In this article, I would like to discuss (and perhaps dispel) the "margin myth" that surrounds Intel's (INTC) attempt to break into the mobile/low power system-on-chip (SoC) market. The points of the bearish thesis are as follows:
- Intel has typically enjoyed healthy ASPs on its PC and server products, which is part of what has allowed the company to enjoy immense profitability. Therefore, as PCs "shrink" or even "die," Intel will be left with a bunch of expensive, unused manufacturing capacity.
- If Intel starts pushing into the mobile SoC business, it will have to accept lower gross margins on these "cheaper" chips.
- Intel is unlikely to have a monopoly position in the mobile SoC space, and therefore will not have the volumes in these smaller, low power chips to keep its fabs filled.
- Intel, as it competes with a bunch of SoC vendors, will face too much competitive pressure to turn a profit.
I believe that such fears are unwarranted and that the mobile SoC opportunity is one that is essentially "pure upside" in the best case, and a nice hedge against weakening PC sales in the worst case.
Intel Can't Fight ARM - The Biggest Myth On Wall Street
One of the biggest myths on Wall Street, perpetuated primarily by those who are not particularly familiar with the technology behind the headlines, is that Intel - a company whose quarterly dividend payments are greater than ARM's (ARMH) projected sales for the year - cannot design a similar/better processor core than ARM. The argument then extends beyond that to state that Intel can't design a world-class system-on-chip.
Look, let's be perfectly honest here. Intel invented the microprocessor, and has been writing the book on power-efficient, high performance CPU architectures for decades. The only reason Intel's products haven't absolutely decimated any of its competition's (powered by ARM cores or custom cores) is that Intel hadn't - until now - really targeted their designs at low power. Does anybody seriously think that Apple (AAPL), which was able to do a better-than-ARM's-own-design core in its first try, has stronger silicon design talent and resources than Intel?
Intel, unlike ARM and its licensees, is a company focused on actually earning money and not touting PR design wins for much lower ASP chips in a commodity smartphone. Intel, instead, plowed its resources into high-end processors with high ASPs that nobody else can match. Despite the hype around ARM, it is noteworthy that Intel's net income in 2011 of $12.5B is more than 12x ARM's entire projected revenue for FY2012 of ~$900M.
However, Intel now smells money in the low power system-on-chip business, and so it is now targeting its designs at that power level. It will succeed, even if it takes a few iterations, and there will be no question of Intel's technical prowess in this area. I would first like to take the reader on a quick detour as to why Intel can now justify plowing so much money into R&D for a low-power processor core and the accompanying SoC design teams.
A Low Power Core Can Be Used Everywhere
If Intel wants to play to win, then it needs to invest heavily to outdo the competition's core and system-on-chip designs. If Intel were just doing it to get into smartphones, then it wouldn't be worth it - the SAM ("served addressable market") is just not particularly large enough for Intel to go after with huge budgets.
However, Intel can expand its low power "Atom" line into a number of key areas that make the investment worthwhile:
- In-vehicle infotainment systems
In addition, the R&D on a low power core translates well into optimizing the firm's higher power core for energy-efficient performance for leadership in those products. There is no incentive to "skimp" here, as the competition will take up any slack that Intel gives.
With that being said, let's get into some nitty-gritty financials (I promise, I will keep tech geekiness to a minimum).
Cheap Doesn't Mean Low Margin
Investors need to be aware of the critical distinction between total net profit and gross profit. The prevailing notion that the mobile SoC business will "lower" margins at Intel is quite unrefined and needs to be looked at closely. Let's talk about margins.
If I sell you a $20 chip that cost me $10 to manufacture, then I have 50% gross margins on the chip.
However, if I sell you a $100 chip that costs me $60 to manufacture, then I have 40% margins on the chip.
One of these has higher gross margins (the cheaper one) but the other contributes more to the absolute top and bottom lines. Which kind of product would you rather be selling?
But the thing is, Intel (and most other semiconductor companies) actually design their products to have quite healthy gross margins. Look, for example, at these fabless semiconductor companies that play in much lower ASP, less "sexy" areas of semiconductors than smartphone/tablet apps processors:
These are good gross margins! Now, keep in mind that these companies are (mostly) fabless, and as such rely on foundries such as Taiwan Semiconductor (TSM).
These foundries also have to take their "cut" (a cost advantage that Intel actually has since it gets paid the cut that the foundries get), which is actually fairly healthy in and of itself:
To illustrate the business models, I have prepared the following images that the reader should study carefully before continuing:
The ARM Ecosystem
The Intel Ecosystem
I assert that if third-party mobile system-on-chip business is viable, then Intel stands to benefit the most from it financially. This is because Intel collects the margins for both a design house such as Nvidia (NVDA), Qualcomm (QCOM), and Broadcom (BRCM) in addition to the margins of the foundries such as Taiwan Semiconductor. All while not having to pay royalties to ARM (albeit these royalties are so small that it's hard to fathom the fabless guys losing too much sleep over it - ARM shareholders should, though, at 76x earnings). There are also some other factors that come into play that people generally don't talk too much about: efficiency.
The big Achilles' heel to the ARM ecosystem model is, in addition to everyone needing their "cut" is that there is a severe inefficiency in CapEx spending. At its analyst day, Qualcomm gave the following slide in an attempt to show why Qualcomm should not fear Intel's manufacturing prowess and vertical integration:
Clever, but deceptive. While this chart may make it seem that Intel should fear the collective of fabless players because the foundry partners spend a lot of money on CapEx, it is important to note that this indicates inefficiency since the CapEx and R&D at the foundries are not additive! If this slide were representative of one foundry partner, then it would be something worth considering seriously. However, the slide previous to this one showed just how many players are spending this aggregate amount of money:
Are investors supposed to believe that because Samsung, SMC, UMC, Global Foundries, and TSMC spend collectively twice as much on CapEx as Intel does, that Intel is at a disadvantage? The CapEx isn't additive with respect to any of the foundry partners as a whole, and the R&D efforts to actually develop new manufacturing processes (this is the hard part) do not "add up." TSMC's R&D is distinct from Samsung's, so they are each spending money to independently reach the same place and then compete with one another.
This is the real area that people seem to completely miss. As my previous chart showing the ARM ecosystem showed, the work done at a fabless design house is done by a different company than the one actually manufacturing the product. Make no mistake - there is a very deep relationship needed between the foundry and the chip designer to get good yields at a good cost.
With an integrated model, the chip design teams and the foundry teams can work together tightly to co-optimize both the process and the design. This allows for better time-to-market in addition to better yields, higher performance, and lower costs. In the foundry/fabless model, there is significant latency and inefficiency between the two parties, and problems take longer to sort out, and more sub-optimal compromises are made to get the product "done."
Intel's Cost Advantage Illustrated
Back to the multi-billion dollar question: margins and profitability. The first argument is that if the traditional PC growth slows/turns negative and these PC sales are being replaced with tablet processor sales, Intel suffers. This is not in dispute (although more on why Intel will be fine in a moment), but it has nothing to do with margins. Let me explain.
As I showed above with the numerous small fabless semiconductor companies, cheap chips can still carry solid gross margins, and the foundries themselves make excellent margins from the manufacture of these devices. So how can a company like Qualcomm make ~50% gross margin on its "Snapdragon" product, and how can Taiwan Semiconductor at the same time make ~33+% gross margins on the manufacture of that same product?
Assume that Qualcomm's Snapdragon S4 sells for $30/chip (it's a high-end chip). Qualcomm sees gross profit of $15 (50%), meaning that TSMC charged Qualcomm $15 for it. At the same time, Taiwan Semiconductor took in ~33% gross margins on what it sold to Qualcomm for $15, meaning that it cost TSMC only $10 to make. Keep in mind that TSMC suffers all of the depreciation, CapEx, and so on that Intel faces. We cannot forget that ARM gets its 1.8% cut from the $30 selling price of the chip - $0.54, so Qualcomm's gross margins are slightly lower than 50% (but negligibly so).
Now, say Intel's next generation "Atom" is a smash hit and Intel can sell it, too, for about $30. Intel pays nobody for fabs, so it keeps the $5 that Taiwan Semiconductor made in gross profit. It also, for the sake of argument, gets the same "fabless design" margin that Qualcomm gets of $15 from the chip. That means that for a similar product, because Intel isn't paying anybody's foundry margin, it earns $20 in gross profit for gross margins of 66%.
These numbers are, of course, not actual data, but the point illustrated is clear: Intel has a significantly better cost structure. This will enable either one of two things:
Intel is trying to build market share here, so perhaps it doesn't want to generate *more* profits per chip than its competitors. The company has the better cost structure, so say it, too, is willing to accept the same gross profits as its competitors on the chip sold. To earn Qualcomm's $15 in gross profit, it need only sell the chip for $23. To the phone vendor trying to save every dollar possible in order to fight the juggernaut that is Apple and Samsung, this $7 savings across millions of units adds up quite nicely. Further, Intel will be ahead technologically (smaller/better process technology allows the chip designer to put in more aggressive features in the same/less area), so the phone vendor gets an equivalent product for cheaper.
More Transistors = More Features
At the same feature level, Intel will be able to command a lower price and still make the same (or better) margins. But the better transistor technology also allows Intel to go ahead and build higher end chips that blow the doors off of the competition's products. Why? It's simple. Intel has years of experience designing high performance microprocessors, so it need only bring that massive experience down into the low target power designs.
These higher end chips would be better than the competition's, cheaper to make, while at the same time commanding equivalent or even higher prices to the best that the other companies have to offer.
Qualcomm Still Is The King Of Connectivity
Now, I've been picking on Qualcomm throughout this, but the competitive landscape will not be so cut-and-dry when it comes to a total systems level view of things. While Intel will very likely lead any of the ARM vendors on the CPU side, and possibly on graphics (although Qualcomm and Nvidia are both very strong here and Intel has not traditionally had a strong graphics architecture), Qualcomm is the king of connectivity. Period.
Qualcomm's next generation "Snapdragon 800" products will actually sport quad core next generation "Krait" cores with integrated 4G baseband silicon. Intel is just starting to ship its first discrete 4G/LTE baseband silicon to be paired in the platform with its apps processors. This has been the reason that Intel's products have not yet shown up in the US, and I believe that until Intel is able to actually deliver a strong 4G platform, it will see limited traction in the US (which is needed to build positive sentiment).
Further, judging Qualcomm based on its ability to lead apps processors is a little naive: most of its profit comes from licensing of its wireless patents. The apps processor business is certainly a growth area for Qualcomm and I expect them to continue to be a major force in the smartphone space, but the real reason that they continue to dominate is that they are, as of now, untouchable in wireless.
The hope is that Intel really speeds up the development of its comms processors and gets 4G out to market, but the fruits of this labor are likely not to be seen until the 2014 timeframe. Qualcomm will be the undisputed provider of smartphone silicon until then.
How About The PC Slowdown?
The PC processors have higher ASPs. While a tablet or smartphone chip commands $20 - $35, a notebook or desktop processor has an ASP of ~$100. Chips in the data center (server, workstation, cloud) have even higher ASPs. So the big question is: will Intel be swapping out $100 CPU sales for $25 - $30 apps processor sales?
I believe the answer to this is actually "no, not really." Intel is making significant investments in sprucing up the notebook and the desktop in a way that hasn't been seen in many years. As the focus has been on making PCs "cheaper" and "faster," people have concluded that even their older PCs are "fast enough" and that an upgrade can be put off. Since the laptop PC hasn't been a "sexy" device, and since even a fairly old laptop is fast enough for general use, the focus now needs to be on selling form factor and selling "user experience."
Intel is making good headway with the "Ultrabook" initiative, and I believe that the exciting designs coming from the OEMs will help to bring back the wallet share to the notebook, especially as these notebook designs are increasingly "convertible" (tablet + notebook in one).
However, even if the upgrade cycles lengthen permanently from ~4 years to ~5 years, Intel will be a big player in both Android and Windows 8 tablets, which could help offset the damage. The Achilles' heel of 4G connectivity just isn't there in the tablet space, so the firm's products are likely to see strong sales, especially as its next generation "Bay Trail" processor can enable fundamentally strong tablet experiences on the familiar and popular Windows and Android platforms (read: high end, iPad-like or better).
Time for some math to quantify the different cases:
Case #1: Ultrabook Initiative Fails, PC Cycle Extends To 5 Years, No Growth In Emerging Markets - Pessimistic
The $100/chip spent every 4 years (~$25/yr) becomes $100/chip every 5 years, which would lead to a 20% reduction in PC chip revenue as a long-term value. We also assume no unit growth from first time buyers (emerging markets, new hires, etc.) and make the very bold assumption that we are in a totally "refresh" market.
Since PC related revenues make up ~66% of Intel's sales, then in this ultra-pessimistic scenario, we see a drop from ~$35B in PC chip sales to $28B, or an impact of roughly $7B in sales.
Case #2: Ultrabook/Better Macro Restores Low Single Digit PC Growth, Tablet/Smartphone Negligible
In this case, we see 2012 as a fluke and 2013 and beyond return to low-single digit revenue growth for the PC client group's products. We assume the big growth story from emerging markets as such has very little impact, and we assume that refresh cycles do not shorten.
In this case, with a pessimistic tablet/smartphone view, the majority of the growth comes from the data-center group (~15% CAGR through 2016).
Case #3: Ultrabook Fails To Ignite Sales, But Tablet Push Succeeds
We are forced into a $100/5 years cycle on the PC side, but we add $25/2 years for tablets/smartphones, and assume roughly equal units of PCs and non-PCs for Intel (this assumes a ~20% market share of these non-PC devices long term). The refresh cycle for these tablets/phones is, I believe reasonably so, expected to be faster than the PC refresh rate given that these are consumer electronics devices that live and die quickly (and come at fairly inexpensive price points).
This leads us to a long-term PC + mobile scenario of $32.50/year, which is a healthy increase from the $25/year in the case of 4 year PC refresh, and no mobile contribution. Of course, it is trivial to extend this to the most optimistic scenario of PC refresh goes to 4 years and tablet refresh is at 2 years, giving us a very optimistic $37.5/yr value, or roughly a 50% increase from today, although the risk to this is that if Ultrabooks succeed, the case for the refresh of the pure tablet at all diminishes.
Conclusion - Key Takeaways
The key takeaways are as follows:
- Mobile chip ASPs are lower than PC chip ASPs, but actual gross margins are quite good.
- If Intel gets mobile chips selling in high volume, then fixed costs become less of an issue and the division can add quite a bit to both top and bottom lines.
- Tablets/smartphones unlikely to cannibalize PCs, but if they lengthen upgrade cycle, and if Intel can take significant market share, the lengthened PC cycle can be offset by shorter tablet cycles (this helps make up for the lower ASPs).
- Intel has a cost advantage over fabless peers, so if the mobile chip business is viable, Intel is certainly viable in it from a margin standpoint.
- Intel has "everything to gain and nothing to lose" with regard to tablet and smartphone market share, so assuming that the PC sales can be stabilized, this market represents a pure upside opportunity to the top and bottom lines.
Additional disclosure: I am short $QCOM puts.