In a recent article entitled "ARM Holdings: The Real Reason That it Ends Badly", Ashraf Eassa argued that ARM (NASDAQ:ARMH) was overvalued because expectations of future growth in ARM processors fail to take into account the "pincer attacks" Intel (NASDAQ:INTC) is about to unleash in the form of the Haswell 4th gen Core processors and Bay Trail Atom processors, the first that will use the 22 nm process of Haswell and Ivy Bridge (3rd Gen Core). Eassa expects ARMH to "become a very good structural short as growth expectations fizzle out. I believe the dominoes begin to fall in 2H 2013, but the wreckage won't truly manifest itself until early 2014."
While I agree that ARM Holdings is overvalued, I don't foresee anything approaching "wreckage" for ARM in 2014. Intel will provide healthy competition for the ARM ecosystem of processor and mobile device makers, but the ARM ecosystem is far from the problem plagued economically marginal industry that Eassa depicts. The ARM ecosystem can and will rise to the challenge posed by Intel and Microsoft (NASDAQ:MSFT) and emerge stronger for it.
Yes, it's Overvalued
In the chart below, I plot some key financial metrics for ARM starting with Q4 2009, drawn mostly from ARM's quarterly reports. While ARM processor shipments and revenue track each other fairly closely and have grown 25-30% per year, operating income has outpaced revenue growth as ARM has slashed expenses. But even operating income growth cannot fully explain share valuation, which put on two growth spurts in Q1 2011 and again in Q4 2012.
The jumps in Q1 2011 and Q4 2012 correlate with announcements by Microsoft involving ARM processors. The first at CES announced the Windows on ARM initiative that eventually became the RT variant of Windows 8. In Q4 2012, Microsoft rolled out its new Windows Phone 8 platform with partner Nokia, as well as unveiled Surface RT, both in time for the holiday season. Both occasions apparently encouraged ARM investors that there would be a vast expansion in the market for ARM processors, courtesy of Microsoft.
That vast expansion hasn't really occurred. As IDC's tablet market share data for Q1 showed, there were only 200,000 Windows RT devices shipped in the quarter, far fewer than the number of Windows 8 tablets at 1.6 million. Windows 8 tablets are effectively cannibalizing Windows RT tablets, but there continues to be some optimism for Windows Phone on the strength of Nokia's earnings report. Nokia (NYSE:NOK) did increase sales of Lumia Windows 8 phones from 4.4 million in Q4 2012 to 5.6 million in Q1, and Microsoft added Windows Phone partners Huawei and Samsung to Nokia and HTC. This seems enough to boost optimism for Windows Phone on ARM and to some extent ARM itself.
I'm still waiting for IDC and Gartner to publish their smartphone OS market share data for Q1, but based on what I've seen so far, I believe the optimism for Windows Phone is misplaced. Between Q1 2012 and Q1 2013, IDC's data shows that both HTC and Nokia dropped out of the ranks of the top five smartphone vendors, hardly an endorsement of the Windows Phone strategies of either company.
ComScore just released survey of smartphone usage share in the U.S. showing that Microsoft had increased its share by just 0.1% to 3.0% from the previous quarter. Kantar Worldpanel also showed that sales market share in the U.S. was just 5.6% in Q1, although this was a gain of 1.9% year over year.
The small but significant share gains of Windows Phone may be enough to satisfy Microsoft, but I doubt that it will long sustain the hopes of ARM investors. If it becomes apparent that Windows Phone sales are slowing down, or if Microsoft decides to jump to Intel for Windows Phone as part of an "all Intel" strategy for Windows 8.x and a phone derivative, investors can expect a correction in share price back to something commensurate with earnings growth, or about $30-35/share.
Eassa seems to approach the ARM ecosystem from the perspective of the Intel commodity processor paradigm for PCs in which the commodity processor maker supplies CPUs to the PC box maker. He claims that the majority of ARM SoC vendors are having margin problems, pointing to Nvidia (NASDAQ:NVDA) and Qualcomm (NASDAQ:QCOM).
But these aren't the most important players in the ARM space; Samsung (OTC:SSNLF) and Apple (NASDAQ:AAPL) are. In Q1 Apple had mobile device revenue of $36.7 billion and Samsung had mobile device revenue of $28.9 billion compared to $6.124 billion in total revenue for Qualcomm or $4.28 billion in revenue for Nvidia's entire fiscal year. Yes I know that Apple and Samsung don't just make processors, but entire devices. My point here is that successful ARM processor manufacturers are vertically integrated device makers. ARM processor makers who attempted to emulate the Intel paradigm have either failed (TI, Freescale) or are on their way to failing (Nvidia, Qualcomm).
As I pointed out in "Qualcomm and the Demise of the Commodity Processor", the ARM processor makers such as TI (NASDAQ:TXN) have become the victims of a major paradigm shift in the way mobile computing processors are made. TI management specifically pointed to the growth of custom SOCs (systems on chip) used by companies such as Apple and Samsung as the reason they were getting out last November.
Eassa also asserts that everybody is "using the same core/graphics IP, in which case your market degenerates into a commodity one". In fact, customization in the ARM world goes quite a bit deeper, with the major companies like Samsung and Apple not just licensing logic designs from ARM, but building their own unique ARM-compatible designs in the AMD Intel-compatible vein. This has been verified by Chipworks which has actually opened up Apple and Samsung SOCs to verify key elements of their logic and fabrication process. According to them, Apple's latest A6 and A6x SOC's use completely original processor core designs.
Also, even if a company licenses a core logic design from ARM, there's still quite a bit of work to do to implement the design in silicon, and specific implementations can be very foundry process dependent, all the more so as process sizes shrink. So there's still a good deal of differentiation that can exist even within a given logic design, such as the Cortex A-15.
Even without this differentiation, Eassa's assertion of "commodity degeneration" shows a fundamental lack of understanding of the role of an SOC designer such as Apple. Building devices has always been about systems integration, combining different pieces of hardware and software into a working, marketable product. PC box makers are systems integrators, and the different choices of hardware and software features they select provide the main means of differentiating their products from competitors. As more functionality was brought into the SOC, it was logical for device makers to want to customize the SOC with just the functionality they wanted. The systems integration role has descended to the silicon level.
Most importantly, the SOC integrators aren't cost disadvantaged for doing so. Intel lists the tray price for the Atom Z2760 SOC (built using Intel's 32 nm process) as $42, while the estimate for the Apple A6 SOC in the iPhone 5 (built using Samsung's 32 nm process) is $17.50, according to iHS/iSuppli. I know this isn't exactly an apples to apples comparison (excuse the pun), but hard data on production costs for either company are impossible to come by. The point here is that I doubt that the cost structure for the fabless SOC integrator is significantly worse or better than for the traditional processor manufacturer.
Eassa also refers to production cost issues with TSMC's new 20 nm process, implying that ARM SOC makers will be disadvantaged relative to Intel's new 22 nm Atom SOCs due later this year. This is actually a recycling of some very old news. The issue of the cost effectiveness of TSMC's 20 nm process was first raised by Nvidia in a public presentation in November 2011, in which Nvidia projected that it would be mid 2014 before the 20 nm process became cost competitive with the 28 nm process at TSMC.
Apparently, Nvidia hoped that by airing TSMC's dirty laundry, they would force a price break from the foundry. I doubt that it worked, but Nvidia's charts made the tech media rounds and became widely accepted conventional wisdom. At TSMC's 2013 Technology Symposium on April 9 this year, TSMC basically said that the 20 nm process was ready for production. What that means in terms of cost remains to be seen, but TSMC claims that 20 nm also offers a 20% speed improvement, a 30% power reduction and a 1.9X transistor density increase. Even if the cost per transistor remains the same (which would be unusual since IC fabrication cost scales with IC area), I bet many manufacturers will choose the 20 nm process just to get the performance benefits.
Eassa, in raising the straw man of TSMC's 20 nm process appears unaware that there are other ARM foundries. Global Foundries, the spin-off from AMD, is preparing their own 20 nm SOC fab at their Saratoga, NY facility. The ARM ecosystem is large enough that I'm convinced that the demand for 20 nm SOCs will motivate TSMC and Global to rise to the challenges of the 20 nm process and make it work.
The Problem of Argument by Analogy
Much of the problem with Eassa's viewpoint is that it's a recycling of an often used argument by analogy with what happened to the PowerPC. Intel was able to drive the PowerPC from the PC market by making huge investments in process technology and microprocessor design improvements. When Apple switched from PowerPC to Intel in 2006, the triumph of the commodity Intel processor was complete. Many in the Wintel world assume that another such triumph is inevitable in the mobile device world.
But the analogy is superficial in that it ignores significant differences between the two cases. In the 90's the PowerPC was the upstart challenger, and Intel processors already enjoyed a market share and usage share advantage through IBM-compatible PCs running Windows. The PowerPC ecosystem consisted mainly of IBM, Motorola, and Apple. Apple wasn't the only computer maker that used the PowerPC, but it was by far the most important. Apple's market share was miniscule, however, and IBM and Motorola didn't have the capital or the will to keep up with Intel.
Today the tables are turned, with Microsoft and Intel the upstart challengers to the well developed ARM ecosystems of iOS and Android. Google, Apple and Samsung are flush with cash, and can easily afford to invest billions in foundries or design houses, whatever it takes to match Intel in process technology and design innovation. Wintel tablets, ultrabooks, and (eventually) smartphones will compete with the ARM iOS and Android ecosystems, and compete effectively, carving out a third mobile ecosystem. But Wintel won't wreck the ARM ecosystems, just stimulate them to greater innovation to the benefit of all consumers. Competition is good.
Disclosure: I am long AAPL. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.