In the last five years or so, ARM Holding's (ARMH) profile in the investing community has risen considerably, and it has become inexorably linked with the mobile revolution, the post-PC world etc. Moreover, Intel (INTC) has gone from trivializing ARM's threat when talking to its own investors (the tiny revenue critique) to completely reorganizing (CEO transition, going node first on mobile etc.) around trying to compete. With the stock now trading at arguably lofty valuations the question really becomes one of growth. Will said growth top out with the smartphone penetration rate, or is there more to the story? The (limited) goal of this article is to touch on some of the primary growth drives which I believe make ARM a compelling investment. The goal is not to try and quantify how much growth could come from each as that would be so much hand-waving, rather it is to point out some of the more material potential opportunities that lie ahead of ARM.
Smarter Phones, More ARM Chips
It's hard to talk about ARM's growth without talking smartphones. The smart phone market is growing at something like 20% per year. However, more importantly smartphones are also continually getting smarter, with superphones being the current preferred brand for really smart phones. Bottom line, increased functionality has led to acceleration in not only the number of ARM chips per phone (average = 2.5, smartphone much higher) but also to a shift to higher royalty rate ARM chips (i.e., Cortex A series). Thus, ARM benefits to an ever increasing degree (vs. many other component suppliers) with the shift from feature phones to smartphones and then to superphones. With Big.Little threatening to double the core count on top end phones (Samsung (SSNLF.PK) and others will likely announce Octo-core phones shortly), this acceleration in the # of ARM cores per phone shows no sign of slowing down.
Microcontrollers and The Internet of Things
We live in a world where even the most tech savvy of us can count the number of connected devices we use on one hand. Our phone, computer, maybe a tablet or TV. Fast forward 5-10 years and the world will look very different. In this time frame we can expect products as diverse as credit cards, lighting, motors, household appliances of every kind, health equipment, and even sensors in the human body to start to be connected. The bottom line is that as the cost (and physical size, the ARM M0 is 1mm squared) of the chips required to do this fall, tremendous opportunities emerge to make existing products more efficient and useful, and to create entirely new products, services, and business models. Cisco (CSCO) (quite possibly conservatively) estimates there could be 50 billion connected devices by 2020, many in entirely new product categories. For comparisons sake, ARM shipped 3.9 billion cores in phones (equating to fewer actual devices) in 2011. As such, the internet of things has the potential to take ARM's pennies per chip business model onto an entirely new level. Long term, I believe the Internet of Things opportunity has the potential to equal, and potentially exceed, ARM's current smartphone revenues.
The A50 processor line, based on the ARM v8 architecture, is ARM's first 64-bit technology. Announced chips include the A53 (A9 performance, quarter of the power), and the A57 (1.4 *A15 performance, 60% of the power.) One thing to highlight is that ARM is very conservative in its performance claims (Intel take note) and, like Apple (AAPL), has a reputation for delivering what it promises. While both chips can run 32-bit code in legacy mode, the A53 is likely aimed primarily at the smartphone space with the A57 being targeted more at computing and server (and superphones in big.Little configurations).
ARM's ability to take meaningful share in computing is currently one of the more interesting unknowns. My own view is that a for a very large chunk of PC users we have reached the era of good enough processing and that puts extra emphasis on price and power/thermal efficiency. With Microsoft (MSFT) slowly moving toward full ARM support with Windows RT now in the market, and Apple apparently toying with moving its full O/S to ARM (I am quite sure they have done the hard labor here, whether it will ever come to market is perhaps not even decided in Cupertino yet) the software position has never been more favorable for an upstream move by ARM.
The current best example of a mainstream ARM computer is the Google (GOOG) Chromebook (which also comes in an Intel flavor) which runs Chrome O/S and is powered by a dual A15 Samsung Exynos SoC. Reviews have been broadly positive and the device held the top spot on the Amazon (AMZN) laptop chart for several weeks. My guess is that this much simpler web based PC model is going to take a considerable amount of share in the next 10 years as high speed internet becomes entirely ubiquitous. The reality is that for many users Wintel have made a fortune selling way more power and capability than many users require, and the Chromebook is on the vanguard of better segmenting the market by need. Browsing, Social Media, document creation and editing, Skype, Angry Bird type games, Netflix (NFLX), and email pretty much sums up what many people want/need. Add virtual instant boot, a virus free environment, auto updating of everything, and that's a disruptive value proposition for a lot of consumers. Quad A15 level performance at 20nm might turn out to be an inflection point for this market. How big it turns out to be, and whether ARM can move beyond it, remains to be seen.
The Server Opportunity
The A50 (which is modular and with the ARM interconnect can initial be used in up to 16 core configurations) is essentially ARM's first serious server chip. Announced licensees with probable server intent include Samsung , AMD (AMD), Broadcom (BRCM) and Calxeda. It is somewhat remarkable in such a (relatively) mature industry that almost all the major server players have announced plans to design ARM based server products. While a critic could rightly argue that many may not see retail launch, the energy around an alternative to x86 is palpable (though perhaps not surprising with Intel becoming a complete monopoly and power becoming such a runaway cost for server farms). HP (HPQ) cites its Redstone architecture as using 89% less energy, and requiring 94% less space, while achieving performance parity with its x86 products.
There is a long way to go, and a lot of software to be optimized, but when a company such as HP quotes such astonishing figures there is clearly potential for ARM to massively disrupt the server space. With approx. 75% of the server market being addressable with the A50 line (vs. around 10% with the current A15 products) 2014 and beyond is going to get very interesting. Notably, AMD's ARM server announcement was attended by Red Hat (RHT), Amazon, Facebook (FB), and Dell (DELL). Put differently, software players, OEM's, and major server farm customers are already engaged in forming a new ecosystem. It is no coincidence that the Intel hype machine has sprung fully into action claiming its rebadged Atom Centerton was actually a microserver contender all along. Calxeda have a nice blog post commenting on said hype.
The reality is Intel is terrified to fully go after the microserver model for fear of cannibalizing their own high margin server chips. While they dilly-dally, the ARM ecosystem is not going to wait. Of course, the next Intel CEO might be more proactive in this regard. If I was on the board, I would not hire anybody who was not going to be.
Clues from the Balance Sheet
While I believe much of ARM's potential growth lies ahead of it, we can already see signs of further revenue acceleration. ARM does not disclose its backlog (essentially contracted but un-invoiced revenue) but we can, however, see the deferred revenue. This can be thought of as the recently invoiced flow of (as yet unrecognized) revenue out of said backlog. They key point here is that deferred revenue is growing much faster (in fact almost twice as fast) as recognized license revenue itself. Put simply, acceleration in the growth of deferred revenue foretells acceleration in top line growth. Further, given that deferred revenue likely relates predominantly to recently sold licenses, this acceleration in the growth rate of deferred revenue gives us material insight into ARM's ability to monetize licenses sold for its highest end chips.
But, Intel's next chip will change everything in mobile, right?
Sure. And was supposed to every year going back for some time. Moorestown was the next big thing, except it hardly shipped at all after Nokia (NOK) left it at the altar. Never mind, Medfield, Intel's first true SoC arrived to save the day and shipped in about 6 phones (yes 6, with the Motorola socket likely paid for). However, the future does not have to predict the past and if Intel (currently) has one thing going for it that would be process superiority. Intel's next next next chip to change their mobile fortune will be a 22nm finfet out-of-order Atom (aka Silvermont) technically still due in late 2013 but actually due in 2014. This will likely compete with second generation ARM Big.Little designs, a tweaked Tegra 4 (with Tegra 5 likely sampling in the Silvermont time frame), a 20nm Krait refresh etc.
Everything Intel has said suggests to me that they will be competitive with what the ARM ecosystem will have on the market in the second half of 2013. I am not sure that is how you steal share. Intel has no mobile ecosystem, their innovation seems overly biased toward process, and the industry they seek to enter has displayed little enthusiasm to let them in (they have seen how that movie ends). You can't count Intel out, but at the same time they have currently shown nothing to truly threaten ARM. The 2011 resignation of their senior VP of mobile, and now the departure of their CEO, would appear to be indicative of a company struggling to adapt to a new world.
Moreover with the PC in decline there looms the stark possibility that Intel's fab utilization is going to fall and the money making machine stutter. Their process advantage has been funded by historic quasi-monopoly returns, as those returns normalize then so to will their process advantage. Many seem to assume their process superiority is a permanent genetic advantage, it is not, it is a function of greater historic investment. As Intel navigates the post-PC world it remains to be seen if their investment scale will be sufficient to keep them so far ahead. I rather doubt it. With TSMC rumored to be building an entire Fab for Apple, Intel's relevance is falling.
Investing 101 fans of static ratios will tell you ARM is overvalued, in a bubble, destined to crash, an obvious short etc. Of course, if investing was as easy as selling/shorting firms with high PEs, it would be a rather simple endeavor. The fact is ARM's growth is largely in front of it and its CAGR is the envy of the semiconductor space. This makes static ratios rather problematic, especially the PE. As an alternative perspective, if we look out a year (and thus ignore Internet of Things, meaningful server and computing revenues etc.) with a 2014E PE of approx. 32, and an earnings CAGR of approx. 21-25%, ARM has a 2014 PEG in the 1.2-1.4 range, very similar to Qualcomm (QCOM), and actually not incomparable to Intel (0.9). In essence, just paying for the current (so basically phone, AV, and microcontroller) business ARM looks to be valued at the top end of what you might expect (there is clearly a bit of Apple-esque hype in the stock). However, if you think the IOT, server, mobile computing beyond phones etc., opportunities are real, it starts to look much less expensive.
However, the reality is valuation ratios really don't tell you very much about the actual direction of a business and can easily be massaged to fit whatever story the author prefers. They are far too often the tail wagging the investing dog. Certainly, If you don't like buying stocks at 52 week highs, or PEs that make you feint, ARM is clearly not for you. I can think of several semiconductor players that can offer you paltry PEs, close to 52 week lows, and all the dividends you can eat to boot. Bottom line, sometimes you have to pay for growth and potential and ultimately the only thing that matters is whether you think that growth will be more or less than the market is anticipating. In the case of ARM, I don't yet believe the Street has fully comprehended the long term potential.