Cray Inc. (CRAY) CEO Peter Ungaro on Q3 2018 Results - Earnings Call Transcript

|
About: Cray Inc (CRAY)
by: SA Transcripts

Cray Inc. (NASDAQ:CRAY) Q3 2018 Earnings Conference Call October 30, 2018 4:30 PM ET

Executives

Paul Hiemstra - Investor Relations

Peter Ungaro - President and Chief Executive Officer

Brian Henry - Executive Vice President and Chief Financial Officer

Analysts

Alex Kurtz - KeyBanc Capital Markets

Aaron Rakers - Wells Fargo

Chad Bennett - Craig-Hallum

Operator

Good afternoon. My name is Tanya and I will be your conference operator today. At this time, I would like to welcome everyone to the Third Quarter 2018 Financial Results Conference Call. [Operator Instructions] Thank you. Mr. Paul Hiemstra, Corporate Treasurer and IR Contact, you may begin your conference.

Paul Hiemstra

Thank you. Good afternoon. I would like to thank everyone for joining us today. Participating from Cray are Peter Ungaro, President and Chief Executive Officer and Brian Henry, Executive Vice President and Chief Financial Officer. Today’s press release is available in the Investor Relations section of our website at www.cray.com. This call is being broadcast live on the Internet and recorded for replay purposes. The telephonic replay will be available shortly after the call. You can access it by dialing 1-855-859-2056. International callers can dial 1-404-537-3406. You must enter the access code 9872838. A replay will also be available in the Investor Relations section of the Cray website for 180 days.

I would like to remind each of you that today’s conference call will contain forward-looking statements that are based on our current expectations. Forward-looking statements include statements about our financial guidance and expected future operating results, our product development, sales and delivery plans, the future growth of markets for our products, our ability to expand and penetrate our addressable market and other statements that are not historical facts. These statements are only predictions and actual results may materially vary from those projected.

Please refer to Cray’s earnings press release dated today and quarterly report on Form 10-Q for the period ended September 30, 2018 as well as Cray’s documents filed with the SEC from time-to-time concerning factors that could affect the Company and these forward-looking statements. Our presentation includes certain non-GAAP financial measures in an effort to provide additional information to investors. Non-GAAP measures, other than non-GAAP outlook, have been reconciled to their related GAAP measures in accordance with SEC rules. Our non-GAAP measures adjust for certain non-cash, unusual and infrequent items included in our GAAP results. Typical adjusting items include stock-based compensation, amortization of purchased and other intangibles and purchase accounting adjustments. When applicable, we also adjust our book tax provisions for certain items, including the impact of non-cash items such as benefits principally related to our net operating loss. You can find a reconciliation of these non-GAAP financial measures to our GAAP financial measures and a discussion of our non-GAAP outlook in our earnings press release, which is posted on our website and which is included with a related 8-K furnished to the SEC.

With that, I would like to turn the call over to Peter Ungaro.

Peter Ungaro

Thanks, Paul and thank you all for joining the call today. I will start with some comments on our third quarter performance, as well as our new product launch today, then turn it over to Brian who will go through our financial results and outlook. I will wrap up by discussing our focus areas going forward, then open the call for Q&A.

We had a solid quarter, highlighted by building momentum in several key areas of our business. We exceeded the revenue target we previously laid out and also had a very strong quarter on the bookings front signing several contracts across each of our product lines computing, storage, and AI and analytics.

Q3 clearly was another quarter that gave us solid signs of the rebound we are seeing in the high-end of the supercomputing market.

Let me give you a quick view of some of the activity we completed during the quarter. In supercomputing and storage, we installed a cluster system along with a Cray storage at a government customer, an XC Systems at a Japanese University and a couple of European customers. We also installed sizable clusters at a large energy customer and a government contractor based here in the U.S.

Overall, clusters were strong again representing a larger percentage of revenue than usual. We installed sizable clusters to our storage solutions at a government laboratory, as well as at a commercial customer in the U.S, In AI and big data analytics. We completed multiple installations around the world. We installed CS-Storm systems for used for AI at several sites including the Japanese University I just mentioned, a large automobile manufacturer, a government customer and a European weather center.

A Fortune-500 pharmaceutical company significantly increased the size of its CS-Storm and another government customer chose our Urika-CS analytics platform. Now I want to shift to our exciting announcement that we made this morning. The launch of our new supercomputing platform which I believe will be a game changer in our industry.

Code-named Shasta, this revolutionary system features an entirely new software and hardware design. It is said to be the technology that underpins the next era of supercomputing characterized by Exascale performance capability, new datacenter workloads and explosion of process or architectures.

Shasta will not only feature revolutionary performance at scale but will also be the most flexible system we’ve ever created supporting a wide range of system interconnects and processor architectures to enable customers to fit the design of their systems to closely match their applications and datasets.

Shasta’s flexible design allows it to run diverse workflows and workflows all-in-one system all at the same time from scientific modeling and simulation, to AI and analytics. Shasta will ship with our new Cray design and developed system interconnect called Slingshot, which builds on our industry leadership and scalable network performance and adds new capabilities to broaden our market reach.

Slingshot will have up to five times more bandwidth per node and is designed specifically for data-centric computing. The Slingshot network features groundbreaking congestion control to isolate applications from other network traffic, as well as Ethernet compatibility for datacenter integration. It will also allow us to pull our storage solutions right onto the supercomputing network, tightly binding compute and storage resources in the system.

Shasta's flexibility also enables support for other system interconnects, such as InfiniBand and Omni-Path. With Shasta, we are eliminating the distinction between clusters and supercomputers with a single new system architecture providing a rich software environment and system interconnect in different infrastructure form factors.

Shasta will come in two options, a 19-inch, air and liquid-cooled standard datacenter rack and a high-density liquid-cooled cabinet. As processor wattage increases over time to boost computational performance, Shasta is future proof and eliminates the need to do forklift upgrades, driving down TCO and providing customers investment protection.

It will support a wide variety of processors and accelerators including those from AMD, Intel, Maribel, NVIDIA as well as others. As is also announced by the U.S. Department of Energy this morning, the National Energy Research Scientific Computing Center in Berkeley California known as NERSC, selected a Shasta system for its NERSC-9 program.

With a total contract value of $146 million, this is the #rd largest award we've ever received in our history. We will be delivering a nearly 100 petaflap pre-Exascale Shasta system, as well as a next-generation Cray storage, all in the 2020 timeframe with acceptance likely to be completed in 2021. This system will be more than triple the computational power currently available at NERSC.

We are extremely proud to have been selected for this award which represents a significant endorsement of our technology roadmap. I couldn’t be more excited about unveiling Shasta today and what it can mean for our future. We plan on starting shipments in the fourth quarter of 2019.

With that, I will turn it over to Brian to take you through the numbers.

Brian Henry

Thank you, Pete. Before I get to our outlook, let me first take you through the third quarter financial results. For the quarter, revenue was $93 million and as anticipated, we reported a net loss. Product revenue was $58 million and service revenue was $35 million.

Please note that the majority of my remaining comments will be focused on our non-GAAP measurements, as we feel that it is the best way to look at our progress. We also encourage investors to focus on our results over several quarters as the variability in any given quarter is typically very large given the nature of our business.

Total non-GAAP gross profit for the quarter was about 27%, with product margin coming in at 16% and service margin at 46%. Product margin was lower than we typically target due in part to mix of clusters, a higher mix of clusters as Pete mentioned including one which carries low margin. These systems are typically more commodity based and as a result tend to carry a lower margin than our higher-end supercomputers or storage.

Non-GAAP operating expenses for the quarter were $45 million slightly higher than the third quarter of 2017. Our GAAP net loss was $22 million for the quarter compared to $10 million in the prior year period with the difference driven in large part by a sizable income tax benefit in the third quarter of 2017. As we have discussed previously, we are not currently reporting the potential future tax benefits from our loss.

On a non-GAAP basis, our net loss for the quarter was $19 million or $0.47 per share compared to $13 million or $0.33 per share in the third quarter of 2017. Our third quarter GAAP operating results include $3.6 million for depreciation, $3.5 million for stock-based compensation, and $300,000 for amortization.

Shifting to the balance sheet, cash, investments, restricted cash at the end of the third quarter totaled $184 million. As you likely know, our cash balances tend to be volatile due to multiple factors as we encourage investors to focus on working capital, which is less volatile. Working capital decreased by $31 million to $294 million at the end of the quarter, compared to the last quarter. About half of this decrease was due to an increase in long-term restricted cash associated with our working capital revolving line.

Inventory decreased by $6 million in the quarter to $142 million with 32% or $46 million out at customer sites and in the acceptance process.

I would now like to take a moment to discuss our outlook. For 2018, while a wide range of results remains possible, we continue to expect 2018 revenue to be in the range of $450 million. Overall, non-GAAP gross margin for 2018 is expected to be in the range of 30%. This is slightly lower than our previous outlook driven by several items including the product mix issue I mentioned earlier as well as higher overhead rates.

We expect non-GAAP operating expenses for the year to be in the range of $190 million. Adjusting items for GAAP to derive non-GAAP results are predominantly driven by stock-based compensation and are expected to total about $14 million for 2018 with about $2 million of that going to cost of sales and $12 million to operating expenses.

Based on this outlook, even though we are expecting a loss, we are anticipating nominal tax expense for the year on a GAAP and a non-GAAP basis. For earnings per share purposes, share count should be somewhat lower than 41 million, though it is dependent on a number of factors.

For 2019, while a wide range of results remains possible and it is still early in our planning process, we continue to expect revenue to grow modestly, compared to our 2018 outlook.

In conclusion, we are making progress toward our goals for the year and I remain very confident in our ability to drive long-term revenue growth and shareholder value.

With that, I will turn it over to Pete.

Peter Ungaro

Thanks, Brian. We have two key focus areas for the rest the year. The first is to win new business for 2018 and beyond and then second is to continue to expand into our key growth areas of commercial customers and big data solutions. We maintained our momentum on the first one with strong bookings during the third quarter driven in large part by the award from NERSC that I mentioned earlier. Total backlog increased by 32% in the third quarter to $615 million.

We now have all the contracts we need in order to achieve our outlook for 2018. While we have work left to do complete the remaining acceptances, several of which are likely to come down late in the year, we are on track to deliver 15% revenue growth compared to 2017. And while we are certainly not back to revenue levels where we were a couple of years ago, and where we need to be as a company, I am pleased with our overall performance so far in 2018 and with what we plan to complete over the remainder of the year.

As we mentioned over the past couple of quarters, a significant portion of the opportunities in the market are for 2020 and beyond. As a result, while we expect our business momentum to continue to improve, especially from a bookings perspective, we only expect revenue to grow modestly in 2019 as Brian mentioned.

The high end of the supercomputing market continues to improve and customer activity remains extremely high. Our competitiveness remains strong as do our win rates.

As we look ahead, our Shasta system will be key to building on this competitiveness. While we have been in this market downturn, we are continuing to aggressively invest in our technology roadmap and have taken a long-term approach to maximizing the significant market opportunity, we see coming in the next few years.

Ultimately, Shasta will expand on each of the key advantages of our current offerings to deliver to what we believe will be the most performing, scalable and flexible system on the market.

Our second focus area for the year is to continue to expand into commercial customers and Big Data markets. The commercial market continues to improve driven in large part by a rebound in the energy segment. Some of our largest energy customers both in the U.S. and around the world are beginning to reinvest in supercomputing and storage solutions. Manufacturing in Aerospace also continues to be a solid contributor to our business.

Overall, we expect revenue from commercial customers to increase substantially in 2018 and to represent more than 15% of our total revenue for the year. While it remains fairly concentrated with a handful of large customers, as well as typically lumpy, I am very pleased to see this market continue to improve and we are confident in our ability to delivery value-added solutions which drive increased competitiveness for our customers.

In big data, we are focused on three high growth areas: analytics, high-performance storage and artificial intelligence and deep learning. Each of our solutions here are designed to enable our customers to seize on the ongoing explosion of data and the desire to harness new ways to analyze it and gain better insights.

Our ClusterStor solution continues to lead the market at the high-end delivering the most powerful, robust and cost-effective storage solution on the market. On the analytics and AI front, our new Urika-CS and Urika-XC software allows our customers to run their analytics and AI workloads right on their Cray supercomputers versus having to purchase a separate system.

As we look to the future, Shasta will play a substantial role in enabling us to expand further into commercial customers in the big data market. Enterprises can seamlessly incorporate a Shasta system into their infrastructure as Shasta is designed to fit into enterprise datacenters with its standard rack packaging, Ethernet compatible interconnect and modern system software architecture.

With Shasta’s flexible design supporting a wide range of processors suited for different algorithms and applications, enterprises can start as small as they want and expand their system over time to support their workloads.

Before I wrap up, as with last quarter, I understand that many investors are focused on the Department of Energy’s CORAL-2 program. This program is for Exascale supercomputers in the 2021 to 2023 timeframe covering two systems with an option for a potential third and an expected budget of over $1 billion.

As you may know, we submitted proposals for this procurement in late May of this year. At this point, we are still not able to comment any further regarding this program.

In conclusion, I am pleased with our performance in the third quarter and incredibly excited about the launch of our next-generation supercomputing platform and huge win at NERSC. While we have work left to do, we are in good position to drive growth for 2018 and beyond.

With that, I’d now like to turn the call over to the operator to begin the Q&A.

Question-and-Answer Session

Operator

[Operator Instructions] Your first question comes from the line of Alex Kurtz from KeyBanc Capital Markets. Alex, your line is now open.

Peter Ungaro

Hi, Alex.

Alex Kurtz

Hey guys. Hey, Pete. Can you hear me okay?

Peter Ungaro

Yes.

Brian Henry

Yes, good afternoon

Alex Kurtz

Good afternoon, Brian. I just want to clarify, congratulations first on Slingshot. We’ll have some questions on that. But I just want to clarify your last comments about CORAL-2. I think it will be helpful for investors in the market to understand that when those projects are announced, is it safe to assume Secretary Perry and the DOE, will be the ones headlining that announcement and not something on an earnings call from Cray or IBM or whoever?

Peter Ungaro

Well, that’s impossible to predict, but I would expect that, you saw today that Secretary Perry and the Department of Energy announced the NERSC system this morning. And so, I think that that’s probably likely, but not in our control to predict.

Alex Kurtz

I guess, it will be unusual for you to comment about Exascale and CORAL-2 on an earnings call about having a similar path of announcement from leadership at DOE correct? Is that a safe assumption?

Peter Ungaro

Yes, most likely.

Alex Kurtz

Okay, just a quick question about Shasta and Slingshot, how do you see Slingshot helping you in the pre-Exascale and Exascale competitive bidding environment? And then, after that question, just my last question is how are you going to transition between XC and areas in Shasta as you go through 2019? Thanks.

Peter Ungaro

Yes, great questions, Alex. So, we think, the Shasta architecture and Slingshot as a system interconnect, position us very well for – as we see as the next era of supercomputing which really includes Exascale-based systems. The win at NERSC I think is a great kind of good housekeeping stamp of approval on the architecture and gives us a great customer to launch with.

And as we’ve said in the announcement this morning on both the NERSC system as well as our Shasta announcements, Slingshot in particular gives us a lot of really important features that jut can adjust and control and adapt a routing that really are important as you start to scale up to these very large systems.

So, we think we are very well positioned and we think we have great technology roadmaps for customers whether they need an Exascale machine or a single cabinet to run a AI workload.

Your other question was, the transition between our current platforms and Shasta. And as we mentioned, we are going to start shipping Shasta at the end of 2019. So, 2019 is really going to be a transition year for us overall. Our plan for all of our customers is to be able to take applications that would run today on the XC platform and be able to seamlessly move those over into Shasta.

So we really expect that it should be a pretty smooth transition. Of course, I think that we are going to get a bit of a situation where some customers will want to wait for that next-generation of system and so there could be a little bit of a lag there in 2019, but we are going to monitor that pretty closely and we’ll let you guys know if we see any difference in our views on our 2019 outlook.

Alex Kurtz

All right. Thanks guys.

Peter Ungaro

Yes. Thanks, Alex.

Operator

Thank you. Your next question comes from the line of Aaron Rakers from Wells Fargo. Aaron, your line is now open.

Peter Ungaro

Hi Aaron.

Brian Henry

Rakers.

Peter Ungaro

Hey, Aaron, are you there?

Aaron Rakers

Yes, hey guys can you hear me?

Peter Ungaro

Yes.

Brian Henry

Yes.

Aaron Rakers

Okay, sorry about that. So, a couple questions if I can as well, kind of building on what Alex just asked about, but maybe a little bit differently, it seems to me that the Shasta architecture, one of the attributes that just kind of blurring the lines between the clustered platforms and out of the true high-end scaled out platform in a supercomputing space.

So, similar to the lines of Alex, how do you manage that line between your CS products and that of the new Shasta platforms? Do they effectively blur and as we see Shasta moving forward, does it effectively open up opportunities to deploy those into more commercial opportunities?

Peter Ungaro

Yes, I think that’s a great question, Aaron, because it’s one of the things that we thought a lot about is we were kind of the initial design of Shasta. Today, customers have to make a choice. If they want a flexibility of a more scale out cluster like our CS line, they have to go into that platform.

They get that flexibility, but they don’t get a lot of the other things like they can't get our interconnect, they can't get our end-to-end software stack. They are kind of locked into that space or they decide that they want to be on our higher end XC line and they get that proprietary system interconnect, they get this software stack that’s enabled on that system.

But they don’t have a lot of the flexibility that they have in the cluster line. And so, we really were pushing our customers into one or the other bucket. As are – as every company out there doing today. So it’s no different for any other company. With Shasta, we just completely take that off the table.

All customers can get our end-to-end software stack from top to bottom. All customers can get the Slingshot interconnect or Omni-Path or InfiniBand if they want that and all customers can fix which kind of cabinet infrastructure do they want to be in or more traditional densely packed cabinet that’s liquid-cooled and has warm water cooling and things like that or a more standard 19-inch commodity stalled cabinet that’s more typical in commercial datacenters today for instance.

So, we really blur all of that and in my mind, with Shasta we won’t have two product lines. We will have one product line and we will just get to pick which kind of cabinet, cabinet will just be a choice. Software will just be choice. The interconnect would just be a choice. So it’s really going to bring these two lines together.

Aaron Rakers

And then, kind of a secondary question, just to think about the business momentum that you had, you mentioned that you had $615 million of backlog exiting the quarter. That was up 32% year-over-year, I think it was up from $465 million in the prior quarter. The simple math would suggest that that really just adds in the new DOE project. I am just – I am curious of how much bookings growth has been or the backlog build has been outside of that new DOE project for Shasta?

Peter Ungaro

Well, as you know, backlog is going to come down every quarter with revenue. So, we started the quarter at one level, that was going to come down by the revenue that we got $93 million in last quarter and then we got to build that back up. So we got to make that back up and then go above that.

And so, it’s certainly a huge part of our backlog growth was with NERSC, no doubt about that. But we also had a number of other wins. And it was good overall quarter for us as far as wins go and NERSC was kind of icing on the cake.

Aaron Rakers

Okay. And then, the final thing with Shasta kicking in, in this year, should we expect – thinking about some of the Exascale or some of the other pre-Exascale platform opportunities that are out there, should we expect 2019 would be a very active year on announcements with Shasta as it relates to those opportunities? Is that really what we are looking at, as this kind of transition your plays go up?

Peter Ungaro

Yes, I think, as I mentioned, I think 2019 from a revenue growth perspective is going to be modest, but I think from a bookings perspective as I mentioned, we are expecting big things as we build back up to the revenue levels that we were at before the market took a dip in and hopefully beyond that, we definitely are very focused on building up our bookings and our backlog going forward.

And I think Shasta gives us a great opportunity to do that. So, absolutely, I think next year is going to be a strong year from that perspective.

Aaron Rakers

Great. Thank you.

Peter Ungaro

Yes, thanks, Aaron.

Operator

Thank you. [Operator Instructions] Your next question comes from the line of Chad Bennett from Craig-Hallum. Chad, your line is now open.

Peter Ungaro

Hi, Chad

Brian Henry

Hey Chad.

Chad Bennett

Hey, Pete. Hey guys, thanks for taking my questions. So, going back to the interconnect question or Slingshot question, take us through the thought process of getting back into I guess, the interconnect market. Obviously, we sold off our hardware assets a few years back to Intel and obviously retained the software capabilities there because we thought the hardware part of the market was going to be somewhat commoditized.

So, help me understand the process of getting back in. Help me understand the gap you fill so to speak relative to the other two players in this space, especially when we move to Exascale and higher scalability and parallel systems and maybe lastly, do we believe the interconnect is going to be more valuable this cycle than last?

Peter Ungaro

Yes, those are great questions, Chad. You packed a lot into that one. So, I would tell you that we have great partnerships of course with Intel on the Omni-Path Interconnect and we talked about that on past calls a number of times and we’ve built a number of very large systems with Omni-Path as we have with Mellanox with InfiniBand.

But we also view that there is a gap in the market for a very unique kind of high-end network that is able to really be driven at this new data centric kind of workload design that people are moving towards. And so, as we are thinking about our own interconnect roadmap going forward, clearly we want to add value into what’s happening with Omni-Path and Intel and Mellanox with our software stack.

But we also saw an opportunity to do some things as servers started to transition from having PCI Gen-3 capability to PCI Gen-4 capability which you are going to start to see coming out next year and the following year, we saw an opportunity there to kind of step in with a new interconnect overall. We also – we are seeing that one of the things at very large scale and we were doing a lot of research around this was how to deal with all of this congestion control situation.

So, you know, typically when people think about networks, most people talk about bandwidth and latencies. Our CTO, if he was here, Steve Scott would say, it’s kind of like looking at your morning commute at 3 in the morning.

It kind of doesn’t matter when all the roads are clear, what matters is what’s your morning commutes in a rush hour time and that’s the same that was happening in supercomputers where as soon as you have a lot of traffic in everything on that system, applications were slowing down and we were doing a lot of research to see is there real good ways to get around this both for single applications, most things called tailed latencies, which is kind of those toughest kind of last bits of a message that passes through the network.

And so, we really came up with some innovative concepts and ideas that we wanted to implement in a system and so, we decided to develop Slingshot along with our partnerships with the other - around the other interconnects. Lastly, I would just say that the Ethernet compatibility was also a big part of it. So, a lot of our commercial customers have internet, Ethernet-based datacenters, a lot of what’s happening in the AI and analytics space is leveraging kind of Ethernet-based systems and even a lot of our scientific customers wanted to connect the supercomputer interconnect out to their other devices within their datacenters including scientific instruments and also of course into storage because, getting storage closer and closer to the interconnect is a big part of that. So, we just felt like we had a lot of innovative ideas that we thought were unique in the marketplace that we can build around and so, we are pretty excited with Shasta that we are able to support not only our own interconnect with Slingshot, but also for those customers that want to support InfiniBand or Omni-Path, we can support those interconnects too. So, our goal is to have the best system whether it be a Slingshot network, an Omni-Path network or an InfiniBand network and I think we are going to do that with Shasta.

Chad Bennett

Will Slingshot include your own custom ASIC or using something else?

Peter Ungaro

Yes, no, it will definitely be rather the custom Cray agent.

Chad Bennett

Okay, got it. And maybe one last one for me. I imagine from a R&D spend standpoint, we are incurring and have incurred any incremental R&D related to the decision to produce our own interconnects. I guess, is that a right assumption and should we expect any material change to the R&D runrate heading into next year?

Peter Ungaro

Yes, that’s a fair assumption. We definitely have been investing in both the Slingshot interconnect as well as Shasta over the past few years and that’s going to continue. We have a roadmap for both of those products obviously as we go forward.

We’ve really made a strategic decision because we see a very large opportunity opening up in our market as our traditional scientific model, since modeling in simulation starts to integrate with this world of AI and analytics, and we see a much bigger opportunity out there and then when you add in the Exascale opportunities for instance, that are going to happen all over the world.

We really see a great opportunity for us to significantly grow over time in the market and key for us is the technology innovation and that’s what we have been investing in and that’s what we will continue to invest and as long as we see that, that opportunity out there and we are pretty convinced that it’s out there for us. So, I think that you could assume that we are going to continue on this path as we go forward.

Chad Bennett

Good to hear. Exciting stuff. Thanks for taking my questions.

Peter Ungaro

Thanks, Chad. Yes, appreciated.

Operator

Thank you. [Operator Instructions] Your next question comes from the line of Aaron Rakers from Wells Fargo. Aaron, your line is now open.

Peter Ungaro

Hey, Aaron

Aaron Rakers

Hey, thanks. I want to just go to the model real quickly. Obviously, this last quarter you had, it sounded like the mix effect impacts your product gross margin. But I am just curious as we think about this product transition and maybe beyond that, how we should think about that product gross margin line?

Brian Henry

Well, I think what we are doing, as Pete described, both of the systems, a 19-inch rack or a more optimized rack for density are going to be similar and have a lot of Cray value added if that’s what the customer wants. And so, I think that we have a good opportunity to expand our margins over time. On that fact, and you know, the Slingshot is one thing, the software stack is another thing that what we can do in packaging, yes, another. So, we are pretty excited about where we are headed in the competitiveness of our platform.

Peter Ungaro

Yes, I would definitely say that, as we get into the Slingshot, Aaron, so, maybe out in the 2020 timeframe when we are shipping that for a full year we hope to see – hopefully little upside on the margin side. As you are also asking, Aaron on the transition time, I think 2019 is going to be a year of transition. We are seeing a bit of a shift this year.

A lot more clusters which have lower margins than a traditional XC platform or storage. And I think you’ll see that continue a little bit into 2019, because a lot of customers are really needing this flexibility, because their workloads are becoming very heterogeneous and a mixture of kind of traditional modeling and simulation in AI and Analytics.

So, we are seeing a bit of a shift of that in our product portfolio. So, I think, in 2019, we will still see a higher percentage of our revenue out of clusters than we do in the high-end machines kind of similar to this year overall. Would you add anything to that Brian?

Brian Henry

No, I agree with that. As you go on to 2020, I think, it accelerates as we move through the year.

Peter Ungaro

Yes.

Aaron Rakers

Okay. Perfect. And then, just real quickly as a follow-up not really a follow-up, but another question. It’s been a little while since we’ve talked about Cray’s opportunities possibly in cloud and the relationship that you had or you have with Microsoft. I am just curious is there any update there on how possibly Cray is positioned in some of these cloud opportunities and whether or not that over the next twelve months, we could look at that as an adjacent or incremental revenue stream at all?

Peter Ungaro

Yes. I think we are really hopeful, Aaron that does going to start to take off. As we’ve mentioned before, we have been working very closely with the Azure team at Microsoft in getting a Cray supercomputer put up and running in the Azure framework.

We don’t have anything to announce on that today. But that work has been ongoing and we really see cloud as another opportunity or another channel for us really into the market, because a number of our customers that don’t really have the datacenters to house a supercomputer are very interested in being able to access one and use one and this provides them another way to do that.

Also, we have a set of customers that are already up, for instance, in the Azure cloud and have their data up in the Azure cloud and would love to have the ability to have a supercomputer up there to run some of their models or AI models.

So, I think that it’s definitely an opportunity that we are excited about. I think it’s going to take a little while to really kind of formulate and pick up in the markets. But it is something we are – still very positive about and very hopeful for in our partnership with the Azure team at Microsoft been going great.

Aaron Rakers

Okay. Great. Thank you.

Peter Ungaro

Yes. Thanks.

Operator

Thank you. Your next question comes from the line of Alex Kurtz from KeyBanc Capital Markets. Alex, your line is now open.

Alex Kurtz

Thanks. I want to follow-up on the question from Aaron Rakers, real quick. So, Brian, is it – I am sure you’ve done a lot of analysis on the BOM cost for Slingshot. Is there a chance that once you get this things and volume, there is a shot on goal of getting to 40% product margin for the compute business, not blended product margins, but just for Shasta? Is that sort of stretch goal may be?

Brian Henry

Yes, I would put it in a stretch goal, only because, we utilize a lot of the other party’s technologies. Other processors, the accelerators, the memory and things like that. And we are in a marketplace that we are very specificated buyers and they are kind of keep an eye on what’s out there and there is some limit as to how significant our differentiation can go. Now, do I think it’s possible, yes, but I don’t think it’s how we are viewing it today as the – the part of it is, it’s a very big system or often quite congested.

Alex Kurtz

Now I understand, but apples-to-apples, is Slingshot or Shasta are going to be a higher margin outcome than Aries and XC?

Brian Henry

We’d be very disappointed if it isn’t better and again, we get to have our value added on kind of two choices and cover a broader market with our value-add and then we can in general today.

Alex Kurtz

And Pete, just to be clear, because I'm not sure if I asked this the right way on the additional questions on – the initial round of questions. Shasta Slingshot is going to be your go to market technologies for CORAL-2, correct? I mean, that’s what’s in the bid?

Brian Henry

Yes, we bid Shasta technology for CORAL-2.So, if you look at our CORAL-2 offer, it contains an evolution of the Shasta technology that we have announced here today. So we are very excited about it.

Alex Kurtz

Shasta with Slingshot in the bid?

Brian Henry

That’s as far as I’m going to comment, Alex.

Alex Kurtz

Okay, fair enough. Appreciated guys. Thank you.

Operator

[Operator Instructions] Your next question comes from the line of Ben Meir, individual investor. Ben, your line is now open.

Unidentified Analyst

Hi. Peter and everyone at Cray. Thank you very much for your time and taking my call. I wanted to say that the Shasta announcement is very exciting and its flexibility. The new network sounds great. Wanted to thank you guys for your good work.

Peter Ungaro

Thank you.

Brian Henry

Thank you.

Unidentified Analyst

I had two questions for you guys. The he ARM processors are new to the HPC arena and the HPC community has a fairly large software stack as required for functional operation. Other ecosystems have their own requirements. At the level that’s appropriate for this call, could you talk about preparations that you guys are making for enabling this the ARM architecture ecosystem and maybe any other work for GPUs that you guys have done?

Peter Ungaro

Yes, Ben, that’s a great question. So, in fact, we have today support for ARM processors right in our XC systems, as well as of course our clusters. So, we have taken our entire end-to-end software stacks from the operating and system environment, the system management environment, all of our tools, our compilers, libraries, all of that and brought that over to ARM.

So, today, for instance, in our XC product line, if you buy an XC with the Marvell ThunderX2 processor in it, you have the entire Cray software stack that goes along with that. So, we’ve really brought our entire stack over to that environment. Now, clearly with ARM, there is a lot of other applications in the ISDs out there that need to port their stuff to ARM.

So, from a system infrastructure side, we have everything that’s needed. We are of course working with the opensource ARM community and incorporating all of those tools and offerings there as well as with ARM directly in doing that. But there is work to be done as more and more applications, end applications get supported.

And we think that’s going to come over time. So, for people that have their own applications, they offer many opensource applications out there today, they run already on ARM and there is some great examples of those out there. But others will come over time. There is not a huge ARM ecosystem in the server compute business like there is for instance for X86, of course and that’s going to take some time.

GPUs you mentioned, I would answer it the same way of course. So GPUs that are in Cray supercomputers, we’ve brought our entire stack over to GPUs and we could take advantage of all of those. They are a little bit further along of course. The GPU computing has been going on for quite a while now and if you look at specific areas like steep learning for instance, I would say that there by far have the majority of applications out there running on GPUs. So, it’s kind of different by each of the processors and such.

But overall, our strategy has been to make sure that for those processors that we support, we have a software environment for our customers that they can count out end-to-end. So, that’s one of the big advantages I think that we have right now is our software stacks, not just while our hardware is great, but our software stack is really differentiating today in the market.

Unidentified Analyst

Okay, that’s really good. It sounds like Cray is ready to go and just waiting through everyone else to catch-up. So, I – yes, okay, good. So, I understand there is a large amount of variability in your business income. It’s nice to see that bookings have grown over the last quarter.

I was looking at GAAP revenue. I know you guys like the non-GAAP. In GAAP it looks like the loss per share has doubled. Can you just talk about plans to reboost those trends? Or is there any plans to leverage your expertise and HPC or maybe starting with cloud to start selling high margin software, whether it’s developed in-house or sort of a value-add thing?

Peter Ungaro

Yes, so the way that we really think about it is, that our – the last two years, our market has been very depressed and we talked about this a lot in previous investor calls. So, one of the things that we really believe and fundamentally that we are starting to see is, the market is returning and we think the market is going to come back and be even stronger that it was previously.

And so, from that perspective, we believe that there is a lot of growth that’s going to be out there in the market that we are going to be able to take advantage of and we don’t have to scale our infrastructure or kind of OpEx with that market opportunity. So, we feel like we are going to get a lot of leverage in the model as the market grows and as we grow, hopefully even faster than the market grows.

And then there is certain opportunities that are out there that we are talking about, Ben, that are these three Exascale systems that can fundamentally change the game for us in a big way. These are systems that are not measured in the $10 million or $20 million range, but are measured in hundreds and millions of dollars are in the case of the Department of Energy Exascale system across two machines over $1 billion.

And so, those can kind of be that icing on the cake that we like to say on top of what we think is going to be a good market. And then lastly, the whole world of artificial intelligence and big data analytics which is bringing a lot of new customers to the table for supercomputing technology and we see growth there too.

So, we are going to continue to manage our business pretty aggressively from an overall cost perspective, but at the same time, we are going to make sure that we are investing in technology to have that growth in – as Brian was talking earlier, to be able to have that margin expansion and be able to be not just profitable, but hopefully, strongly profitable over the next few years. Brian, anything you want to add to that?

Brian Henry

Just a reminder to everybody. We vary by quarter and even by year fairly significantly. It’s just kind of the nature of what our business is and so we can have a set of quarters and maybe backlog and things like that will become more important to just see our progress for a little bit because it will be a little bit before we will get full kick in, in revenue on some of the timing of things.

So, I would just say, look, quarter-to-quarter we are going to be pretty lumpy. Book-to-bill was good this quarter helped by a very large order. But that is a bit of Cray’s business that we get those from time-to-time added into kind of what we do. So, I think, 2019 from a booking point of view, will be continue to be good, 2020 should be good. So, we have a good chance there.

The revenues will trail a little bit and we are in this period, where we have revenues trailing. Now we are investing, right now and have been for the last year-and-a-half or two years really to get the opportunity that we see out there. We wish the current businesses are stronger than it was. We can’t control what happened in the market over the last few years. We are still kind of paying for.

But I can’t tell you the activity level in the business is up a lot. Opportunities, we see, look is good, if not better than ever and that’s kind of how we are looking at it. So we are making the investments and play it longball, more than we might have played, but that’s where we are and we are quite confident that it’s going to pay off for us and the shareholders.

Unidentified Analyst

Okay, so this sounds like you’ve got a plan for scaling and maybe the current loss per share is, just an aberration for this quarter. Did you, well, I’ll let you guys go and just I’ll thank you for your time. Thanks for answering my questions.

Brian Henry

All right. You are welcome. I just want to warn, we expect a loss for the year. We will have losses from time-to-time going forward particularly on a quarter basis. But we are focused for shareholder value in the long run and that’s what we are investing for and feel real good about the opportunity to achieve that. This may not happen in the short run.

Unidentified Analyst

So, actually, let me follow-up on that. So, how long from bookings to revenue realization is typical for you? I understand every project is different and it may slip by a quarter or two. But is there a six month or a year sort of progression?

Brian Henry

I mean, it does vary dramatically and we can have something that’s fairly significant that we can get in, build very quickly. We are pretty efficient at that and get it shipped and get it accepted. And then we could have somebody who is really picking a point in time to have when they want to spend the money, get the new technology or whatever and we have – it was just kind of a luxury for Cray, people, our customers want to be on the leading edge of technology often and they pick points that they think are those fit and we have a lot of potential opportunities that kind of are trying to match.

In this case, where Cray will be best for them or the technology be best for them and that’s why you can get an order now and depending on the complexity and the timeframe may not be delivered for a year-and-a-half or two years. In the case of Exascale, there is a very outline of that being a pretty long booking to delivery date in the future.

So, those things are just important to say they vary a lot. And that’s why I would say, kind of keep track of kind of how we are doing on bookings over time from the backlog perspective is probably the best way to think about.

Operator

Thank you. That ends up the question and answer session. I will hand the call back to Cray Incorporated.

Peter Ungaro

Thank you. I just like to wrap up and note that Supercomputing 18,SC18 which is the biggest supercomputing industry conference of the year starts in a couple weeks in Dallas. It’s sure to be an exhibiting show and I hope to see some of you be there. So, thank you all for joining the call today and for your continued support of Cray. Have a great evening.

Operator

Thank you and that concludes today’s presentation. You may now disconnect.