There is a war under way today between some of the largest technology companies in the world. There are allies and adversaries, friends and enemies, and even some "frenemies" that could play both sides. There will be winners in devices and a winner as the semiconductor components supplier to the device makers.
Let's take a look at the players:
Apple has been dragged off to the woodshed after last week's earnings and guidance were soft.
Everyone knows the story here, Jon Rubenstein developed the iPod. The iPod begat the iPhone, the iPhone begat the iPad, the iPad begat the iPad mini. The iPhone is "the Franchise" at Apple. It generates 60% of the gross profit and maybe as high as 80% of the net profit. Apple ships about 125 million iPhones per year. Apple is at war with Samsung. The two companies are currently pursuing tens of lawsuits in many different jurisdictions.
Apple is also at war with Google (GOOG). According to the departed Steve Jobs, this war is of the thermo nuclear type. This war is over the Android mobile operating system that Mr. Jobs feels was a rip-off of iOS by Google, presumably through Eric Schmidt who was serving on the Apple board of directors at the time.
Apple has removed Google Maps and YouTube from the "i" products, thus impacting Google's click revenue from these assets.
In response to the Apple "Thermonuclear war", Google has gone directly into the mobile device business through two channels; the Nexus operation and the Motorola Mobility acquisition. Nexus has three devices that compete directly with Apple; the Nexus 10 tablet, the Nexus7 tablet, and the Nexus 4 smart phone. These products are credible competition to the iPad, iPad mini and the iPhone. Apparently the counter attack by Google is to sell these products at or near burdened costs, thus applying pressure to Apple's high margins. Google is applying the most pressure on those products that are the most important and profitable to Apple, i.e. the Nexus 4 16GB at $349 unlocked vs. the iPhone 16GB at $650, and the Nexus 7 16GB at $199 vs. $399 for the iPad mini 16GB. Google can afford to sell mobile devices at cost in order to trump any further intrusion by Apple or anyone else into the Google click based business model.
Google's Android operating system based smartphones, available from a multitude of suppliers, are outselling the iPhone by a 3:1 ratio. Obviously all of these Android devices come loaded with Google Maps and Google owned YouTube.
Google has entered into a cooperative agreement with Intel to optimize the Android operating system to work better with Intel x86 based application processors. Intel has also signed a multi-year cooperative agreement with Google owned Motorola in which Motorola will use Intel Application Processors in return for technical support from Intel. Apple might be a little upset with Intel over this since Apple is a large customer of Intel and Paul Otellini, the CEO of Intel, is on the Google board of directors. The dark side of me thinks that Apple has insisted on Paul Otellini leaving Intel as a condition to Apple moving their Application Processor fabrication business from the hated Samsung to the only slightly more desirable Intel.
The Samsung components divisions have paid a huge price because of the Apple declared war (this one is of the conventional type) against Samsung. Samsung used to supply the screen, battery, Nand memory, DRAM memory and Application Processor for the "i" products. Samsung, at one time, supplied 60-70% of the dollar value of the components in an iPhone. They are now down to supplying just the Application Processor at about 12% of the value of the iPhone supply chain. Apple would like to replace Samsung as the supplier of the Application Processor, but the only rational replacement supplier is Intel and Apple apparently doesn't like the Intel/Google lash-up enough that they are willing to cut off their nose to spite their face. Intel also is not very excited about manufacturing an ARM based Application Processor for Apple at a time when the company is trying to get x86 based products in all mobile devices.
The Samsung smartphone division uses the Android OS throughout their product line and is outselling Apple by a 2:1 ratio. At this point Samsung is the largest supplier of Application Processors at 450 million units, including their own captive business and the Apple foundry business. Apple spends about $5 billion on foundry work with Samsung.
To pour a little salt in the wound, Samsung just announced a smartphone based on the Tizen operating system which was jointly developed with Intel. Oh, that Tizen phone will use an Intel Application Processor, probably because Tizen will work really well with Intel devices because of the Intel involvement in the Tizen OS design.
Lenovo group is a Hong Kong company best known as the buyer of the IBM PC business. Lenovo has just passed HP as the largest maker of PCs in the world. The company also offer tablets computers and is in a hyper-growth phase in smartphones, especially in China. Lenovo was the first company to adopt an Intel based SoC in a Smartphone design. Lenovo is in the envious position of not being a participant in this war, but is quietly pursuing their business interests.
Intel is the largest semiconductor manufacturer in the world by dollar volume. The company is the leader in semiconductor fabrication technology. Semiconductor technology is driven by "feature size", often referred to as process "node". Intel is producing over half of their output on a 22nm Trigate process node with very good yield. The next best in class is producing at 28/32nm planar node and struggling with yield. The difference in chip size due these different processes is much more dramatic that the 32 to 22nm numbers would suggest. Because of the arithmetic square relationship, assuming the same number of transistors the 22nm chip has the potential to be less than half the size (area) of the 32nm chip. Therefore, the 22nm wafer, with virtually the same production cost, can produce over 100% more chips. The manufacturing cost of the 22nm chip will be about 50% less than the equal 32nm chip. Intel is moving to 14nm node beginning this year. Those chips will be one fourth the size of the 32nm chip and one fourth the cost. These much smaller chips are also faster and consume less power.
The common opinion is that Intel has missed the mobile business, or is struggling to get involved. The truth is that while the smart phone business was exploding Intel was preoccupied with gaining a near monopoly lock on the data center server business which is now a $10-15 billion business for Intel with 80%+ gross margins.
Intel also, unfortunately, partnered with Nokia (NOK) and the MeeGo operating system which was a near total failure, thus wasting two years of Intel's effort in the mobile area. Now, Intel is re- focused on the mobile business with a beginning wave of products that are competitive with the best current products in the market.
Since Samsung is supplying 60% of the Application Processors, Intel, at the moment, is shut out of that business. The remaining 40% of the business is entrenched with suppliers like Qualcomm and Nvidia and ARM based. While those who understand the semiconductor business can see the benefits to both companies of Apple moving their foundry business to Intel, there are impediments to consider. The first is that the Apple "A" chips are designed by Apple with a core processor design from ARM Holdings. This processor, from Intel's viewpoint, is a competitive product and as such is not something that they would normally want to manufacture. Intel feels that their x86 based Atom core is a better solution for mobile applications and view making ARM based chips for Apple as undermining that position. The second is the deep cooperation with Google/Motorola with the purpose of enhancing the Android operating system to work with Atom processors, in order for Google to build products that are directly competitive with the Apple smartphones and tablets.
Microsoft either doesn't have a mobile strategy or the strategy is so sophisticated that mere earthlings are unable to understand it. Windows 8 seems to be working, Windows RT is all but incompatible with the full Windows 8, and Windows phone is just getting started with Nokia.
Microsoft has a huge built-in advantage in the form of the existing base of PCs. The smartest thing that Microsoft could do is to offer Window 8 for free for any tablet that also has an associated PC. They could carry that through to Windows phone for free for any smartphone that, likewise, is associated with a WinTel PC owner.
The power of a Windows PC, Windows tablet, and Windows smartphone, all using x86 architecture perfectly matches the Apple OSX/iOS line-up, but with better uniformity. If Microsoft made Windows free to any tablet or smartphone backed up by a registered Windows PC owner, they could restore the WinTel model that has served both companies very well for decades. I'm not sure that Microsoft management sees the problem or the opportunity.
ARM Holdings (ARMH)
ARM Holding is an intellectual property company that has designed microprocessors since the 1980s that are able to be licensed and used to build the Application Processors used in virtually all of the smartphones made today and many of the early tablet style computers. These ARM microprocessors actually live as files in CAD (Computer Aided Design) systems until they are committed to a silicon Application Processor design by an ARMH licensee such as Qualcomm or Nvidia. These Application Processor designs are, in turn, manufactured (fabbed) by silicon foundry companies such as TSMC or Global Foundries.
These ARM "cores", as they are called, have, in general, been light on computational power, but very low on power consumption. Since mobile devices are battery powered, low power consumption is extremely important. ARMH has worked very hard to make their microprocessor cores good enough to power the new smartphones and tablets. The term "good enough" is actually used in ARMH promotional presentations.
The immediate challenge for Intel to beat ARM processors in the mobile business is for Intel to build Application Processor SoCs (System on a Chip) that are much higher performance and much lower power that anything the ARM licensees and their foundries can build, thus making the Intel products overwhelmingly compelling to the smartphone and tablet makers. Intel has a huge and durable advantage in manufacturing. Intel manufactures at least one node advanced from the best foundry available to build the ARM based silicon. That one node Intel manufacturing advantage confers significant power, performance, and cost advantages to Intel.
With all the competitive noise surrounding ARMH vs. Intel, it is interesting that the cores in the new SoCs, whether Arm based or Atom based make up a very small part of the overall die size. Most of the chip size is made up of Graphic Processor Units (GPU), Video processors, Image processor, GPS, USB, Bluetooth, WiFi, memory controller, and other functions necessary to make a smartphone work. So, the ARM vs. Atom debate is mostly an emotional one.
Both companies are fabless semiconductor companies that design Application Processor SoCs based on the ARM core. Both companies use TSMC (TSM) for fabbing their products. Of the two companies, Qualcomm is the distinct leader since Qualcomm has the lead in the LTE baseband technology that encodes and decodes the radio transceivers required by all smartphones. Since both companies use ARM cores and TSMC for a foundry, they are at a potential disadvantage to the vertically integrated Intel. Both companies pay royalties to ARMH and, of course, the 45% gross margin charged by TSMC for manufacturing. Perhaps a bigger problem for QCOM and NVDA is that there is no way that they can duplicate the Intel/Google relationship.
TSMC is very large contract silicon foundry based in Taiwan. The TSMC business model is interesting in that it is a "time-shared" fabrication factory. Morris Chang founded TSMC to service companies who have good product ideas and the ability to design integrated circuits, but have no production facilities. TSMC has enabled hundreds of semiconductor companies, some of which moved on to become industry leading companies.
As a foundry, TSMC has the responsibility to maintain leading edge manufacturing technologies as well as a multitude of older technologies suitable for less advanced customers. The latest leading edge technology at TSMC is 28nm. TSMC has had a great deal of trouble getting the 28nm (and the preceding 40nm) processes into volume production with high yield. TSMC is about 1 1/2 nodes behind Intel. As such, their customers, such as QCOM and NVDA will also be 1 1/2 nodes behind with their products, thus putting TSMC customers at a large disadvantage to Intel.
TSMC is trying desperately to catch up with Intel, but listening to their earnings conference call today, it seems that a TSMC 20nm process will not be in volume production until 2015. Since the TSMC 20nm process is a planar process vs. Intel's Trigate process, there is a high possibility that the TSMC 20nm process will have worse performance on speed and power than their 28nm process, and that is if it works at all. There was no mention about the expected TSMC 16nm finfet process, but if 20nm takes until 2015, 16nm finfet will be 2017, if ever. By contrast, Intel will be ramping 14 nm Trigate later in 2013. And 10nm will begin early production in 2014-15, putting Intel 3-4 years ahead of TSMC.
The danger for TSMC being this far behind Intel is that if Intel secures enough leading edge business, if Intel's products are preferred over QCOM and NVDA, over time the market for foundry work at TSMC may diminish to the point that the cost to build TSMC advanced technology foundries might not be justified by the available business. This means that, while TSMC will not fail, they will be stuck at the 28nm node while Intel moves on to 15, 10, and 7 nm. That would not bode well for TSMC customers such as Qualcomm and Nvidia. An IC at 7nm would be 12% of the size of a 22nm IC with the same number of transistors. Without raising prices, Intel could generate extremely high margins on the new processes. Intel could make multi-billion transistor ICs that could sell for a few dollars at 90% gross margin.
Apple could be the victim of severe device price attrition driven by Google/Android/Motorola and be the loser in their own declared thermo nuclear war. Apple should make peace with Google and move as fast as they can to Intel for the "A" chip manufacturing under the best circumstances available to them.
Google needs to defend their click based business even if it means selling mobile devices at cost. Samsung and Lenovo will suffer collateral damage if Google continues to sell high quality mobile devices at cost.
If Apple moves the fab business to Intel, Samsung will have lost a total of $20-30 billion in component business from Apple in the course of the "war." If the Intel mobile products are good enough, the Samsung mobile division might have to cut off their own chip division and use Intel products.
How Does This All Play Out From Here?
I believe the Mobile business will self organize around an Apple Axis and a combined WinTel and Android/Intel Axis. Let's coin a new word for this; "AnTel", there, you heard it first here.
The device winners will offer a "set" of mobile devices that are compatible with each other and will be operated by a common hardware architecture and a common operating system. The set will be comprised of a conventional WinTel notebook computer (fast boot, 2.5 lbs, SSD, and 10 hour battery time, 13" and 15" models) and, what looks like, a seven inch tablet computer with high end products on Win 8 and Intel iCore processors and low end products will use "AnTel" with Atom based SoCs, and an AnTel smartphone also with an Intel Atom based Application Processor.
Apple will have the equivalent set running OSX and iOS. Right now there are three companies that have the required three member sets of mobile devices. Apple has all three. The Macs are powered by Intel x86 and the Apple OSX operating system. The iPad and iPhone run on ARM based chips using the iOS mobile operating systems.
Samsung has all three products in the set. The PCs are WinTel Ultrabooks, the tablets and smartphones are Android systems using ARM based processors. Samsung might offer a WinTel tablet in the future, but will likely stay with Android and ARM on smartphones at least until an Intel "superchip" becomes available.
Lenovo has all three products in the set. As the earliest adopter of Intel mobile chips and the largest PC manufacturer, one might guess that there is a "special" relationship between Lenovo and Intel. Lenovo has gone from nearly nothing to second place behind Samsung in the important Chinese smartphone market. Sitting on the sidelines of the mobile war will have advantages for Lenovo.
Dell (DELL) and Hewlett Packard (HPQ) are missing the smartphone element of the set of devices required to be an ultimate winner in the device market. Both companies have WinTel PCs, of course, and both companies have various models of tablets. Either or both companies could enter the smartphone business in order to round out the set of mobile devices. Both companies seem more concerned with selling their companies that innovating in the mobile space.
Google is supplying smartphones and tablets through both Nexus and Motorola. They are missing the PC part of the set, so will probably only fulfill the role of price and margin spoiler to Apple's iPhone and iPad business.
The winning semiconductor supplier to the winning device manufacturers will have to have the right processor for each member of the mobile set. Qualcomm can supply ARM based processors for smartphones and tablets, but not x86 processors for PCs. Qualcomm is a fabless semiconductor company, for better or worse. Nvidia can also supply ARM based processors for smartphones and tablet, but not x86 processors for PCs. Nvidia is also a fabless semiconductor company in a weaker position by far than Qualcomm.
Only two companies can supply the x86 processors used in virtually all PCs, Intel and AMD. AMD has been in total disarray for several years and can be disregarded at this time. Intel can supply a full range of x86 processors to the PC segment and is currently holding nearly 90% market share. Intel can also supply Core processors for high end tablets or Atom SoC processors for lower end tablets. Intel is also supplying Atom based SoC processors for smartphones. The Intel Atom processors can and are running on Android systems, both tablets and smartphones. Because the Atom is an x86 processor it can also run full Windows 8.
If the first part of this argument, that the device winners are likely to use the same hardware architecture across the three member set of devices, is correct, then the only architecture that can supply products for all three pieces of the set is x86. The only x86 manufacturer that makes solutions that are fit for all members of the set is Intel. The possibility of an ARM based chip taking any PC market share is highly unlikely. On the other hand, the possibility of ARMH losing significant market share to Intel x86 in smartphones and tablets is almost a certainty.
The device winners will include Apple, Samsung, and Lenovo. Dell and HP could make the list if they make a few correct decisions in the near future. Intel currently has at least twice the capacity required for a flat PC business and are adding more. The only obvious reason for this excess capacity is for a high powered run at all of the mobile chip business.
I call Intel the single major winner as the semiconductor supplier to the device winners. In the process, Intel revenue will double to $100 billion plus. This self organization of the mobile business will take place in no more than two years from today.