to see if there’s something valuable there. And by the way, we saw this happen in the PC industry, didn’t we? I think NVIDIA; one of the great things about NVIDIA was how we leveraged the system builder and white box, parts of the industry back when we started building the GPU business into something big like it is today.
And what it allowed us to do was deliver more value for the money that people pay. Their consumer cares about that and it’s indicated by the fact that they’re trying out these new ways of getting value on their mobile computing devices. Jen-Hsun had this graph in one of his slides. By the way, there’s a typo there. That’s units in millions. But this is the quarterly Tegra business back from starting before Tegra 2 started shipping up until now.
As you can see, the business is generally growing. But it’s also, it steps up on the successive generations of products and the rhythm of those products has been approximately a year with Tegra 1, Tegra 2 and Tegra 3. And each generation, we’ve captured more opportunity. We’ve created more capabilities that have caused us to get more design wins and more interesting design wins over time. But there’s also cyclicality to it, and this is the thing we decided we wanted to break. We wanted to make sure we could execute at a faster rhythm and a more parallel execution. So we now are with Tegra 4. we have two products in our portfolio, coming out almost simultaneously, very close together. And shortly, thereafter as Jen-Hsun has already mentioned, Logan, which is the next product after Tegra 4.
When I think about what that does for us today or where we have to go through a period of being relatively flat, as we catch up to the one quarter bubble we inserted it with Tegra 4. we’re seeing great up tick on Tegra 4. it’s got every reason to be highly competitive than it is.
But we see the ability to rapidly move to the next stage. And then as you look back at our $723 million last year, the areas we played in like premium tablets phones and normally subsidized tablets and white box tablets. We have a more competitive position going forward, more ability to attack in those markets while also being able to attack completely in new markets.
And so there’s a couple of things that contribute to a real competitive discontinuity for us. if you look backward, Tegra 3, Tegra 2, we had great tag lines like Dual Core, Quad Core CPUs, which I think got people excited, but we also delivered the best experience and part of that was the visual computing we’re able to deliver, that affected everything from house not be (inaudible) to the TegraZone games you might be able to get, and a bunch of other things and that created some success for us.
What we didn’t have was LTE and as Jen-Hsun mentioned, LTE is come on faster than we’ve been expected, the transition has been rapid and you can see why people have an insensible desire for bandwidth and whether that’s wired or wireless people have that desire and so if you create a solution that offers it people are going to want it and that’s a leverage point having LTE is a leverage point. We didn’t have it last year. We didn’t have a portfolio last year we had one answer that allowed us to target one segment of the market once per year. Now if you look at where we are going this year, all of the things on the right are beginning to come to the market during this year.
Graphics, we already have a solid leadership but when you think about log-in that leadership is a four or five year advantage putting 40% of your SOC into the GPU does not create what’s it called infiltrator does not create infiltrator, right. More transistors alone does not create infiltrator what creates that is the GeForce platform. Brining the GeForce platform and particularly Kepler into mobile is going to radically change our competitive position in the market.
Furthermore, we are – and I’m going to talk a little bit more about this but on the architectural front aggressively pursuing niche approaches to create better performance and efficiency in the classical CPU processing and of course we also have LTE it gives us the tailwind of having leverage of our own LTE solution, which is already been proven in AT&T using our previous generation modem, but now with the new generation modems will allow us to have strong LTE solutions in every category that we play in. And having the portfolio allows us to bring these premium features into more mainstream categories. The family Integra 4 or Integra 8, basically leverages the architectural capabilities, Tegra 4 into two products that both have all those capabilities. So when I talk to you later about computation of photography features like HD, high dynamic range photography or other things that I’m going to show them or later. Whatever I can do on Tegra 4, I can do on Tegra 4i, it’s just in a smaller footprint.
And what you find is that, we’ve modeled devices or any device for that mater. The experience you can deliver entirely depends on the performance you can fit into that form factor and this pro forma the optimal way to do that is Tegra 4i because it’s smaller and therefore and focused on the powered envelope of these zero to three watts where as a tablet you need 3 to 10 watt per envelope. In the other words, that’s the maximum you might actually use during that time. And the amount of compute, I can do in five lots if I have equally efficient processors is more than the amount of compute I can do in two lots. So I have to build specialized products to give you the optimal experience in each of those segments.
With Tegra 4i, I does for you is because of the efficiency, it gives you the performance of the most premium, high end product from our competition that’s coming out. He has 800, hasn’t started shipping yet but there has been publicly announcements made about it. Tegra 4i is about half the size in terms of the size of the die. Because of its efficiency is able to outperform as 800 and most of the benchmarks. Meanwhile, because it’s a lot smaller can be at the not the next tier of below it, but tiers below that, right into the mainstream LTE segment. The lowest cost quality LTE product you could buy today is the AD930 or the S400 and in the future there is a follow on to that. Both for the AD9030 and for the follow on to that Tegra 4I can be sold at the same price. You can build the same level of phone the two tiers down from what you can build and a net save 100. But at the same performance of the SA 100. So the same prices below is the and low end or which I’m calling the mainstream, this is not the value segment. The $700 is the mainstream high volume LTE market, which is three times the size of the premium market.
Tegra 4i nails in for this and this is the value product of the performance discount nearly that I’m talking about. Why is the value so important? Well, there is a lot of indicators they are showing, the people are demanding more and more premium features and capabilities for their money. This is a just a chart taken from our recent survey done in China. One of the top most important things to you contributing to your purchase of a smartphone, not surprisingly number one was price. But look at number two, three, four, five – the very next thing right after price is performance that are screened that are camera capability. So 76% of the people said price is the number one thing, but 68% of the people said performance was the number one thing or performance say was key thing that contributed. What this means and there’s a lot of other evidence to the effect is that, that over time people come to expect to more and more capability for their money. Another great example of that is today the average revenue per user for data, for mobile data is about $65. In other words, the total amount of data that people purchase on average, this is at Verizon, is $65.
Many times, this represents more than one SIM card or more than one device. One person spends $65. 10 years ago, when there was about 400 times fewer data users, only elite users, people like you guys who traveled and had to connected, right, you spent $60 so you could have a data plan with 30 time less data than what today’s data plan has so that you could be connected when you needed to be.
It was considered a very elite premium experience. Now people are spending, 18 years olds are spending that same amount of money in order to get access to 30 times the amount of data. They want the experience. They’re willing to allocate money to it. They’re not in the same category as that person 10 years ago, that 1% who could have spent even more to get more data. But they have a high expectation in terms of what they want.
And this is why Tegra 4i is so important. It gives you experientially the ultimate in performance, but fits it in a package that can be low cost and low power. And it’s gotten rare reviews like from Walter forward concepts. Now, let me talk a little bit about our radio, our modem. The thing about our modem that’s a little bit different from any other currently shipping even 3G modem or LTE modem like Qualcomms is that ours is a fully software defined radio, it is, there is nothing in the digital processing that’s done using a hard algorithms, everything is done in software.
Now of course this is a Holy Grail of modems, no one believed you could do it. My friend Steve is going to talk to you more about how we do it but let me just start by saying that it opens up some pretty important capabilities to you, number one because you’re reusing the circuit since it’s programmable and software you aggregate the uses of transistors, any given algorithm you are going to run and you run a bunch of algorithms over the course of a few milliseconds when you are communicating over the air any of those algorithms all we use the same small piece of silicon because it’s program, we got a bunch of firmware running on that.
That’s why our modem is a lot smaller than anybody else 2.5 times the density, but equally important is that you have the opportunity to optimize the algorithms that are being run. You can implement new algorithms, you can implement new techniques that increase the throughput, one of the things that people don’t think about very often is LTE is 100 megabits per second or 150 megabits per second or average the throughput is many times lower than that because that 100 megabits per second is the thing that you want to have a big number on the box but that’s what happens when you’re standing right next to the cell tower with no interference of any kind and the reality is that (inaudible) and what you really want is to optimize actual bandwidth you’re receiving real world conditions like watching the World Cup in the café and I really want to see that play but is a lot other people that also want to see that play at the same time.
So there is lot of interference or demand. the real world throughput can be optimized, because we have a software-defined radio, because we can optimize the algorithms for real world throughput. And of course, flexibility allows you to very quickly get through the challenges that are faced during the certification process with carriers et cetera. And it even allows you the prototype, new ODI inaudible technologies.
And there is real spectrum challenge coming. the growth of data usage is exponential as you can see here from the chart on the left. This is measured in petabyte. So 6,000 petabytes is we the expected demand, what’s the thing after pedax, right 6 exabytes of data aggregate demand in 2015 for mobile, for total mobile bandwidth across all the users.
This exponential growth is attempting to be serviced by a spectrum allocation that’s relatively flat over the last 10 years. The amount of spectrum available for cellular data has been relatively flat, and if you thought about every possible place you might go, harvest spectrum like takeaway UHF, other areas that have radio frequency spectrum available to you, the most optimistic estimates of that would be, you could almost double the amount of spectrum you have today over a decade.
In other words, it’s growing 7% per year and the usage we want, the demand for the data is growing exponentially. Well, what about improving the efficiency of that spectrum? While the underlying protocols aren’t giving you much, if you look back six years ago when HSPA REL7 came out and you moved forward to today with LTE. the advancement in terms of the total improvement in spectral efficiency, in other words for every megahertz of radio-frequency, how many megabits of data can actually get over the air, I’ve improved it by a whopping 20% over six years, the spectral efficiency is beginning to max out, we don't have a way to obviously improve the efficiency and we don't have a way to obviously dramatically grow the allocation, so we have to come up with major architectural innovations in order to get there and software defined radios can help us do that, we can use creative ways to aggressively allocate the cell size or to create adaptable cell sizes or to create algorithms that have 1X, 2X, 3X, 4X, 5X improvements in terms of the realized throughput, while not changing the ideal case significantly.
And that turns into a real world benefit, if I can double the efficiency of the actual throughput per megahertz or double the capacity of the system.
Right now there is nothing on the horizon that I can do anything like that other than some of the innovations we're making, because we have a software defined radio.
Icera, the company we acquired a year and a half ago it has been around for a while, they started shipping modems seven years ago in 2006, they've even been making phone calls and one of the questions I get often is yeah, Icera was and has been and is the leader in data, the highest performance of data sticks on the planet have come from Icera.
The first guys to do LTE data on the AT&T network. The most pervasive other than QUALCOMM in all data products. But what about voice, you guys don't know how to do voice well. Icera did their first voice certifications on a phone, on a voice device back in 2008 five years ago and has been successful. But I would like to show you that Tegra 4i, the chip that we’ve just recently finished building the integrated version has the modem and our processor in it can make a phone call because I get ask this question so often of voice phone call. Is anybody there? Hello.
Philip J. Carmack
Philip J. Carmack
Is this Steve?
Is that you Phil?
Philip J. Carmack
Yes, this is Phil. How are you doing?
I’m happy to hear from you.
Philip J. Carmack
Well, I’m actually at the Investor Day here today and I would like to have you come and help me out in a little while if you can.
When that was today, I think we can move on. Wow. I want to find straight away and come in let’s say it.
Philip J. Carmack
Don’t give me heart attacks like that. You are the feature guy, so looking forward to having in you here.
Okay, great. Talk to you soon.
Philip J. Carmack
All right, thanks Steve.
Philip J. Carmack
Yes, he can make a phone call. And there is a lot more better things to do than just make a phone call, but we’ve done that. We know what we’re doing in this part of the world and this part of the technology, we are meaning Steve and his team and all the guys in Bristol. And Steve is going to come up here and talk a little bit more about the steps that we’re going through to get both the tablets with i500 and Tegra 4 and the phones of Tegra 4i to market. And we are going to break it down and maybe an enough detail that you’re going to be even bored as we stand between you and lunch. But I’m told by our U.K. guys who run the modem business that the only thing more dangerous than being between your audience and lunch is being between your audience and (inaudible) is that right? So that would be like caring or something like that. So we’re okay.
Anyway, there’s a lot of milestones that we went through along the way. As the industry has progressed trying to build on the 3G with the HSPA standards, it was actually a time of fair amount of turmoil in the standards bodies where they fairly rapidly were throwing out new ideas like turbo decoding in order to get up to the 21 megabits per second with HSPA plus, whereas HSPA started as a 3.6 megabits standard and built up from there to get all the way to 21 and then even 42 with dual cell.
And there were some pretty important innovations that had to happen along the way. And guess what, Icera was the leader in each of those steps along the way, the first to actually ship, the first to actually be able to get the carriers up and running with most of these new standards and you can see it right here and world leader with LTE as well.
As carriers were just first beginning to bring up their infrastructure on LTE, they ran more of their tests on Icera modems than on any other modem during the development process. And so we’re not new to this industry. We’re not new to this technology. We recognize what has to happen in order to make it work. And Steve will tell you more about it. Now I would like to shift gears a little bit and talk about, I made the assertion that computing matters; that visual computing matters for the mobile industry.
One of the things that the consumers been conditioned for is this exponential growth in terms of computing capability due to Moore's Law and Moore’s Law in the past has been used to give you more and more performance and generally at an exponential growth rate in other words every 10 years the performance grows 110 times not 110%, 110 times.
The problem however is that purple line at the bottom that you can barely see above the axis is the rate of growth in energy density for batteries, the battery that goes into your cell phone or tablet, the amount of energy it can store per volume so the amount of space it takes up has grown at a combined annual growth rate of 7% per year over the last 10 years for a total of 2X over the two year period and while 7% per year might sound good to a bond investor, it doesn’t sound very good to a technology guy like me who is used to Moore’s Law the 1.6X per year.
This is real problem because we’ve conditioned the world to lot more compute, we have all these ideas to use more compute like infiltrator or computational photography or reality augmentation or all kinds of things but we’re not, we are not creating energy at a fasten up pace to be able to keep up with the computing demands.
The first is IPC. IPC stands for instructions per cycle or in other words, microprocessor, how much work you get done every clock, I had a 2 gigahertz processor that means the clock is running 2 billion clocks per second, right. Now during each of those clocks, how much work do I get done, our current architecture, the Tegra 4 gets 1.9X, the amount of work done per clock compared to our competitor, now those of you who have been in the CPU industry know that when you’re fighting hand-to-hand in CPU battles, you feel pretty amazing if you have 10% or 15% lead.
Having a 90% lead in terms of architectural efficiency is a pretty stunning development and it happened because of the focus on architectural efficiency, things like more aggressively looking at branch target, predicting branches, Java does, Java is nothing but branching. Things like using a unified caches instead of dedicated caches where dedicated L2 caches per processor make it trivially easier to increase the clock frequency than they get harder to store the entire working set of the most important algorithms, like Java into the cache.
So that give us a 1.9X advantage. I’ll show you a comparison, demonstration in just a minute. For the other things we did was we said one of the biggest consumers of power in your phone is the backlight, tablets even more or so. A tablet might consume one or two watts in backlight power and that is just a – that’s power that yes you need it, you want to be able to see your screen, you wanted to look beautiful.
So you wanted to be bright, but it burns a lot of power. Well, we give created this computational technology called PRISM. And |Tegra 4 implements the second generation of PRISM. What PRISM does is it in real time every frame fully analyses every pixel in the scene along with the brightness that you have chosen and it figures out a way to completely redo all of that. And it changes the backlight, it cuts the backlight power in half generally, while not decreasing the brightness.
So you get the same brightness, but at half the power. Super important, opens up more of the envelope for doing computation. Lastly, we have this 4+1 architecture one of the things that happens is when your phone is sitting in your pocket or your tablet is sitting on the table, if doing things if not doing much but it’s doing things it’s checking your news feeds. It’s checking your e-mail to see if it needs to think or whatever it’s doing.
So it’s not in a deep sleep, it’s doing something but it’s doing very little and we sometimes call that active standby. But during that time it is very few MIPS and if you turn out really fast processor, it’s going to burn a lot of power, and that’s going to use up the total amount of battery you have available to you during that 80% of the time that is in your pocket, we can't have that happen. And so the fifth processor in this 4+1 is unique processor that allows you that extremely low power in cases where you need to be alive but you need to be doing very little, it’s in many ways like a hybrid type system.
This breakthrough allowed it to be identical to the main processors, this is not as of today a big little solution as you may have heard off. This is a solution where they are architecturally identical before and the one identical, so the exact same software, the exact same kernel runs in both cases, the system doesn't even know it's running on this extremely low power processor.
Given all those innovations, I would like to show you a comparison using our most efficient processor ever built in Tegra 4i. What you see here on your left is the Tegra 4i based phone, on your right is the QUALCOMM S600 base phone, this is their latest and greatest, the number you see on that phone, the 736.6 milliseconds and the 875.0 milliseconds that's the amount of time it took to run this benchmark which is called the SunSpider. SunSpider is the industry standard Java benchmark and as we add our competitors in the industry have known, Java is the most important indicator of web performance and one of the most important things to run fast.
For this benchmark, the lower the number, the better because this is timing how long it takes, that's how you can tell an engineer about the benchmark not a marketing guy, because lower is better. The faster you run, the shorter it takes you to run and so you can see there we're 38% faster. Now we run it and we measure the power. So here we are 30% to 40% faster, you can see the power is fluctuating around a bit, we actually instrumented this and measured it many times per second and calculated out the total power but at this performance level, we are about 10% to 15% lower power. Let’s find and go back to the by the way maybe I didn’t explained that clearly, the blue things are the power meters, sorry about that. I just forget to explain that.
So 40% higher performance and lower power it was a pretty amazing breakthrough to be able to give you both. Okay. Take me back to the slides just representing what I just said I mentioned in the beginning that HDR has really taken off now these software, these apps that are being downloaded for Android today are software based high dynamic range photography applications, when you take out your smartphone and you want to take a picture and there is something bright behind and the face is going to be washed out or the backlit thing is going to be washed out, you are not going to be able to see what you are doing, you want HDR images, the problem with HDR images or photography has been that it takes like five seconds to take the picture because what it’s really doing is taking two pictures, two exposures and so it has one where the back isn’t blown out, the light behind you hasn’t blown out and one where your face is at dark, it takes those two, those two different exposures and then it merges them together.
So first of all it takes time and you might miss the shot, second you are not allowed to move during that five second interval because there will be blur because it has to add these two photographs and then blend them together.
What it really needs is to be real time and this is one of the things we have implemented with our computational photography engine which is called Chimera, Chimera allows for real time high dynamic range photography like 0.2 seconds or even video, full video rate, 30 frames per seconds with HDR. You don’t have to stop, take two pictures and merge them in the software.
Our engine allows you to do it and this engine as I said is called Chimera. To do HDR at the levels I was just talking about, you need 25 times the computation of a CPU based HDR photography application.
However, once you do HDR, you always want HDR and then there is orthogonal features you would like to have. In Chimera, this collection, computational photography capabilities uses the GPU, it uses special purpose image sensor processor, ISP and it uses software as well to manage it.
It allows for many X like 50X or 100X or even 200X performance advantage and more importantly performance per joule advantage, energy density advantage on computation.
This allows you to do other things like, we’re about to do for you a tap-to-track demo. Tap-to-track is where you’re trying to take a photograph and let’s say for example, you’re at your kid’s soccer game and you want to stay focused on your kid.
Now there’s no easy way to do that. Sports photographers pay $10,000 for their cameras so they can focus fast. That’s the single most expensive feature that they add to a photo journalist camera is they can focus fast.
What it can’t do is recognize an object and stay focused on that object and can keep refocusing pretty fast, but it can’t recognize an object because it doesn’t know what a person is. Computational photography can know what a person is or what an object is. But if I also use computational photography to recognize an object that once again doubles the computation, I need it beyond HDR. And I don’t want to turn off HDR, so that I can do tap to track, I want both. so that means I need 50 times the computation of what a CPU might be able to get me. Well, we’ve achieved that and I’d like to show you a demo, tap to track with HDR running.
And Doug here is going to do that demo for you.
Unidentified Company Representative
Great, thanks, Phil. So in the scene, we have a, this is high dynamic range scene and you can tell from the picture being broadcast by that $50,000 camera, we can even capture the scene accurately, because the background is blown out, because it’s exposing for the foreground.
All right, what we want to do, what you can do today, whether that is camera phone, as you can force exposure on a particular object or what happens is if you recompose the picture, it adjusts automatically, you don’t lock it to the object, you lock it to the particular place in the screen and that’s clearly not what you wanted.
What you wanted was to be able to select an object that you care about and to be able to recompose the photo, move it around or the example, Phil gave actually there will be motion in the scene itself someone running around in, running in and out of different lighting situations and you maintain proper focus and exposure on the object that you’re tracking for as long as you, as long as I selected, until select something else that will stay the object selected.
If that object goes leaves the scene, I come back it will reacquire the object, even it maintains persistent tap to track, which is something that only us can do in the industry.
Philip J. Carmack
So what you’re seeing here is, this is what you’re seeing HDR. You wouldn’t be able to see both the church and the mountain, if this was an HDR, is that correct?
Unidentified Company Representative
Philip J. Carmack
What you’re also seeing is object recognition in the form of tap to track, in other words, stay, be aware of this church and keep that both in terms of exposure and focus that’s the thing I want to focus on and expose on. Actually it is computationally recognizing that object and if I move around it still stays on locks on to that object.
(Inaudible) Thank you. There is a whole body of I’m pretty excited about it. There is I’m not a photography expert but I’m an enthusiast, I have a $10,000 camera, but mostly because of the glass, but actually this is way more cool to me, and I think you’re going to see because we created this engine called Chimera which is the computational photography engine you’re going to see a whole host of new algorithms that come to there to enhance photography, and photography as you can see from the number of downloads has a broader appeal, everybody in the world cares about photography, people who use Facebook care about photography, the number one purchaser of digital still cameras today right now is young women in the age of 18 to 25, and so photography matters to a broad section of the population and these features are critical to making that photography experience good on your mobile device.
And fortunately as we look forward our answers get even better, we’ve been in the Tesla business that Jen-Hsun mentioned, CUDA business it’s something we have been working on for a long time, some of the early work that was done there had a lot to do with imaging and unique things you might do with images and mapping some of these advanced algorithms into GPUs, face recognition where some there might be at a immigration centre at the airport, there might be a camera with a whole rack of GPUs in the bathroom is doing actual face recognition and saying that Phil walked by, let me know if Phil walked by, that type of thing. So we’ve not the bunch of those algorithms historically as developments have happened in CUDA, and here is just some examples of algorithms that might be relevant for various forms of either computational photography, reality augmentation or other things you might do with imaging. And in general, you’ll see that you get a 5x to 50x, 60x advantage compared to doing a high performance CPU implementation of the same algorithm.
I mentioned there is this 100x problem or 50x, 100x compute improvement over a decade and 2x battery then for the improvement that’s a 50x. You need architectural breakthroughs to get that and you want that because you want these new forms of visual computation that will change your life.
Before I finish, of course I should talk a little bit about games. Games matter to us. TegraZone, is a great indicator of the importance we’ve placed on games and the connection we have with the consumer with respect to games. TegraZone users has grown by 81% over the last 12 months. As Jen-Hsun mentioned, we just crossed over to where more game developers are developing for mobile devices than anything else. A year or two ago, this would have been swapped, PCs would have been number one and mobile would have been number two.
Five years ago, mobile would have below Xbox 360 and PS3. So an indicator of where things are going, where the revenue is going to be, it's pretty clearly defined here when you look at where the developers are going. 10 years ago, I'm not sure anybody expected to see what is $50 billion worth of blockbuster games in the PC market or $80 billion in the future. I don’t think people expected it to be that big, but because of the developer community it became so significant and so compelling that it becomes a major and an important form of entertainment that everybody participates in. And this is now well underway in the mobile world and points out the relevance of what we at NVIDIA are doing for mobile. As I said earlier, 76% of revenue now in the U.S. for Android app revenue is game revenue. In some other countries it’s even higher, I think it’s 96% in Japan, but the number is big, the most important revenue source for mobile apps.
Before I turn it over to Steve, let me just as a last part give you a little bit of view of our roadmap. A few years ago, two years ago we showed a roadmap that looks sort of like this, when we introduced Tegra 3, an aggressive ramp of computational capability into mobile. And the first three generations that are shown here had to do with taking the original Tegra 2 to Tegra 3 with Quad Core to Tegra 4 with computational photography and things like that.
And what is not really shown here is the fact that starting with Tegra 4 these are now families, and so these architectures are the architectural name, but they associate with the family. Tegra 4 is the current one. The next one hasn’t named because it’s not a product yet, Logan. Logan will be a family like Tegra 4 is a family now. We’ve already bitten the door. We already have that capability now built into our system to deliver a portfolio not just one product. And so you’ll see a more rapid fire set of products coming out with different, that attack different product categories. Not only then will you see this first Logan being the first case of an integrated Kepler in a mobile device, allowing you to do what you saw on the KAYLA system earlier, which is a good way to develop for Logan. But you’ll see a whole family of products over time. And this will allow us to take our advantage of GPUs across the entire spectrum. I showed you how we’re going to be able to attack a much larger spectrum, consumer products just because we have a family, because we have LTE and other competitor capabilities. But being able to take that competitiveness and have this weapon of the visual computing that comes from our GeForce, Kepler GPU and just follow on Maxwell is key to our ability to really change the game for people.
At this point, I would like to invite Steve to come forward and talk about a little bit more detail about our motives.
Hi. So I was one of the co-founders of Icera and I hate to tell you why LTE modems are so much more exciting than all these GPUs and graphics stuff. In case, I’m wrong though, I’m talking before lunch, not after so I’m quite relieved about that. So I see this modem technologies being around quite sometime, we’ve been certified with 95 carriers, we shipped about 20 million modems in the data space and we are shipping our first generation, LTE modems on the AT&T network today.
What Phil talked about in terms of software-defined radio, it’s an extremely powerful technology. it’s allowed us to keep the pace with companies with much, much larger budgets. It’s allowed us to innovate, it’s allowed us to add new features at a pace that is simply not possible, with conventional ASIC approaches. What's also not so obvious is that it gives us a significant (inaudible) advantage, so we are about 2.5 times smaller than the competition in terms of pure silicon area comparing like-for-like features.
We currently got two new LTE modems, completely on track at the moment, I'm going to go through some detail about the pain of career certification. We've done a lot of them, we're being through a lot of certification for voice and data products, I think for the first one out of full head of heap, that's moved on to lot since then. We are now into our full process generation of fundamental software defined radio architecture, so back in 2005 we launched a product with SoftBank in Japan, so this was the first 3.6 megabit HSDPA device on to the AT&T network, we had legacy support for 2G GSM, GPRS and EDGE and we also upgraded it entirely through software at HSUPA.
Each generation that comes along, what we typically do is we increase the computing capability of that platform, we make minor architectural tweaks and we increase the memory size, that's allowed us to make a pure backward compatible solution that the previous generation software can now get instantly forward ported on to the next generation. By doing this, this gives us a huge advantage in terms of software, bring out both the physical and the protocol stock.
All the way through now to the 40 nanometer, this was our first generation LTE multimode solution supporting 50 megabit CAT 2 LTE, 42 megabit dual cell and 28 megabit per second MIMO. The current technology on which the Tegra 4i and the i500 platforms are based allow us to do a 150 megabit carrier aggregation LTE as well as additional standards like TD-LTE for the Chinese and TDSC for the Chinese markets.
So as I mentioned our LTE modem technology already certified and it’s being shipping on AT&T for a number of months now. So certainly a good proof that our protocol stack can stand up to the rigor of the AT&T requirements.
So when we started Icera our challenge at that time was to try to implement a soft modem architecture that could support 3G voice. It seems like quite a long way time ago in terms of compute. What this graph shows you is the number of arithmetic operations per second required to implement the physical layer of a particular standard. It doesn’t actually include all the additional things like shuffling data around, (inaudible) leaving and general permutations that go alongside that. But it does give you a rough order of magnitude of the compute complexity of a particular standard. 3G voice there requires something like 3 gig operations per second these are sort of adds and multiplies and so forth in order to implement that standard is pure software.
Scaling that up to Cat 4 LTE 20 megahertz you’re looking at more or like 65 times of that what an excessive 200 gig operations per second in terms of number of arithmetic complexities required to do that. We had to develop a absolutely revolutionary process or architecture to implement those in a power efficient manner, in a very low compute requirements.
So we wanted to try to do this typically less than a gigahertz in terms of actual processor speed. So we develop the processor architecture that allowed us to take the various wireless signal processing algorithms that we needed to implement, do them entirely in software and focus the process architecture on that problem. This processor although general purpose in terms of wireless signal processing is not general purpose in terms of more conventional DSP architectures, and for this we needed a very carefully crafted processor design.
What might be less obvious to you is that what we get from this is the additional fringe benefit of being able to completely innovate entirely in the software. So rather than having made the decisions at design time as to how the particular algorithms we’re going to use on the platform are going to be implemented, we can dynamically change that at some point in the future either through field upgrades or through purely dynamic software switching at one time.
So this slide is an attempt to give you an insight into why it is so significantly smaller than a conventional modem. Conventional legacy approaches for the design of this type of solution end up with a quite a silicon area plot. So typically what you have to do is for each function that you’re trying to implement, you have to put down a dedicated piece of hardware, you have to add quite a lot of additional memory for that particular function. And you’ve got to do that for all the particular pieces of the span that you’re looking at.
So in the case of LTE, you’ve got things like [crafted LTE] processing to do. You’ve got equalization FFTs and so forth, all of these consuming quite a lot of silicon real estate. And when those particular functions aren’t being used, the silicon is effectively dark and unused. So you’re wasting silicon area. What we’ve done is very much turned that problem on its head. So what we do now is we actually have a general purpose processing engine that can run through all these algorithms alternately and effectively use the same transistors over and over again very, very different algorithms and the net result of doing that is that we have a significant advantage in terms of silicon area.
So I think it will be a good time to take some of the theory that I have shown you here and take it out to practice. So I’m going to give you a quick demonstration of our 100 megabits per second LTE running on our T4i integrated Tegra device.
So what you can see here is a Tegra 4i form factor device sitting on top of a tester. So this piece of test equipment emulates the functions of the radio network. So it includes the base station some of the core network functionalities.
What we’re going to do here, if we can stream our video to it through those two RF cables that are actually connected to the right hand side of the tester. And then in this way, we can emulate what would happen if you own an LTE network and you were downloading and streaming a live video sequence.
So if you’re looking closely now, what you can see here with these peaks of up to a 100 megabits per second. So obviously, in order to sustain a video on that side the form factor, you don’t need to sustain 100 megabits per second all the time. So what’s happening as you’re downloading a large chunk of data, it’s being buffered and stored, and then that process is repeating, say every 20 or 30 seconds when the next lump of data for the video stream needs to go through. But approved point of the integration success of the Tegra 4i device and the ease with which we have managed to bring this up quickly to the LTE modem as well as the application processor.
I’m now going to talk about the lengthy and arduous task of carrier certification. So this is a complex process, it varies significantly between operators and requires a huge amount of effort to get through. We’ve done a number of times with a number of previous generations of platforms. So we do have experience in doing this, I think this is now going to be our full certification for example that we’re going through with AT&T, a significant amount of effort.
So this gives you the high level, the typical milestones that you’re looking to go through. The initial piece is what’s called PTCRB casually named, personal communication system, test certification requirements for the equipment in Europe is GCF, but these are a bunch of tests that you run on a piece of test equipment using predefined scripts. And you typically have to go through something like 2,000 or 3,000 of these tests in order to get certificates for each product that you’re going to launch on.
Then you have to go out and then you have to test and do NVIOT as its called, so this is testing against the major infrastructure vendors like Ericsson and Alcatel Lucent for example, improve from a protocol perspective you can interoperate with those devices. Then depending on the operator you then have to go out and do configuration, do some additional testing called Network Operator IOT now with a particular configuration and mix of infrastructures that an operator is using, and at that point you hopefully will get to a type approval process or through all particular tablet.
Now this is a sort of where we will go to this is specifically, I’m going to talk about AT&T. AT&T is one of the more complex operators to certify with they have a very rigorous process that goes on top of this, some other operators instead having to pass test you have to sort of pass brown envelopes, but that is sort of more of an Italian field really.
Anyway so with Tegra 4i, the AT&T certification just to give you a snapshot of where we are today. So we need to pass something like 3,577 tests, we’re about 80% through those. So these tests are looking at things like testing the same, the device RF bands, you’re looking at physical add performance, you got to check that you are not emitting power RF unwanted power into bands, you’re not supposed to radio hitting into, you got to check how sensitive, so in other words how lower power signal you can receive, you got to do a bunch of measurements, you got to test hand over procedures and call set up procedures.
Each one of these tests some of them can take 2.5 hours to run. So you can be looking at many weeks to simply go through the process of running these tests assuming you got no issues. So this is really just to prove out sort of the fundamental operating of the device itself and prove that you’ve got a stable and fully functional modem.
Networks and operators of our each infrastructure has its own set of proprietary tests, these are more protocol variety. So we are not generally testing out how good your physical app performance or how good your IRF is. They are just checking that you can reliably set up course and so forth with the infrastructure. I mean there is a smaller number of tests to go through here, again in case of AT&T you have to look with Ericsson and Alcatel-Lucent and this particular process will take us a number of months to get through and again it’s quite an expensive procedure to go through this. So today we are at about 83% through that 80% through the PTCRB process.
Now AT&T have a process called Adapt, so this is for certifying the reference platform, in other words, the chipset and the bullet that we use and that we give to our customers at a starting point for their commercial products.
Now the process of adapt is really to give the reference platform a boost off and speed up the whole certification process of getting the products on the market. And so actually NVIDIA and our friends in San Diego actually there are only two companies that are going through this process. So it's expensive mechanism to ensure that they’ve got high quality products on the network. An example of a product that’s didn't do this and had actually no problems at all (inaudible) was the iPhone, if you remember that product when it launched.
So in terms of adapt, we have three phases to go through. We are first 100% through Phase I and about 20% through Phase II of the 2,100 test that we have to get through. The trigger for entering that – exiting that phase, obviously you've got it by the end of this whole process of Phase II and Phase III, you could have completed all these other things but many of these activities can go along in parallel before the end of Phase III you’ll present it with a certificate that says your platform, your reference platform has been certified through AT&T and at that point you are ready for lab entry and we aim to be there by about September of this year.
In parallel with all this there is a whole bunch of additional testing that’s going on for field testing, so this tends to be somewhat more arduous us with voice, say you are looking at corner cases for co-drop statistics and so far and then you are doing additional testing on the stability of the platform meantime between Fridays and so forth and looking at the (inaudible).
So by the time a customer takes this reference platform and goes into lab entry for product it is called, you’re looking at some additional requirements to testing and tell themselves, that’s obviously not part of the reference design and again looking at very specific carrier customizations and so forth. And if you’re looking at top right hand corner there that gives if you have very, very good eye sight you can look at exactly what happens in order to take that through to type approval and a production smartphone device.
How long it takes a customer to go through that process can vary typically between 2 and 3 months, it really sort of depends on how experienced the customer is with working with AT&T and our current plan is factoring in as you see to the end of January before that is actually achieved for production smartphone products.
And so quickly I have to now hand back to Phil to wrap up the final as possible.
Philip J. Carmack
Thank you, Steve, great work going on there, we’re excited about where we are I think where we expect to be making good progress and finding as usual that we’re able to adjust to the situation at hand and work through carrier certifications, and the team is doing great there. As I said at the beginning, we are, we and Tegra are here for a very important reason. We truly believe that the insight we bring in computational, in visual computing in many different ways that it applies to is relevant and important to the industry.
We’re leveraging that to create a whole new as I said discontinuity from a competitive point of view. So we can go and track that large market that is out there. And we have a pretty clear path on what we needed to do get there, the key elements are coming together quite well. We’re excited about it, we’re excited about where we’re going this year and into the future. Thank you for your time.
Unidentified Company Representative
All right, thank you, Phil. We are exactly on schedule, so right now we are going to take a break. One of the things that I had mentioned earlier and we want to make sure that you have is interactive time with the executives. So our lunch is going to be right outside over here, we will take a break for lunch and let’s say we will start up again at 1 PM all right. We will start up at 1 PM and all of the NVIDIA executives, the people who presented this morning Steve, Phil they are all going to be available.
So take advantage of this time and then after lunch, we will get into GPU compute. We will get into the financials and then we are going to have the final Q&A. All right. So 1’O clock PM let us start up again.
Are we good to go? Few more people, are we got to keep the show on the road, and we’ve got to get these guys out at (inaudible) in an hour and make sure we have time for questions. Okay we are going to pick up again and as I mentioned on the agenda one of the things we do want to make sure that you guys understand is the increasing importance of GPU computing, and one of the events we recently had was the GTU Tech conference and it’s a huge event with a lot of things going on. But Dan Vivoli who is going to come up and present next is our Senior Vice President of Marketing, but he is also the guy who basically created GTC. He is one of the original employees of NVIDIA and he has been kind of shepherding the whole GPU computing innovation engine and been at the heart of GTC.
So we couldn’t think of anybody better to kind of give you an idea of what really happens at GTC and why you guys should care about what happens at GTC. So with that I would like introduce Dan Vivoli, he is our Senior Vice President of Marketing.
Daniel F. Vivoli
Thank you, Rob. All right, so I got to draw the post-lunch coma session, and when Rob told me I was supposed to talk after lunch, I volunteered to ride my unicycle and juggle, which sadly I know how to do from too much time as a student where I should have probably being studying, and when he said that probably wouldn’t be practical for you guys, I decided that maybe I should talk about GTC, so if you look at the agenda GTC is kind of like the Sesame Street, the thing that doesn’t belong with the others on your agenda, it’s not a business, it’s not a financial thing, I don’t have a bunch of bar charts about performance for revenue or anything like that, but it is relevant.
Now how many of you know what GTC is, GPU Technology Conference. How many of you don’t, let’s try it that way. It is okay to say you don’t, if you don’t. How many of you have been there? How many of you feel like when you were there you got a good feeling for what was going on there, really I don’t believe you, and I will tell you why because I have been there for four years in a row and every time I go, I feel like the first day I went to Disneyland as a kid. I feel like I didn’t see anything, I saw stuff but there is a bunch of stuff I missed, like I need a lot more time and it turns out there is a lot that goes out of GTC that you prior don’t know about.
I want to share that with you today. So I’m thinking about how to talk to you about GTC, the first order is we think of it as probably one of the most important early indicators for us as to where our ultimate business is going to go over time, and a lot of you probably think of GTC or the GPU Technology Conference as something for Tesla, it turns out, it is not true, it is for Tesla, but it is also technology being built, being innovated for all of our products because as you know we’ve had a plan for long time to move CUDA into our entire product line including Tegra. So Tesla to Tegra. And so at the GPU Technology Conference, a lot of stuff happens that actually turns into applications later that are important for our company.
Now I tried to figure out what’s a good analogy for this conference, and it’s the bringing together of all these really intelligent people doing cool stuff and we learned with CUDA early on that we want the real smart ones. The real smart ones, the people that are out there, doing algorithms with it. and it turns out that we’ll also learn that if you get them together, they’ll share with one another, they talk each others’ language, whether an astrophysicist actually talks the same language as the Medical Imaging guy, they talk in algorithms.
And so we figured out, we need to get them together. so I try to figure out what the right analogy was here. And first, I’ll think about is a cauldron. It’s a cauldron of innovation, well that’s not the quite right analogy, because that’s kind of a bad thing, I thought well,, it’s a cornucopia, another term people use for stuff like that, and that’s not quite right, it’s kind of fruity and yummy and quite right for this.
So I came over this idea of about furnace. the reason I think about it this way is, because there is this furnace, I don’t know how it’s to explain it. it feels like a furnace of innovation where this is energy put in, but more energy is coming out. And it’s gotten to the point now we’re like the sun, which is the ultimate furnace it’s self propagating, it’s self fueling, and it’s kind of at the point where it’s now the energy that radiates out that’s I think fueling a lot of the growth of our product lines into the future.
So that’s what I want to talk about today, our furnace for innovation GTC. Okay. So first of all, if you could measure something like intellectual density, I’ve never seen a measure like this, but intellectual density would be something like IQ per millimeter, per diverse field. I’m pretty sure there’s no place on earth like GTC. in fact, I would guess that there is absolutely no place like that are on earth like GTC.
This is who is who of anybody that’s doing anything interesting in science or innovation. and it’s across so many disciplines, it’s impossible to really mention them all. And so it’s this kind of intellectual density that makes it special. It’s 3000 people this last year, 3000 people from all over the world that came together to share their ideas on what to do in GPU computing and of course, there’s a research institution that you know about or care about that’s doing important work, they were there.
Also it might surprise you that we have spend just for researchers, there is a lot of companies that are there now. One of the indications of the growing importance of GPU computing and GPGPU is that a lot more of the industry is involved now and actually using it and doing a lot of the research and practical application of GTC and these are some of the GPU computing and GPGPU and these are some of the people that were there.
So it’s big, it’s dense with respect to the kind of people that are there with their intellectual capacity. I want to take you through and just give you a feeling for what the show is like. So what I would like to do now is I want to start with just showing you the intro video that we used to kick-off the show. And then I’m going to dive into some things that I think will help you in shorter period of time as I can, make you really feel like you were there. I want to give you feeling for what really happened there, okay. So, let’s role this intro video.
All right. so that was the kick off video, now three things strike me about that video as we look at it. The first thing is compared to the first GTC, four years prior we couldn’t really have done a video like this because there really wasn’t that many real practical applications yet that we could highlight, but then you notice the second thing that I noticed in this video, which is that there is a wider array of diverse set of things being done that are pretty interesting and pretty important by the people working on GPU computing. But the further most important thing about that video is that GTC is not really about us. In fact, we view a successful GTC if we hold this conference and we are in the background because the whole point of it is to bring all these people together and let them share with each other and then great things happen. And I’ll show you some examples of how that’s manifested itself over time and how that later turns into applications which are showing the revenue.
So let me walk through some of the components of GTC that make what it is. The first thing is posters, when we’ve done our first GTC this is something we borrowed from the academic community, this idea of doing posters sounded kind of weird to me. But if you walk around GTC, you see these literally posters all over the place. And there is a 150 of them this last year. And every single poster is a hypothesis and an algorithm and then some results and then a conclusion about somebody's idea about some algorithm that they could address with GPUs. And there is a 150 of them that are new and different this year.
And they're really fun to go look at it, in fact wander around the show, you can spend hours just looking at these things. One of the things we do is we have a beer and poster night, the first night it’s sort of a nerds night out and the people that have their poster or they sit by their poster, they stand by their poster, have a beer, people come around and they talk about it. And they debate or share ideas or whatever and then people kind of mingle that way.
And so we do after two hours because we don't want to give them too much beer, but they stay around in terms of going out for hours after that because it’s such a new thing.
So let me show you kind of – and this sort of like the technical epicenter if you think about the core of the technology there and these are algorithms here is one about weather modeling, weather modeling as you can imagine is one of the most complicated stimulations there is and so getting an accurate weather model to be done quickly is really hard, it turns out there is a lot of elements of weather modelling that are very conduced to programming and of course GPU’s are being deployed in a variety of places. This particular poster talks about simulation called Scale 3 and in this case by using GPUs, we are able to make this part of the weather model go five times faster and this is a critical path of weather modeling in this particular case. So there's one in climate modeling.
This one is kind of still maybe near to your heart, some of you might know little about money in here. This is about trying to calculate given a specific amount of money or capital, what is the proper rating for a portfolio stocks to maximize the return. So this algorithm done by Riskcare, in this case they were using a 583 stock portfolio. I'm not sure why they picked that number, that 583 stock portfolio and in this case they were able to optimize their algorithm so that it was 60 times faster by combining a GPU with the CPU instead of adjusting it on the CPU alone and that means that the stimulation ran at 40 seconds versus 30 minutes which is pretty substantial. You can imagine that, that can change somebody's actual wealth potentially if you do it right.
That's finance, we have weather finance, here is one in a totally other part of the world. This is gaming. Now this is kind of interesting to me on a number of front, the first thing is this has nothing whatsoever to do with the rendering of the graphics for this game and in fact it has nothing to do with the PhysX, all things that we've talked about many times with respect to gaming, this has to do with the load balancing of users from massive multiplayer online games.
So one of the big issues is if you're trying to manage these games, you want to get as many users as you can. You, like all of the users to be able to be in the sane world as all of the other users. It turns out that one of the problems that folks like the World of Warcraft have or the people with Command and Conquer, you can’t scale your users past how many can fit on a server. It’s hard to spread across servers. That limits your number of concurrent users to, in a particular realm or a world to a hundred or thousand users. So that’s a problem for them.
Interesting thing here is they borrowed an algorithm from another part of the scientific community called N-body. N-body is a very popular algorithm in astrophysics and chemistry for modeling particles. They took that kernel and they turned that into an algorithm that they could use to manage the work load across servers to solve this problem of the constraints of servers and users.
So I think it’s a cool example of this cross pollination that we talk about. And that’s what GTC is all about. Now how can we get one field to teach another field an algorithm that might help them? So those are three of the 150 posters. So you can imagine how much fun it would be to walk on and look at all of these. Another thing we do at the GTC is the Emerging Company Summit. We’ve been doing this every – one of the – the four years that we’ve been doing GTC and this actually started as a seed of an idea based on early work that we did years ago with Keyhole. Anybody remember a company called Keyhole?
Now who did they become? Google Earth. So we worked with Keyhole years ago and we invested in them and we nurtured them along and we recognized that people didn’t quite get that this company had something that could be of value, but they needed a little bit of help. And of course, they ended up being something that everybody knows about now. And so this idea kind of germinated into a group within the company that tries to martial along, start ups that are doing interesting work in GPU computing or GPUs.
And so during the GTC, we have an Emerging Company Summit where we invite start ups to come. We provide them sort of a rigorous VC type environment. We have bankers there. And they get to tell their story and with any luck they will end up stroking a partnership or getting some funding or perhaps even become a company that gets bought by someone else, that’s what they want to do, but the idea is to nurture this ecosystem. And so we have had 37 this last year, there has been 221 companies total that have gone through this and there has been some early (inaudible) that Jen-Hsun talked about, a reference Jen-Hsun talked about it earlier today about MRIs and CAD/CAMs and yet the interest in trying to reduce the amount of time people spend in them, one of the companies that represented this year called Morpheus, they are building an MRI machine that cuts down the time in MRI from an hour and half to eight minutes which is pretty substantial and it is based on doing GPU computing.
So the Emerging Company Summit, that’s another piece that happened there, which is in itself can consume you if you wanted to hang out there. This is a picture from the exhibition floor which I think is important for couple of reasons; one, the first thing is that we have an exhibition as ecosystem large enough now with GPU computing that have 100 exhibitors showing off their wares and these are lots of companies from the system world, software world, applications world, add on world they are all there and so we had a big exhibition. Also this is – it’s fun to walk around and watch people and mingle again. There is a lot of good problem solving that gets done and here you can see this is one of the evenings where we have light dinner and so people get to walk around and hang around the exhibition floor.
Now it’s not all super techie there, there is some fun stuff we have there. This is a one of the demos that we had just for fun this is year, Shield, I think you know Shield. One of the things that we discovered is that Shield makes the perfect controller for a personal; it turns out that you can feel the controls. There is a high res display built in for the camera on the drone and that’s got wireless. And so the drone guys are coming to us and they are saying we really want a huge Shield for this thing. So we had one there and people loved it. Now I’m sorry to report there was one crash of the drone, I’m even more sorry to report that I’m the one that crashed, private pilot with a 1000 hours of flying time and I can’t fly a personal drone apparently. Anyway this was a nice fun thing for the exhibition.
And then now let me get into the kind of the meet, this is the talks, we had 450 sessions there and these were really super interesting talks and some of them are hands on about how to use the stuff. Now interesting thing here as the talks ranged from anything from very deep algorithmic kind of scientific mathematical things all the way to the other end of the spectrum which were, how are you using GPUs to make your company better. How are you using application that has GPUs in it to make your company better today? I will actually give you a taste for some of those that I thought were pretty interesting.
First one is that that one far end of the spectrum, the mathematical end, one of the presenters talked about how, [Ed Carl] talked about how he cracked the 4 quadrillion to digit of Pi, I didn’t know this, but apparently for the last several 1000 years, there has been a race between mathematicians for people to figure out the next digit of Pi and in fact the 3.1416 which we all kind of memorize that was figured out about 280. By 1940 or 1950 we heard about 100,000 digits to the right of decimal place and in 1995 there was this breakthrough in algorithms and those arms raise came about, people started using supercomputers to figure out these numbers.
Now before Ed did his work, the record was set by Yahoo on 8000 node, 8000 core supercomputer and they calculated 2 quadrillionth digit of pi. Ed built the computer with $5000 in his garage with one CUDA GPU and doubled the number of digits to the right of the decimal point with that machine. So I don’t know what this algorithm means to science, but it probably means something. Some of the work he did will probably end up somewhere. Some one will figure out a way to use it for something else. But it was a neat example of, what I thought was an interesting session.
Okay, another one here I’m going to talk about is speech. But before that, let me just kind of give you a way to think about this. So our GPUs, we design these things from the very beginning to be able to do a bunch of computation and then spray pixels to the screen and we’re spraying a bunch of pixels to the screen, right. So we build these processors to do this. Well, it turns out they work in reverse pretty well too.
We can take in a spray of information or streams of data and then compute it on the fly in real time and do something useful with it. And there’s several different application areas where that makes that kind of paradigm, one of them is speech recognition. And so there were several talks on speech recognition. This one in particular, I just picked out as an example. It’s kind of interesting.
You have an application where you’re trying to solve for three things in speech. You want a big vocabulary, you want to make sure you can understand accurately, but you also want to do it fast because if you can’t listen fast enough, you can’t respond quick enough either. And so what these guys did is they built a million vocabulary speech, vocabulary and they were able to decode that speech five times real-time and that is far faster than they could do before.
Before, they would do it about half real-time. So if you can imagine trying to listen to somebody, but you only can understand, only get half of what they say, it’s not useful at all. So it goes from being not useful to useful to be able to do this at five times real time. That’s our hybrid project.
Another one in this kind of streaming stuff and in doing application work on is vision. And we’ve talked a lot about here. There’s lots of work being done on vision. And also has a similar problem. It has a vocabulary that needs to understand, and needs to do it quick, and has to be accurate. And so, we show the demo here earlier, where we highlighted the church and we recognized that church next time around and kept track of it. But the computer didn’t know that was a Church, it’s just new to recognize that object again. The trick is to say, oh that’s a church, oh that’s a catholic church that been better, oh that is a protestant church.
So that the idea is to be able to teach the computer not just to see but to recognize. And so these guys are doing some really neat stuff, where the state-of-the-art for this has been to be able to identify and figure out what one type of class of things are say vehicles, cars has been tens of seconds to identify, but it’s a car. What they were able to do with the GPU is recognize a dozens of different types of objects. And they were able to do that in less than a second, half a second. So what we’re showing here on these different examples is, how quickly the computer can recognize that’s a horse, even with partial information it recognized it’s a horse. So this turns out to be great, it has never been done before. Being able to recognize all those different object classes and to do it in faster than a second is never been done before.
And you can imagine all the different places this could lead to help me or anything from a safety and a car potentially really cool apps for your phone all different either this could result in good apps. This one real quick, this is just another, I thought interesting example of vision, this takes it a step further, this says hey look when you look at a scene you don't just stare at the whole scene and take in everything equally, your eyes move around and your brain and your eyes weight things differently. What if we could make the computer smart enough to do that too then we can put our computational effort in the places that make the most sense and therefore get more auto-provision system and that's what these guys are doing. So that was a pretty neat are that you could layer on top of some of the other speech work that we do.
Now lets (inaudible) all kind of earlier things, let me get to stuff that is more in practice today and play today, this is ESPN. ESPN sent 16 people to GTC. They have a technology group that does all kinds of cool stuff and I'm a sports man, I’m sure lot of you guys are sportsmen, you've seen a lot of their work and this is a couple of examples, the first one at the top is this is the home render and they have all these four kit cameras all over the stadium and when the home run gets hit it, it lands out, they have the bubble popped up for 25, how that pop up so fast. Well there is somebody sitting at the screen at a high res screen he clicks the mouse where the ball hits and the system takes the information from all those cameras around the stadium, does all the math and figures out exactly how far it went and they do that with using our technology, which I think is pretty cool.
On the bottom of this, this is another one that I noticed before I realized how they were doing this. I know guessing football game, they have what they call the virtual camera which is – you're watching already in high-def. If you're watching a football game and then they zoom it on a foot of a player and that's also in high-def. Well how do they do that? Well it turns out that the regional stream was 4k they are showing you a HDTV ready high-def version of the game and if they want to zoom in on something they can take the virtual camera and move around and you also get that in high-def and so that takes a ton of processing power and they are doing that using our technology as well.
Okay so ESPN, who knew? And then, this is another one I thought was pretty interesting, because a lot of people won’t be able to relate to this.
Harley is a real interesting company. They are 100 years old. They are from the outset build motorcycles that seem pretty traditional, a little, almost old school in the way they’re looking, and the motorcycle turns out but they are very, very modern internally. And their brand is one of the strongest brands by quite long distance. Their brand is one of the strongest brands in the world actually. And it might surprise you also their brands are not just appealing to older people, they are riders are now half of them are under 35. So the reason that‘s all important because they spend a lot of time in the design and how these things look, and sound and feel.
And what’s shown in the box here at the bottom, is their design process and what this gentleman was doing in his talk, we’re showing about how they were changing their design process early on, where rather than relying on physical models and markups early on, they were using more visualization and modelling upfront, to do some of the work that used to be things that were build out of clay. There’s still both physical models near the end, but they were shortening and building these processes much early. He told a lot look really great stories about how much faster they were able to build stuff. I’m not going to have time to go through all these, but may be Dan, if you could scrub the top one a little bit. This one was really compelling and, he, Vivoli was talking to the audience, he designed a motorcycle seat from scratch and he was using 3D Studio Mac. You made the point and say, hey look this is a game developer tool, but it works awesome for making motorcycle parts, because we subdivisional services, we now have this part, you can send it to a 3D printer and within an hour or so they have a prototype. This was unheard of for years so ago, big deal it’s fascinating to watch though.
At the bottom is an example of them using a practical Bunkspeed, Bunkspeed is a real-time ray trace vendor, or you want to start the bottom one again just quick on it?
And with this it’s the ability to get real time photorealistic ray tracing for different looks for the motorcycle. This way they can build custom motorcycles or they can make decisions about the design. He talks about also and they did the motorcycle for, I forget the name of the movie now. Right now, I wish I would have write it down, they did had to call do motorcycle for a movie, and they never could have done it before this technology, because they don’t have enough time. They had to build this motorcycle like a month. But now they can do that with this kind of technology they can build custom products very, very quickly.
So last one I want to tell you about is IBM. And this is probably the best example of how far we’ve come in the last few years. So this was two different sessions done by IBM. One of them about how to use GPUs effectively in a computer environment, and the other one is how to use GPUs to manage the data base. And so the thing I can’t understand I hear is this isn’t about if you should use GPUs, this is the largest server company in the world. This is how do you use them, which is I think a big distinction if and how. So that is the sessions now there is a lot there and I gave you a sampling of it. I’ve a present for you today. For those of you that are interested this is a copy of a little book we gave everybody there at the show.
So there is all of a session and posters and stuff in it. And that I’ve given you a $1,000 value. You like money right, a $1,000 value this is a log in and password to access all of the sessions and posters and stuff that we have at GTC. And they really are pretty fascinating to watch some of them. A lot of these guys tell those stories in real live language and they are not just so geeky you don’t understand them. And you can get a really good feel for what is happening at the core, at the center of this innovation that's driving ultimately the applications that are running all of our stuff.
So let’s start handing that out now. Okay, so GTC gets a lot of, as you’d expect, gets a lot of notice from the press, we had 165 press from around the world there, lots of biggies from all over, and lots of great press from it. So that's another measure in my mind anyway as to whether or not GTC has momentum or is important is how much press they come in droves.
How we measure? Like we measure a lot of stuff, we measure a few things with GTC to see if it’s growing, and if it's having a bigger impact in the last one. And this is four different measures, attendance, the number of posters there, number of sessions, and the different countries represented, and you can see they’re all moving up nicely, which is great. You want to see that, you want to see us catalyzing more and more innovation and that's all great.
But it turns out that the problem of GTC is, that’s a big event here in San Jose, a lot of smart people who can’t afford to come here, there is lot of smart people that don't have the time to come here. And so, if you look at it another way, this is the first GTC, and then the last one that we just did, here is that growth and attendance, but what we've been doing for the last few years, it’s finding ways to take and create GTC like events around the world to bring the same sort of collection of innovation together in the same catalog, the same sort of innovation.
And if you add these different things up GTCs in other regions, things we call GPME, that’s in campus events, we’re actually taking this same sort of idea to a much broader base than even that 3000 that came to the GTC this year.
Okay, so Furnace, GTC has helped you get a little better feel for what it’s about. It's not just a geek conference. To us, it’s a peak into one of the key things that will drive our business which is software innovation based on GPU computing. And it’s not just for Tesla, it’s for all of our products and it’s up into the right. But you guys all like to right, up into the right stuff.
Okay, so thank you very much.
Thanks a lot, Dan. Okay, from GTC in the research and the innovation that’s going on at GTC to our business and the Tesla business seems like natural progression. So I’d like to introduce Sumit Gupta, the GM for our Tesla business unit.
Thanks, Rob. Any one got Shazam on their phone. Yeah, let’s take out our phone. Let’s get Shazam fired up. Hang on. Let’s take Shazam out. I want you guys to help me identify this song. So just start Shazam and have it identify the song. Let’s do the Shazam salute, going in the air. Did anyone get it? All right, cut the music. Which song is it? It’s Harlem Shake. You just used the Tesla GPU in the cloud. In fact, 10 million people use Tesla GPUs in the cloud everyday for Shazam. And in fact, it’s growing roughly at 2 million per month.
So these are some of the applications and Dan did a great job of giving us an idea of the spectrum of applications that people are using GPUs for. What I’ll try to tell you now is – show you some of the metrics that we use to measure the business, right. There’s a lot of people using GPUs. But what’s really happening in the business? Let me share what I used to measure the business. So I’m going to show you four growth drivers today. Kepler, our deal pipeline and how it’s growing, Apps. And I think Apps is so important, I wrote it three times, Apps, Apps, Apps and then I’ll share with you how Tesla is entering the world of big data.
So first of all Kepler, Kepler has been a phenomenal GPU for every business at NVIDIA. So Jeff Fisher talked about GeForce. It’s a same for me. We shipped Oak Ridge National Labs as our first system in Q3 last year. And really this quarter, Q1 is went up the Kepler products by Tesla have been available in volume from every OEM.
So we are now beginning to see our customers upgrade. I talked to you about this last year, how Kepler will drive a refresh for all the Fermi installed base that’s what we’re beginning to see Q1 was the first quarter, and we expect that to go throughout the year. So tremendous speed up, tremendous value, take of Fermi, put in a Kepler, you’re 2X faster. That’s growth driver number one.
Growth number two is from the pipeline. So before we get to the pipeline, I want to share with you the market opportunity. Where do we focus? So all I did here is I took IDCs HPCs server data. So it’s actually a sub segment of the market we go after. It’s not – this doesn’t cover workstation. It doesn’t cover big data, which I’m going to talk to you about later today. This is just the HPC Server Market. And the way I built the time, was HPC says the Server Market is about a $10 billion going 7% a year and I believe that about $3.5 billion to $4 billion of that is my [tam] broken out in the most important segments to us on this chart.
And then I looked at the applications that we have, the momentum that we have in each of these segments and based on that, I believe I have $1 billion sound, right. And again, you can see how a lot of it is in supercomputing higher education and increasingly in things like Biosciences, a Computer-Aided Engineering.
And in fact, if you look at our revenue, it reflects well into the sound. This is FY13 revenue first segment and you can see that we do a lot of business in supercomputing, oil and gas, higher education defense, and Jen-Hsun mentioned nearly every CT, high-end CT scanner that’s shipping in the market today is how the Tesla GPU in that, right.
So there’s a number of things that you can take away from this. One that is our revenue base has diversified from just being supercomputing to being many different segments. This actually helps reduce the lumpiness in the business. Two, we have very good commercial success, in other words, enterprise success, oil and gas, defense, these are commercial enterprises and we actually have Mark E customers in that. I will share some of the names in oil and gas in a few minutes, but we have proven value proposition with Mark E customers in these segments, and we have a number of emerging markets up in the right CIE, media and entertainment, finance, there the applications have just got an end place. So we’re beginning to see the revenue ramp. So what I’m going to do next is, I’m going to take two of these supercomputing, and oil and gas and go further in to show you some of the things that we measure to give us the health of the business.
First thing we’ll do is, look at the pipeline. This is a view of our sales force pipeline like we measure ourselves, pipeline if it’s 12-month pipeline at any point, it’s looking forward 12-months all the deals that we have visibility into 12-months ahead. When a deal closes, the pipeline goes down. When a new deal opens, the pipeline goes up, right. So you can see ups and down. And fundamentally, what’s have, so good example is in Q4, we started shipping Kepler, we had a bunch of backlog and that’s why you see a dip in Q4 in the end there. Right, we start shipping Kepler pipeline went down. We added new deals in the pipeline went backup. But the key message here is our pipeline has roughly doubled in the last one year right. And the number of customers we engaged with, the number of markets we engage with has gone up. So, number of deals – size of deals has gone up. This is a key early indicator of future success for Tesla.
This is another key early indicator of future success and Jensen shared this with earlier, but I want to reiterate one key point here. We have made extremely fast progress in the top 500 in the last couple of years. In particular Kepler has been a big beneficiary and we will continue to be a benefit. But, we are still in only 10% of the systems in the top 500 there is a lot of runway here for us.
The reason why these systems are so critical is that, these are leadership systems. These are the systems at the most important sides in the world. And they have a network effect, if you’re in Oak Ridge National lab, thousands of users access that super computer to run their applications. So, all the systems lead to more and more GPU applications being developed, they lead to more systems being put into place at smaller government labs or academia or industry, which also using the same systems. And really what’s happened here is, we talked about how the OEMs or shipping systems last year. But what you’re seeing now is the impact and everything in green here, all the names in green are new over last year. Another reason why – Dan did a great job of this, there is a lot of value to these folks for GPUs but I still want to share one example and this is an example for an application called Amber, it's a very important research application, it essentially allows you to simulate biomolecules. So a good use of this would be to find a new drug candidate and figure out how – what the efficacy of this drug candidate is with a virus or bacteria, so you know they try to match them up.
It turns out that just to stimulate 46 nanoseconds, and a nanosecond is a billionth of second, 46 billions of a second in the life of this biomolecule used to take one full day of computer simulation time on 96 overs, right? 46 billionth of a second in the light of a biomolecules, whole day with computers. You can now do that using a single Tesla K20 GPU, 1 K20 GPU now gives you 81 nanoseconds and 81 billionth of a second of stimulation time in a day. You can put this in your PC, the reason why this is revolutionary is getting access to one of these is difficult, the big computer, getting access to one of these is drive down (inaudible) right? That revolutionizes the life of a researcher.
In my opinion it fundamentally changes the base of research and that's the reason why we're seeing such good adoption of GPUs at these leadership sites. So I talked about super computing, let me now talk about oil and gas to give you an idea of what’s happening there. I’ve shown a similar chart to you last year, the key point I want to make here is, that we now have critical mass in the oil and gas industry. I’m sure a lot of you have worked with the oil and gas industry in the past and you know very well that enterprise businesses take a long time to adopt the new technology.
We started working with Schlumberger in 2009. The way these guys work is, they put in a small system to kick the tires, then they put in a prototype, then they put in a pilot, then they put one application on GPUs right in production, and so on.
And they continuously add more applications as their confidence in the new technology builds. And that’s really, this is another way of looking at a pipeline, but anyone who just tried using GPUs, will make real money from them in a few years, right. So this is another view of how we are making progress and are at a tipping point in the oil and gas market.
The other thing we do is, once we enter a market, we actually really try to understand it. We have vertical specialists, we call them industry business development experts. And what they do is, they deeply understand the market with their technical folks who work with them, find out what are the challenges in that market, and how can our products bring value, what additional things can we do to bring value to that market.
Oil and gas is a beautiful example, we started with Quadro by enabling large 3D visualization. We figured out working with them over the years that what they really wanted to do, was to visualize terabytes and petabytes of data. They have ships going in the Gulf of Mexico with all kinds of sensor data coming back, and they want to be able to visualize the volume. In fact, one of the demos right there, the middle one, is exactly this demo where you can visualize an entire dataset and that’s the third box there the remote large 3D data rendering. So we developed a new middleware software called Index to enable this. Similarly for Tesla, we started in seismic imaging. We’ve now developed a new library that enables them to do reservoir simulation and we expect that to come online next year. And of course, GRID it has huge value for this market.
So I only wanted to touch on two of the segments in my dual pipeline oil and gas and super computing to give you an idea of what are businesses and how we are growing the business, how we are actually at a critical mass in some of these key verticals. But let me transition to this third growth driver, which is apps.
So this is another indicator or another measurement that we do, which is what are the apps that on GPUs, what are the number of users or what is the kind of installed base of these applications. And really the key message here is we measure and we are constantly tracking the most important applications.
We are working with those developers. We have just like the gaming group has technologist group that works with the game developers, the Tesla Group has a bunch of technologists, all computational scientists who work with these application ISVs.
And the idea is once you have these applications then you can go after the users, but that has another challenge even after we have Adobe CS, which has roughly 3 million users, we still have to reach out to those users, show them the value of the GPU before they’ll actually start using GPUs. And that’s some of the hurdles that we face as we grow the business, so let me share with you one way in the addressing that particular hurdle what we found really early on was that when people try GPUs, we end up buying them. Simple, right? If I give you a GPU and I say, here’s ANSYS. Try it on this GPU, on this system or the GPU. It turns out the guy sees speed up, right? ANSYS is a good example. A product designer will fundamentally be able to design a higher quality product if they can run 2x faster, because they can run twice the number of simulations. So their simulated model is so much better.
The IT manager is happier because he is getting two times the throughput out of his data center, huge value. He is servicing his users with 2x the throughput, right? But we have to get them to try it. And so we lost this program that we called the GPU Test Drive, really simple, GPU is in the cloud, you can access them, try them and if you like them, you can buy them.
So what we do is we measure the number of people who are registering to try GPU Test Drive, right. So that’s another early indicator of future success because as we get more registrations, the more people try them, we know we’re going to have a good pipeline of run rate business out of these apps.
So this is one of the things we do. We do a lot of things to get more application momentum. I just wanted to share with you one. And we continue to really, really focus on developers. The way I see it is developers leads to applications which leads to revenue. So developers is the base of everything we do. GPU Technology Conference is all about the developers and we measure it in many ways.
These are some of the measurements, the number of – the accessibility of the GPU. The biggest competitive advantage I have with Tesla is GeForce. The fact that you can walk down to a [price] and buy a $200 GeForce or it’s already in your laptop means that you can start developing for my product, much, much easily than any other product out there, right.
We measure the number of downloads, every minute CUDA is downloaded, one download a minute. We measured the number of supercomputers and we’ve measured the number of people learning parallel programming the GPUs. So, a good example is there is two new massively online courses from Udacity and Coursera on parallel programming with CUDA GPUs.
In the last six months, 50,000 people have registered for these two classes. In the last six months, 50,000 users have registered for CUDA, learning CUDA through these online classes, right. It’s a terrific momentum story there. So, I guess, I will transition into the fourth growth driver for us. And I’ll start by saying that we’re moving beyond HPC into the world of big data. So Tesla is entering the world of big data.
Now this is important to understand. I think I’ve been talking to a lot of you last night at lunch and even during breakfast today. And the one thing I kind of got feedback on is, aren’t you guys just HPC, it turns out that we started in HPC, because we were a startup business unit and we still are a startup business unit, because that’s the market where it’s easiest for us to get adoption, the high performance computing market likes higher performance processors best, right. It’s easiest to go, sell it to them to show them the value.
But fundamentally, we have a horizontal processor, which can accelerate any application that has a lot of data or a lot of compute. And this emerging market of big data is actually just terrific for GPUs. So what I have is four different examples in big data where GPUs are applicable. The first one is analyzing Twitter and Jen-Hsun spoke about this briefly. So fundamentally, 500 million tweets coming in at any minute, sorry 500 million tweets a year. And the ability to just analyze these tweets and figure out patterns, look for sentiment, be able to do all kind of analytics on it, is what Salesforce is doing using GPUs.
Shazam, we just tried that and you can see Shazam has – Shazam is growing extremely fast in this market, but they’ve experienced a 3x growth in the number of users and 3x growth in the database. The Shazam CTO was at GPU Technology Conference and the key insight he gave, which I loved was, he said, we used to have ideas of new services, new applications we wanted to implement, but the infrastructure cost was too high. GPUs enable us to get to have scalable growth to be able to implement these new ideas, right.
So that's the second one. So twitter analysis, audio analysis, audio search. The third one is video. It turns out – anyone a sports fan here? I am guessing there are sports fans here, right? Yeah, there is a few, excellent. Rob is sports fan. So if you guys go to ESPN.com and watch any video there or if you have the ScoreCenter app on your phone, right, any video in ScoreCenter, it has been transcoding using NVIDIA CUDA GPUs. And this is actually technology that comes from a company called Elemental Technologies. And I don't know if I need to tell you this, but video is exploding on the web, it's exploding on tablets, one of the more – after gaming, watching video, I think, is the second most popular thing on tablets. And some of the statistics here are just amazing that the volume – new streams of volume of people watching is so high and GPU's are in the middle of this whole thing.
So the fourth thing that I skipped over is visual shopping. So Twitter, analytics, audio and imaging. So this is more computer vision imaging where Cortexica is building – has built an application that is a smartphone or a tablet’s application. You take a picture of the nice shirt this gentleman is wearing and it can match it with a database of all the shirts that eBay has and shows you all the shirts that are similar, right? Extremely valuable in moving the sales cycles from text based search to visual search. And research has obviously shown that when shown in image you are more likely to buy, okay?
So, four very good examples and diverse examples of how GPUs are being used for big data. So it turns out, Rob asked me, hey, do you have any other prospects for big data except the four you are showing that you already have, I have four customers and he said he wasn't satisfied with four customers. So I made a list of prospects. I was able to find a few, Rob, it turns out there's are a few more guys out there doing big data, okay?
So we think it's a nice big market opportunity that's opening up to us. We intend to move forward and go after this market opportunity as the business matures. So I talked about four growth drivers today, Kepler, our deal pipeline, applications, and big data.
So that concludes my talk. Thank you.
Thanks a lot, Sumit. Okay, we are on schedule, we did get the feedback also to, I think, we got some feedback that’s a little cool in here, so we're going to tend to heat up a little bit. We're done with the final session, prepared session. And I think Jen-Hsun had talked about a number of things that are driving our business that obviously, the financial results of NVIDIA. There’s a lot of components that go into the financial results of NVIDIA and our investment strategy. But it’s in no small part due to the work of Karen Burns and the finance team. Nothing in our financial result happens by accident.
All right, it’s all very, very important work done by a lot of people who really know their stuff. So with that I’d like to introduce Karen Burns, our Vice President of Finance and the Interim CFO of NVIDIA.
Thanks. Don’t clap. So I guess I don’t have to introduce myself, but it’s really exciting for me to talk to you today. So I appreciate this time. So I’m going to talk about three key things today. And that’s delivering exceptional financial performance, investing in our business for future growth beyond the PC. And lastly, generating cash and returning capital to shareholders. And as you all well know these three things don’t run each other.
I’m going to show you how our growth that we delivered in the last three years generated the cash at $2 billion, which we are now going to return $1.2 billion of it to shareholders through the end of this fiscal year. And we will also invest $1.6 billion back into our business. And two key strategies that we’ve identified for further growth. So what I hope to achieve today is just to show you how NVIDIA operates, and shed some light on what we’ve accomplished in the past.
Let you know again what those investments we’re making today for the future. So you can see the same path, I do for how we’re going to grow. So this is our achievements over the last three years and I think what’s really key here is what Jen-Hsun talked about earlier, our three strategies. This is foundational strategy. He talked about visual computing leadership. First and foremost, we are going to push the technology further and further, because that’s what our consumers want. That richer visual experience gets some excited about our next generation time and time again. It’s core to our business. And he also talked about making that applicable to GPU to more and more applications. That allows us to expand our TAM. That’s what Tesla is all about.
And then finally, he talked about devices. It’s all about volume. We’ve got to get into more and more markets. And we have all the tools to do it. So, over the past three years let’s talk about what we did and I’m going to show you that’s the same strategy we deployed and that’s how we grew to this extent.
In absolute dollars, on the top line, our greatest growth drivers were in GeForce and in Tegra. And in GeForce, that was starting with Fermi in FY ‘12, building on Kepler FY ‘13 that GPU leadership that we obtained through those years, grew our share, grew a richer product mix, growing our ASP. Now it’s more than just the GPU. Fish talked about how important it is for us to engage with our content developers in the community. We’ve got to push the content, push the computing, bring them together for more excitement to the market.
That’s how we build the momentum and get people excited about our next generation. They keep buying and buying. And I’m pretty amazed I really want to meet a gamer because I don’t have $1,000 to spend on a GeForce GTX. I mean it is amazing the amount of money they’re willing to spend. So there is an insatiable appetite for the level of technology we’re bringing up. We’ve learned that visual computing matters to people, what things look like.
So that’s GeForce and that’s how we grew. And then the other key component was Tegra. And if you can remember, in 2009, we decided to exit the chipset business. We grew up from 0 to 800 million, we stopped, we took those engineers, moved them to our mobile division, we’ve licensed a lot of our core technologies to Intel for $1.5 billion. And what a great decision in hindsight, because look at our mobile business today versus what our chipset business would be doing now and FY ’12 with Tegra 2 as you’ve heard today, we grew our revenue to $340 million, that was just in smartphones and tablet devices. From 0 to $340 million and then in FY ’13, we grew another 50%, same two devices, smartphone and tablets. And as Jen-Hsun mentioned, more devices, more opportunity for us; more units, we can sell more volume; more volume, more revenue. So bringing that to our business was one of the best things we did. The only regret I would have is that we didn’t have Icera here sooner that we didn’t invest in her, we’ll invest it soon. If anything, our only regret is that we don’t have LTE today selling in the market.
Nonetheless, we grew gross margins in the past three years 13.3% and people asked me all the time how did we do it Karen, even within our own company? It is pretty amazing and a key component is definitely what we talked about already, GeForce. That is 9% of that total, huge. But it's more than just on the revenue side. We have a whole team and throughout our company that has to – that has to touch on this number
R&D and process technology are working together before a new process note even comes out. So we can hit the blocks with the best yield possible. Then they're driving those yield improvements all the way through maturity. Our operations team is relentless. They do assembly and test cost reductions, packaging reductions, shipping cost reductions everything.
Next is EPS – OpEx. We have a really tight focus on OpEx. We have precious resources; we need to use them wisely. We’re fighting against large competitors in a big field. That grew our EPS, that management 38% over the last three years.
So the next couple of slides just summarize what I talked about and I just wanted you to have them for your reference. And then for a visual, Jen-Hsun shared with you this chart and you can see as I talk through it that the growth coming from Tegra was really, really strong. That’s why it’s a key, key component of our company going forward. Now, all of this happened and if you think of it from year-over-year while the PC industry declined. And that’s why it’s so important why we’re so focused on growing beyond the PC.
Many of the companies in our fields decline 4% to as low as 17%, and we grew in this past year, year-over-year 8% on total revenue. Pretty amazing considering not only PC decline, but our HDD supply, as you recall, the shortages that we had and 20 nanometer supply constraints. So you can see all the headroom that we have going forward.
So, how are we going to be strategic about our investments? Well, Jen-Hsun shared this slide with you. I just want to make a couple of more key points. And it’s starts with because we have two main processors, the GPU and our Tegra. And the GPU’s core, it’s our foundation, it’s what we have to invest in continually to push the bar, it’s why we are still relevant, that’s why we must invest, pushing the bar as they show of computing. It’s a full scope team. Software is so important to this business and it allows us like CUDA and the example to go into new markets, we’re now serving in supercomputer. So world’s leading supercomputers were in, that wouldn’t have been possible without CUDA. I think it’s pretty amazing that we took what was once used just for graphics for general computing extending the life of our GPU and it’s relevant.
We will continue to invest in it, our software is key that allows us to go into so many more market places taking that same GPU and to GeForce that allows us to go into the professional space with Quadro and Tesla as I mentioned before.
And, of course, another key – the other key component of our business, our processor is Tegra. We have to invest in Tegra. The time is really now, we can’t let off the gas. A $300 million investment with LTE coming around the corner integrated opens up a much, much bigger TAM. That is the volume more devices that we need access to, it’s more than just smartphones, we need to go further into the mainstream. It’s that volume that we need to capture. Now that same processor, we’re going to take and we can bring into many other types of devices, more devices, more units, more volume, more revenue, auto, is an example. When you hear auto, it’s not a brand new market to us, it’s taking the same processor. You don’t even need to do another tape out, it’s the same processor going into a different type of device that’s leverage.
And we’re going to take Tegra and put it in the SHIELD, because the people that buy those $1000 GTX Titan cards, they actually like buying things. And this is another thing to buy, another fun way to play and do what they do. So we think SHIELD is a great opportunity for us and we’re going to go for it. We believe in it. And then GRID, GRID is an amazing opportunity. Everything you hear cloud, cloud, cloud applications, we want to be there, we’re relevant. We have something value both to provide, our same GPU core. So it’s the same processor.
Now, as Jen-Hsun said, it’s just – think of it as, there is a lot of that in cards all set together in an IT room. And now you can access it on any device anywhere. So that’s our strategies as people have this explained today, our team and that’s where we’re investing in our future.
So this is just a summary, again, I wanted to, I know there is a lot to take in, so I wanted to give something I refer back to. And so lastly, it’s really the cash, the generation of our cash that growth that we fueled in the past providing the $2 billion that we have today, just over the last three years, we have $3.7 billion in the bank, it’s a record. over the same timeframe, we grew three cash flow $2.4 billion, if you think about the $2 billion just over the last three years, that’s $2 million a day generated from our business.
So we feel very confident about it, and we’re really excited about the opportunity to return more to shareholders. We’ve grown as well. You can see the U.S. cash balance. So we’re going to pay $1.2 billion out with no further tax implications. Just to give you a sizing of what this return is, it’s 15% of our market cap. It’s 32% of our overall cash and 67% of our U.S. cash balance and we have an ongoing commitment to return more cash to shareholders.
So, just to summarize, I hope I’ve shed some light as to looking in the past at our great performance over the last three years, how that happened, how we used the fundamentals of our digital computing leadership, our applications, porting to more to make it more relevant and more and [be a] more devices. And that’s how we chose where to invest, how to invest to grow our company further. Thank you very much.
Unidentified Company Representative
Thanks a lot Karen. I think you’ll have to stay up here. That’s our prepared material. So, again, we wanted to make sure that we lead time at the end. So can I ask all of our executives to please come up to the front? We’re going to have microphones. All of the presenters, all of our executives, please come up.
I think Chris and a number of people are going to be walking up and down with microphones. If you can wait till you get the microphone before you do your questions and we make sure that everyone hears the question okay. So let’s see. Who are we missing?
Unidentified Company Representative
Unidentified Company Representative
Hey, Jen-Hsun, are you coming up? Okay.
Unidentified Company Representative
Unidentified Company Representative
You count as an executive. Okay, all right, let’s kick it up. I think we’re all here. So who wants to go first?
Shawn R. Webster – Macquarie Capital Inc.
Yeah, thank you. Shawn from Macquarie. Two questions, on the Tegra side, Jen-Hsun, I think you mentioned that you thought it would or you’re hoping it would be flattish this year. If I take what you had guided for Q1 and kind of extrapolate that forward to the second half programs you expect, [I guess], like 100% sequential in Q3 and Q4? I was just wondering what markets or what things we’re going to drive that kind of growth to get an overall flat year in Tegra? And then I had a follow-up question.
Android devices – Android devices, Win RT devices, between the low point of a transition and the ramp up for each one of our generations. If you look at Tegra 3, doubling is not atypical from the bottom of the transition to the transition the ramp up difference I should say.
I’m not surprised by that. It is the case that because of the strategic decisions that we made to pool in Tegra 4i by more than two quarters to pool in Kepler graphics into mobile earlier. And to as a result delayed Tegra 4’s ramp by about a quarter and we are going to have that extra this year we’ll have the ramp will happen a few months later, but the ramp will happen. We didn’t win all the design wins but there is whole bunch that we did win and Tegra 4’s performance capability is un-denied. I mean, nobody argues that it is the best performance on processor in the world today.
And so there are devices that make sense for it. Tablets makes sense and you are going to see a whole lot of new android devices that you didn’t see last year. I wanted to be clear with you and there is nothing I can announce right now because we are the component suppliers, we are not in this case the device themselves. If it’s a device that we may Shield we are delighted to more than talk to you about it. If it’s a device like GRID, we are more than delighted to talk to you about it.
We need to be thoughtful about other people’s products and globally there are android devices being made in consumer electronics, in consuming, in computing, in TVs, in PCs and Android is not going to be just a phone operating system, Android is a computing operating system and so it will just hang for just a little longer and the only thing that I would change is the word hope, I didn't say hope, I said expect and so there was...
Shawn R. Webster – Macquarie Capital Inc.
Thank you for the correction there. Maybe on in terms of on the back of the IDC announcement yesterday maybe I was wondering if you could share with us what you’re seeing out there in terms of inventories, maybe what your customers are thinking in terms of the overall slope of PC demand this year, if we are going to have a recovery in the second half for example what would necessarily give the catalyst for that to occur. Thank you.
Philip J. Carmack
We guided to a rather muted first quarter already and so far nothing that we are seeing is inconsistent with that. We’re not immune overall to the overall PC conditions; however the segments of the marketplace that are being disrupted by the tablets tend not to be where we play anyhow.
When people say they are not buying net books anymore, they are buying tablets, we're not buying, that is another way of saying they are not buying, why buy a cheap PC when you can buy a good tablet. I think that that sensibility is going to last quite a long time, I meant there is it makes a lot of sense I think to you guys, it makes a lot of sense to me, cheap PCs tend not to be what we participating in anyway, we are in the premium PC segment, we're in the segment of the marketplace where you buy our computers because you need to do some multimedia work or you love using photoshop or you like editing videos or you like playing games, those are problems that a table can solve today and so we're in the segment of the marketplace that’s on balance are not as much as disrupted by the tablets today. But we’re not immune to Windows 8, it’s adoption or overall PC market conditions. But we’ve guided to that and so far it’s not inconsistent with how we see it.
Jen-Hsun, this is (inaudible) Research. Jen-Hsun, can you talk about a little bit about the coming up transition to 20 nanometer playing a process, why would it be better or worse than the last transition we had at 28 and maybe just a tad towards the movement to 60 nanometer FinFET, how big a challenge would it provide the manufacturer and yourself and whether that could be a stumbling part?
The way to think about that is I learned two things in 40 and in 28, one thing each. In 40, we learned that we needed to set up a process R&D division group. The process R&D group is intended to work with the foundries years and years in advance. Do test chips; map the process into our cell library design style. Our cell library design style tested against their process, working closely with their R&D team.
I think the foundries that we do work with; I think TSMC would tell you we are world class there now, absolutely world class, I mean the best in getting process readiness, notice 28 ramped up in yield beautifully. The challenge on 28 was the underestimation of supply that in combination between us and TSMC; us being the industry and TSMC could have done a better job.
We do a better job there today. We share forecast, roadmap alignment on a monthly basis now between our planning group and theirs, but in addition to that, I learned the other thing, which is we need to be thoughtful about putting too many projects into each node, all at the same time. Ramping your product line from top to bottom overnight is probably unnecessary and something that we have to be more thoughtful about.
Just facing it in a strategic way, smart way makes a lot more sense. It takes a lot of pressure off of them, it takes a lot of pressure off of us and it de-risks the entire scenario much better. And I hope that applying these two lessons to the next two nodes will be helpful to us, but we’re running test chips and we’re doing process, readiness development with them and we have a deep understanding of each others needs and capabilities of those nodes already and for quite sometime, for quite sometime.
we have a team that’s completely dedicated to just doing this, process readiness.
We didn’t use to have this team. we used to treat it as part of our product development and that’s a mistake, product development tends to be looking at a horizon that’s about two years in advance. We need process development to be looking four, five years in advance; four, five years in advance; three, four years in advance, two, three years in advance and then one year in advance.
And so that methodology has been set up inside our company where I'm just incredibly proud of the work that they've done. So we’ve learned those two things, and I’m hoping that that those lessons will carry forward and be helpful to us and I expect them to.
Good afternoon. Thanks for a great day and Rob thank you for turning the heat up. Jen-Hsun you mentioned that the time is now, Karen you said in your presentation time is now, and you said that you can't look to put off the gas in terms of spending. As you guys look forward, do you see that you didn't have to be spending at discontinued clip or do you see the opportunity to be able to take your foot of the pedal just a little bit.
I think that our current level instead of ramping into growth investing just a little bit too late. Where we are right now, we had to step up and then hold. And so these step ups and hold allows us to build up the engineering team and let the engineering team go to work, instead of large being two or three engineers behind, they never catch up.
And so we’re stepped up at this level and I'm comfortable with this level. The current level that we are at allows us to invest about $300 million of incremental dollars in to Tegra, which allows us to contribute earnings, Tegra is contributing at $600 million, last year at $750 million it contributed, because the quarters are lumpy, it doesn't contribute every quarter, but that doesn't matter, they will contribute this year, so Tegra is a profitable business now.
What we now need to do is to allow them to continue to deliver this multichip family every single cycle and crank at a few cycles. I think our position is going to be stronger and stronger and stronger. Tegra 2 we only did one chip per cycle, Tegra 3 we did one chip per cycle, Tegra 4 we did two.
Next generation would do two plus a little and we better serve multiple segments. And then I’ll smooth out our business, secure our position. And take a little bit of guess work out of this whole conversation. And it allows to transition our business from our starter business if you will where Tegra is today to much more of a growing steadily growing business like GeForce is.
It’s I think that the investment level is good, it is profitable. I expect them to do about flat this year half of a year call it $750 million last year. It is a little bit that back end loaded but there are more devices this year, more types of Android devices and more customers this year. And I know that that many of these projects are highly secretive right now, but they are highly secretive by a lot of very secretive people. And I don’t want to be the first to leak it.
And so give us a chance to come in and surprise you and hopefully delight you.
We did surprise that you gave us the dividend. So thank you 1 by 2 times. I appreciate that.
Thank you, thank you.
And then one adjustment question on Shield. And so you can talk a little bit about that. so you said it was in production and then Q2, can you give us a little bit maybe just a little bit of insight in terms of what kind of pricing range you’re going to be looking out for Shield maybe? And thanks so much.
Yeah. Announcing the price of a consumer product is a big deal, it’s a big deal in the sense that the consumers get all excited about it.
And so, if I could just ask you to give me the opportunity to announce it in a way that brings joy and delight to a lot of people. It’s how these things are, it’s a consumer product. We are finishing our final tooling, I really love the final devices they’re even more exclusive than once that you are feeling here.
I expect that we will sell this device like consumer electronics device not subsidized. But because the content is so abundant and because the content is so affordable, this is the new model. The consumers decide what they love to buy, what they love to own, they buy it and then after that they can enjoy a world of a universe of content in a really affordable way, instead at the other way, which is you buy into a relatively affordable platform and then after that you spend $1000 on content, so the equation has really inverted.
I think modern consumers tends to like this new model. And so we’re going to sell this device for a profit. Now recognizing of course is our first device of its kind. And so, and also it’s a retail product. So the margins I mean going to be like the margins of technology. However, our margin dollars will be much higher than the margins we will get from Tegra for example.
Dean Grummels – Stifel, Nicolaus & Co., Inc.
Hello this is Dean Grummels from Stifel Nicolaus, thank you very much for informative day. Could you provide an update on Tegra 4 and 4i a design win pipelines, perhaps compared to Tegra 3 at a similar point in development?
I would say with the exception of one project that we just couldn’t meet the timing on, everything else is better. I feel like I don’t know how to answer that question without being specific. And was that a clever enough answer for you? I was…
Dean Grummels – Stifel, Nicolaus & Co., Inc.
Okay, that’s great. That’s fine.
Okay with the exception of, I would say with the exception of one that we couldn’t make the timing on, which is, it’s kind of a June product and we obviously just couldn’t make that. But shortly after that I think you’re going to find Q3, Q4 there’s a lot of fun projects coming.
Dean Grummels – Stifel, Nicolaus & Co., Inc.
Okay. There’s a quick follow-up. would you except that LTE carrier certification be a long pull in the tent for revenue, or do you think certification will stay well ahead of other, the usual limitations?
This year, LTE is not a critical path. The only, the i500s LTE is involved in this U.S. revenues, but I’m not expecting i500 to be a critical path this year. Okay, so they were, the i410, the last generation product, it is already certified at AT&T multimode LTE i500 software comparable, it’s going to get carryover, it’s faster, it’s higher bit rate, but architecturally exactly the same. So we’re going to, we’re going to fly; Steve is going to be up all night. But from my perspective, we’re going to fly right through it. Okay. And so there will be a lot of peers suffering, but I won’t know about it and that’s the most important thing. So I don’t think this year, any of the tablets will be affected by that. The tablets, the clamshells all of that is, I think very, very low-risk.
I’d frankly think that the phones in voice modem is low-risk, they went through it before. And it’s low-risk in the sense that I know they’re going to succeed and all the revenues are over for phones are kind of starting very late in the year and beginning of next year. So again from this year’s perspective, I doubt that it’s going to be a real factor, but the important thing is we’re going into next year. We’re going into next year with Tegra 4 (inaudible) Tegra 4i running and who knows what happens on top of that.
(Inaudible) at Barclays. Two quick questions for you, one is you talked about how the move to large screens and PCs was big driver in GPU adoption for you. When you look at the tablet and the handset market, I mean that you’re not going to have that driver to keep pushing larger and larger higher performing monitors. How do you see that playing out over longer term?
Well, just so I can take the rest. Phil, you want to take a swing at that?
Philip J. Carmack
Absolutely, I’ll take a swing at it. First of all, I think the analogy with the move to higher resolution on monitors was that it became more available to have a decent or pretty high quality monitor and as a result of that it became worthwhile to do the visual computing behind it that everyone wanted to have. I would say that that already happened on the mobile side. The resolution, the quality of the display and yes, even the size of the display is evolving. This device which a lot of people carry has a 4.7 inch display. Just two years ago, the Canonical device was a 3.5 or 4 inch display, VGA resolution. This is 720p and this year you’re going to start to see 1080p, 5 inch, fairly manageable size phones.
So I think we’re right in the swarm of that very environment that Jen-Hsun was talking about that changed the whole visual computing side of PCs. I mean, I remember the day when I bought a PC because it had 80 characters wide for my spreadsheets, right. It was characters. It wasn’t graphics. There was no UI. There was no rendering of any kind.
Then, let me tell you, it was a pain. It was a pain in the butt. And when it went to being able to do graphics, I was at the forefront of buying 12/10, 16/12, 19/20 right, 19/12 and now 25/16 and beyond. So there is a crazy escalation of resolution happening on all mobile devices. That’s great for us because we’re able to actually provide the experience that justifies that resolution and there is a crazy sequence of visual computing happening.
Unidentified Company Representative
And Christian, you know that – let me make sure that you do. You know that processing is not related to the physical size. Processing is related to the pixels, the number of pixels. And so look at how many laptops today are not 1080p and look at how many phones are about to be 1080p, right.
So the phones and tablets and these touch clamshells; these types of devices, the resolution is so high. That's where the processing goes. We process, the way we count processing, the horsepower and visual quality is operations per pixel, not operations per inch.
Thanks, that makes a ton of sense. And then just one quick follow up. Maybe that's for Karen. When you guys say you were profitable in the Tegra business last year, I think in your 10-K this year you showed a bit of an operating loss there. Is that just a difference in how you buck it, when you guys think about that business, how you buck in R&D in such and how the auditors look at it or I guess it was the delta?
Yeah. Well, piece of it is SG&A, because we were showing you just R&D. And the other piece of it is the different allocations and how we look internally, how we manage the business.
So for example, we’ve shown previously that Tegra – if you want to think about it from accounting terms, okay and we talk about this in inside, do we think about it as a business or do we think about it as accounts. Because when you’re accounting for it, you want to allocate some percentage of all of our salaries to it, even though we make no contribution to it, right, because that's kind of how people, they allocate things.
And so we used to allocate 20% of our IT department to Tegra, even though they had very little revenues. When Phil started the business and it was kind of – it was lunacy a little bit kind of the way we thought about it. I mean, when he started the business, he started on the first day with three employees. But he was already losing like $20 million a quarter. And Phil used to come to me and say, Jen-Hsun, can I not have any IT, please. I’m just going to run out to the fries and get my own laptop. Can I not have SAP, please? I don’t need to close my quarter. And I said, Phil, don’t worry about that. These are senior accounts in the – don’t worry about that. Well, so that’s – do you see the difference? If you look at it from purely business perspective, the incremental cost of Tegra is $300 million for us. I give film my services for free as I should, right. And so allocating me and HR and the head of HR, it makes no sense. And so, that’s the way you can allocate it versus purely incremental.
Now there is the other debate which is what would happen suppose you were to set it up as a completely different entity, okay. Wow! Then that’s a completely different question, because then he has got to go and start his GPU architecture team, and software team, right. And so all the things that we get for free like Open GL 4.3, DX 11, CUDA, OpenCV, Windows Drivers, Win RT, Linux, right, all of that stuff becomes expensive.
So we don't try to think about, what does he look like as a standalone business because he is not. We should think about him, think about Tegra as an incremental investment to the company and at what point is the incremental business; without Tegra we wouldn't have the $750 million that for sure, right. If the incremental business generates more returns than the incremental investment, then he is profitable. That's how we think about it.
Yeah, thank you.
JoAnne Feeney – Longbow Research LLC
Yeah, hi, JoAnne Feeney with Longbow; I was hoping, Jen-Hsun you could give us an update on the workstation business, what you expect it to do this year? How much of that is driven by the aggregate situation for workstation, how much is perhaps changing because of competitive pressures and whether you think that business disappears and instead moves to the cloud?
First of all, you asked three really good questions. They were just highly compounded. And so let me break it apart. On a baseline basis, I expect their CAGR to continue. More and more industries are moving to digital design. You’ll be surprised how many industries are not in digital design. A lot of people still prototype, carve it out of this, build it out of clay or don’t do anything, just build it.
There are so many industries that do that. There’s a whole maker movement recently as a result of the affordable CAD tools and affordable 3D printing. This whole maker movement, I don’t know if you guys have been following it. It’s such a big deal, okay. People aren’t like knitting sweater. They’re making motor cycles, medical instruments at home.
And so this whole maker movement has really democratized CAD and it’s a really interesting opportunity for us. But my expectation is that more and more people are going to do digital design, digital content creation, more Indies doing video game creation. All of that, that growth is going to continue to contribute to Quadro.
We don’t really see much competitive pressure, okay. We focus all of our energy not on fighting back the competition. But we have 90% share. You focus all of your energy on how to grow the market. And that’s why here; I don’t think there was on comment about the competition. It’s all focused on how do we create a larger market.
Now the Quadro business is interesting and the last point that you highlight is interesting, which is what happens to Quadro with GRID. What happens to Quadro with GRID? I think long-term; Quadro will be replaced by GRID at a premium. Quadro in a workstation is less value than Quadro in a server. ASPs are higher; it’s more useful to the user, to the workstation user because they can access it from everywhere. There’s a value in putting them into the server.
A PC CPU is a few hundred dollars. A server CPU higher value, the up time, the utility, the shareability all greater. And so I think that there is a possibility long-term and I think this is going to take many years for Quadro to eventually become GRID in the cloud, GRID in the server, GRID in the data center. In the middle what we're experiencing is this. They’re getting both. They are literally getting both and the reason for that is because while you’re at the office, where you need productivity to be maximum, unimpeded maximum productivity and that person is one of 20 designers in a very large division of Boeing. The last thing they’re worried about is cost reduction for that person.
What they’re worried about is how can I give this person a remote capability, so that he can continue to work or she can continue to work at home. Just like the marketing people can work at home, why can’t the engineer also work at home and why can’t that data be shared with someone who is actually in China? And so we see right now people installing for the same user, multiple nodes of workstations. One, very powerful one locally; one that’s remote. And so the dynamic that you described is hard for me to predict exactly. But when you go far out, if there’s one of the things that’s always easy, you just go out far into infinity. In infinity, I think it’s all going to be in GRID.
JoAnne Feeney – Longbow Research LLC
And then as a follow-up then, how do you envision your pricing strategy or your go-to-market strategy for GRID? If you’re going to sell hardware, would it be just hardware plus the chip price per, I don’t chip? Is it a software license that has a tail of revenues? How do you envision doing the pricing on this kind of device?
GRID comes in basically two versions in enterprise computing. It comes in GRID VCA which is an appliance and that’s $40,000, okay. And for $40,000, it basically replaces, gosh, 16 powerful workstations for a small medium business. On top of that $40,000, there is a per node, not per node, per GPU license.
And so on top of that, there is a software license that goes per year, okay. Call it 20 or somewhat percent, 20%, okay. And that goes per year. And the reason for that is because we’re upgrading your software all the time. We’re improving your software capability, we’re improving the features, we’re improving the performance of CS7 or we’re improving the performance for SolidWorks next, so there’s a lot of continued work that we do in the software to improve your capacity, your performance et cetera. And so there is a software license that goes with that. In the case of GRID that goes to servers that are sold by Cisco, IBM, Dell and HP, we sell them a graphic card, the MSRP is about $3,000 and it gets discounted down to an OEM, but some percentage of that, okay. And then on top of that, there is a software license that also is per GPU.
Okay. My question is on Windows RT and how you perceive Microsoft support going forward, how much you’ve planned to invest going forward, and then you guys had a lot of success last year in terms of getting the right sockets, but the market didn’t do what you want. What has to happen for that to become a sizable market?
Unidentified Company Representative
Well, I really hope that this time we get the right sockets and the market does what we want. Now that is hope. But my expectation, here is my expectation. My expectation is that Tegra 4 running the next generation operating system called Blue, we have it. It’s just fantastic, because Tegra 4 is just really high performance. It’s snappy, the battery life is even better than Tegra 3 and it’s hard to not like it.
I think Win RT, are we putting a lot of energy into it, like we do PCs? We think about Win RT like it’s a PC. It’s the same team that works on it, the Windows team. Same Vice President, same engineers, same DirectX, the next same DirectX 11, it’s the same everything. For us, Win RT with Tegra is exactly the same as a PC, as Windows 8 running GeForce in x86. In the future, nobody will know the difference. Just go forward to infinity and think, does anybody even know the difference between ARM and x86 or care, right. It makes no sense. And then you work it backwards from infinity and you’re cry meets somewhere around two or three years from now where nobody is going to care. They barely care now, right as we know.
Thanks for taking my question, I have two. One is in the mobile platform. Clearly, that’s almost ready to go into production. If I look at it, you’ve been very definitive about, a statement about that you believe you’ve got better performance than any competitor out there on the processor side and you’ve articulated how you then get there on the modem side. So I guess question will be going forward, are there any technique areas of that platform you think still needs to be improved to get a lot more market share than where you are now?
Unidentified Company Representative
Yeah, I think for Tegra 4 – the Tegra 4 family, the first and biggest thing I think is as we probably already mentioned, actually there are three big things that have to happened, right. We’ve got to get LTE into the market. Connected devices are becoming a lot more than just phones and even just tablets, right. Even the connection rate of tablets meaning wireless, cellular connection is climbing.
So that creates a leverage point for us, creates a tailwind to not only do well in the markets we’ve done well in since we have a better product, even better position than last year with Tegra 4 this year, but also to have the tailwind of people who are thinking of making these connected versions of whether it’s a TV box or an automobile or a tablet or a phone that they’ve got a good road map looking ahead with us with LTE. So we have to deliver on LTE. And I think the next big thing is GPU as we’ve mentioned. The thing after Tegra 4 is GPU. But between now and Tegra 4 products, other than just executing on products, there’s certainly no barrier to building the best mobile computing products ever built. We had the best last generation and we will exceed that lead this generation. All we have to do is get these products done.
Unidentified Company Representative
Now doing those two things as Phil says and I think I’m trying to predict what likely is in your mind. Doing what Phil says opens up our market opportunity by a factor of 20. Stay put, execute, do a few rounds, go a few cycles, play your game. I think you’re thinking in your mind, what about PMIC and audio and connecting and all that bundling. Maybe you’re thinking about that. In the final analysis, nobody gets locked out because of those things. Without LTE, when there is only one supplier on the planet, you’re going to get locked out.
The challenge with LTE is that unless you have an application processor, there’s actually no reason to build a discrete LTE anymore. And you could see where there is ST-Ericsson, or Renesis, or you name it, okay, all of the standalone LTE companies, because they can’t really successfully take their product to market have really backed away from it.
They just don’t have the economic engine to make that business work. And so you really have to have, you have to start with an application processor. Then we don’t need to go get ourselves an LTE. After that, everything else is really a margin story. There’s a few dollars, the delta between making it and bundling it, it’s a few dollars, fraction of a few dollars. So, it’s a margin story. We’re not going to let that get in our way.
Is it fair to say that once you’ve kind of conquered the modem, you think you’ll be fully competitive on both the processor side and on the modem side in terms of the complete system with it’s power consumption or everything else?
Unidentified Company Representative
Absolutely, yeah. Absolutely.
Unidentified Company Representative
And we have a really differentiated position now. The fact that we don’t have one and we’re in the gain, tells you something.
Sure. Can I just ask a follow-up on the notebook business? I think you talked about the fact that you think you interact, at least in line with notebook PC shipments this year. Can you give us any kind of color on what you’re seeing in the market, whether that’s geographies or high end, low end or what if anything else that will help us understand whether you’re going to drag above or below the market and where are the risks of that?
Unidentified Company Representative
We’ll probably – I’m sorry, Trish. I didn’t see you standing there. Go ahead.
Unidentified Company Representative
Jen-Hsun had mentioned earlier that GPUs go into more of the premium notebook segment. I commented that I believe based on our share gains later last year and what we see through the next cycle, we should be able to – I’ll put this as my expectation is we’ve outperformed the total notebook market this year. I think we’re in segment that is probably less vulnerable, not vulnerable to the market in general, but less. And I think we are still benefiting from Kepler and our share gains as results of that.
Thank you. Just a question on margins and I don’t know if you have any specific margin targets in mind going forward and perhaps more specifically as you ramp some of these new programs, you see ramping in the second half and in 2014, what sort of margin implication do you have? Anything that it points out as kind of above or below your corporate average at this point?
Unidentified Company Representative
When our gross margins was 35%, I was called a lunatic for thinking that it ought to be over 50%. I’m about to tell you something that’s (inaudible). There is $10 billion worth of market opportunity associated with GRID that should be very high margin. There should be very margin and there is a software component to it at on top of that. I think (inaudible) about earlier.
And then there is the Tegra component of it, the other $10 million, which is related to, I think it’s probably a 50 point business. And so now the question is which one grows faster. The GRID is probably between net of all the software licenses, it’s probably a 60 – it should be a Quadro like business, so we call it 60 semi-percent, Quadro is running close to 70%, I assume it's running close to 70%, okay. And so it should be like that. It's a server business. The Xeon is not probably anything less than 80 points.
And so we're not a commodity Xeon, we're additive to a Xeon. You see them saying, so we're actually and we're not as big as an opportunity as Xeon. And so we're a premium server opportunity, a differentiated server. So if you call that 70 point margin, we can probably live with that for the group part of it and then you've got the 50 point margin business for Tegra. That's kind of how I think about that.
Okay. And just as a with respect to the automotive market you made some comments at the GPU Tech Conference, talking about the modules, you were just flying in and out, can give a little more comment on that, I mean how do we think of that with obviously the module is going to be higher revenue, there is some passthrough content on that as well, and how material is that in part of the Tegra revenue today?
We have $2 billion worth of design, over $2 billion worth of design wins in auto. It's ramping up, it's approximately doubling and close to doubling each year, somewhere between 50% to doubling each year, 50% or doubling each year. It ramps up to about $450 million in a couple of years, and we like the vast majority of that business to be modules, not just because of the revenue components of it, but because it is really hard to design these computers into each and every car differently. Everybody's car is different, if you designed one module and you certified it from the conditions of death valley to the worst tundras on the planet then you can cover every car, and this way it allows us to design it once, qualified to the ultimate extremes, and then over volume because it's the same volume module we can manufacture quality into it.
I think more and more car companies are recognizing the wisdom of that because computers are hard to build and you don't want to build a computer for 50,000 cars. When they talk about high-volume is because it’s high-volume in revenues and they multiply it by a $100,000 and we are also in the Lamborghini and so they multiplied that by, right, so it's a high-volume business. But to us, it’s still looks like 50 computers. And you don’t want to design a custom 50 computer. And so more and more of these are going to be modules and that's likely the direction, it's like a little computer on a little module.
Steve Eliscu – UBS Securities LLC
Steve Eliscu with UBS. Thank you. I want to ask question about to drill down a little bit on operating expenses, if I look at your R&D expenditures and divide by the number of engineers that you have, the number has been fairly consistent over the last few years. So in terms of thinking about managing OpEx is really worth discussing about your process for deciding your engineering head count. Can you give us a sense as to how you decide what is the right number of engineers you should have to do the projects that you have? And then also because that number is flat and you really haven't been gaining efficiencies even growing your organization focusing around graphics and Tegra. What are you doing at least looking forward potentially increase the efficiency of R&D?
Unidentified Company Representative
Well, efficiency of R&D is not measured in dollars per human, right, it’s really measured in – and it’s not even revenues per human. Because an engineer’s productivity is not measured by how successful the sales team and marketing team or the business team is in promoting or successfully achieving [same wins] not directly related. It's related, but it's not only related. And so it's really about the productivity of their ability to generate complexity of products and some measurement.
Whether it’s – I hate to use line of code because the best code is small. I hate using transistors for engineer because the number of transistors going up for engineer is actually deaf right it leads to a bad place. And so it’s hard to exactly measure that, it’s hard to exactly measure that. I don’t completely agree that our engineering cost is about the same. I mean it’s been pretty competitive in the Bay Area. Between Apple and Google and Facebook, the best engineers are the best engineers and they are sought after by all the best companies.
And the four companies that I’ve just mentioned, the four of us compete for engineers almost every day. And so I think that it’s a – the first thing is to get the best possible engineers, build the most exquisite products and productivity is a harder measurement, because we are not laborers. The best poem is likely the shortest one. And Muhammad Ali, as you know, two words, Harvard commencement speech. Do you remember that?
Highly productive, fantastic poem, right? And so I think in a lot of the work that we, the exquisiteness of the product matters. Kepler is more exquisite than Fermi. It’s fewer transistors per unit of functionality. And so if I measured them based on any kind of metric that I know that’s kind of linear, they were less productive on Kepler than they were on Fermi. But there is no evidence of that.
Steve Eliscu – UBS Securities LLC
I have a follow-up question, you didn’t really talk too much about Maxwell, I guess you touched on GTC and not to get so much into product that in terms of the potential problems it will solve, can you give us a sense as you look a couple of years out where we think for consumer graphics and what the developers are asking for in terms of enabling capabilities as well as on the professional side and the kind of performance that’s required. What are you looking at in terms of the key problems that you’re trying to solve?
Unidentified Company Representative
I must tell you, you could almost guess. And the reason for that is because our company is so well aligned. Okay, anyone of the people from NVIDIA in this room would give you exactly the same answer. Number one for Maxwell, there is likely something that we’re doing that breaks new ground in visual capability, something that is even more beautiful than infiltrated.
Okay, number two, it is likely that Maxwell breaks new ground in programmability, ease of programmability, because we want to expand the general purpose nature of the processor without sacrificing its speedup relative to a microprocessor. For example, if our processor was incredibly easy to program, but after you port an application to our processor, it was only 1.2 times faster than a microprocessor. Then we’ve just caused a lot of people to waste a lot of time, because they now have to buy two processors, and is only 1.2 times faster.
So perf-per-dollar, perf-per-watt, perf-per-effort, perf-per-anything has gone down. We need to make sure that we continue to be five or 10 or 50 times faster. But make it more and more – make it easier and easier to program. Okay, so expanding the reach of our programmability, more applications would result in larger market. We’re in the GRID store, remoting virtualization, CCUs which is concurrent users all of that needs to go up every generation like most law now.
Now that we have the basic infrastructure of this product, we know what’s the measure of goodness the concurrent users per unit of something. Now that we know what to measure, we know how to improve it. And once it’s like clock or once we measure something it doesn’t matter what we measure just keeps going up, yeah and so GRID it’s going to happen. And then last thing, the energy efficiency of Maxwell, it’s going to crush Kepler. And the reason for that is because Kepler crushed (inaudible). And because we know exactly how to measure it now and we know what it means to be good. The engineers know exactly how do keep better. I just described, our GP strategy, our GRID strategy and Tegra strategy.
Steve Eliscu – UBS Securities LLC
Thanks. Jen-Hsun, can you give us an update on the plans for the million square foot campus that you announced couple of months of ago, across the street from your current campus. So in particular, million square feets is a lot of space, so what are you going to fill it with and what does that mean in terms in terms of your headcount projections and OpEx projections going forward.
We are going t built a
lot of space. So what you’re fill out the work and what is that mean in terms of your headcount projections and OpEx projections going forward. We’re going to build a million square feet in two phases. We’re committed to the first phase, we design the campus, so that even if we decided not to build-up the second phase, it would not be an eyesore to the city. The first phase is 500,000 square feet. There were several things that we, first of all we need to space. And as you very well know, I don’t actually even have an office anymore, I’m a vagrant and Mary and I just, we sit where there is chairs and it’s not the worst thing in the world, but I think just as a human rights issue, I think CEOs have to have at least a cubical. And so I think, and there are many others like me, there’re many others like me. Most of my e-staff]they are, well, okay they have human rights issues, let’s just put it that way.
You can only take that so far. We need new space. We decided that we would design a space that makes it possible for us to collaborate in a way that we do work today. If you look at the way that NVIDIA has grown, we grew into the campus of these leftover offices at San Tomas. And we’re in these, here’s a building with 150 people in it and here’s a building with 150 people in it and we find that in today’s work environment, architects, design engineers, software engineers, VLSI engineers, technical marketing teams, even marketing teams really need to collaborate in a much deeper way.
We wanted to therefore create an environment that is a large single force base and that’s why we’ve seen the NVIDIA design. I created a building that is quite large in terms of the actual fore space, it's very – it's not tall, and therefore we could reduce cost, it’s true storey’s only, it's actually interesting when you build something that's only true storey’s, the cost of the building comes down dramatically. And also because this is only true storey’s I can use natural light instead of using recovering energy wasted, let's try to use less energy in a first place. And so this building is going to be naturally lit a lot of times.
While I was sitting here, I was sitting in the back, one of our engineers is back there simulating our building. We are simulating our building. We use iray to photo realistically simulate the permeability of the building, and how light is bouncing off the skylights and the strategically placed windows and mirrors to redirect light all over the building without having to do this, okay. Wasting energy when you don't need it to, and so we’re using simulation technologies to ensure that for the vast majority of the time this building doesn’t even have to be powered.
And so there is all kinds of things that we're doing to be smart, so that we can enhance our ability to be collaborative on the one hand, keep the cost of the building down, don't waste the bunch of technology to do things, simulate it, so that we can come up with the most efficient design, use natural energy, don't try to recover it, which is a very expensive, don't waste it in the first place.
And so we are designing a place that is from an engineer’s perspective very sensible. It's not going to cost very much and it’s for all the people that are, it's going to house, it's going to keep ourselves OpEx neutral and so when we sign up for the first part of the building, and we'll think about the second part as time goes.
Anyone else? Do we get them off?
Well, guys I really enjoyed the day. This is one of the funniest things that we do. I hope that we were helpful in helping you understand where we are in the business, in the first half, why we are optimistic and enthusiastic about the second half, number one. Number two, I hope that we’ve shed some light into the strategic choice that we make with respect to Tegra and that I hope you see now the wisdom behind the choice, although there is a one quarter hit, we ramp up one quarter later that from a positioning standpoint this company is as a result of that choice, quarters ahead of where we would have otherwise have been.
And then lastly, our strategic investment posture, how we think about our core investment, how we make the incremental investment into Tegra and GRID and based on the numbers we've shared with you the fact that Tegra is already profitable on an incremental revenue, incremental investment basis and that the two Tegra, GRID, an additive to the GPU has opened us up to a market opportunity we estimate at some $26 billion.
If you guys have any other questions, we'll just be wandering around and it's great to see all of you guys. Thank you very much.
Copyright policy: All transcripts on this site are the copyright of Seeking Alpha. However, we view them as an important resource for bloggers and journalists, and are excited to contribute to the democratization of financial information on the Internet. (Until now investors have had to pay thousands of dollars in subscription fees for transcripts.) So our reproduction policy is as follows: You may quote up to 400 words of any transcript on the condition that you attribute the transcript to Seeking Alpha and either link to the original transcript or to www.SeekingAlpha.com. All other use is prohibited.
THE INFORMATION CONTAINED HERE IS A TEXTUAL REPRESENTATION OF THE APPLICABLE COMPANY'S CONFERENCE CALL, CONFERENCE PRESENTATION OR OTHER AUDIO PRESENTATION, AND WHILE EFFORTS ARE MADE TO PROVIDE AN ACCURATE TRANSCRIPTION, THERE MAY BE MATERIAL ERRORS, OMISSIONS, OR INACCURACIES IN THE REPORTING OF THE SUBSTANCE OF THE AUDIO PRESENTATIONS. IN NO WAY DOES SEEKING ALPHA ASSUME ANY RESPONSIBILITY FOR ANY INVESTMENT OR OTHER DECISIONS MADE BASED UPON THE INFORMATION PROVIDED ON THIS WEB SITE OR IN ANY TRANSCRIPT. USERS ARE ADVISED TO REVIEW THE APPLICABLE COMPANY'S AUDIO PRESENTATION ITSELF AND THE APPLICABLE COMPANY'S SEC FILINGS BEFORE MAKING ANY INVESTMENT OR OTHER DECISIONS.
If you have any additional questions about our online transcripts, please contact us at: firstname.lastname@example.org. Thank you!