Following Intel's (INTC) most recent earnings report, I was actually expecting the damage to the share price to be quite severe; indeed, with a weak top line guide for Q4, I was almost sure that Intel shares were in for yet another pounding. Fortunately for me, that pounding never came - Intel is up from the date that it reported earnings.
Color me puzzled, but I couldn't for the life of me figure out why the short sellers didn't go and have a field day with the stock as meaningful upside catalysts were largely a ways off (investor meeting in a month, earnings in a full quarter). While the "obvious" answers that come to mind are "dividend yield," "cost control," and "gross margin expansion," I'm not entirely sure that these are necessarily the reason that the shares have held up so well.
And then it hit me. It's got to be the new CEO and his aggressiveness in executing a winning strategy. Let me explain.
Intel's newly appointed CEO, Brian Krzanich, is just the CEO Intel needs. It is clear that he's going to be playing hardball when it comes to pursuing every segment of compute. While, naturally, the products that are coming out now (and that will be coming over the next few years) were under development under Paul Otellini, Mr. Krzanich appears to be taking a no-nonsense attitude when it comes to attacking new markets and doing everything possible to "win."
I, and many other Intel bulls, have always maintained that Intel should be using its manufacturing advantage and - more importantly - scale to dominate this market. In fact, the strategy that BK seems to be implementing is quite brilliant: attack the low end of the smartphone/tablet market with products built on the very high yielding n-1 generation manufacturing process (these are also very cheap since depreciation costs have gone to zilch by the time the new node is out). In this space, Intel can engage in a price war while at the same time likely doing it more profitably than its fabless competitors (since it keeps the foundry margin as well as packaging & test margin).
At the high end, Intel aims to have processors that are so good (i.e. dramatically performance/power than the highest end from the competition) that it can charge a premium. This strategy has a good chance of "working" by the following argument:
- If Intel has a 1 generation process node lead, then when Intel is shipping node n+1, its competitors will be shipping parts built on node n (e.g. Intel is on 22nm when others are on 28nm, and will be on 14/16nm when others are on 20/22nm)
- This means that when Intel's 22nm is fully depreciated (paid for by hundreds of millions of PC and server chips), it will be able to sell these parts more cheaply than the fabless players trying to sell 20nm parts (that will be more expensive to build, and the foundries will get a significant cut - leading edge wafers don't come cheap).
- At the same time, Intel should conceivably be able to ramp its 14 nanometer leading edge parts at the same time that the foundries ramp their 20nm. If Intel's chips at the high end offer better performance/watt, then it can charge a premium to the 20nm from the leading fabless players (and again, can undercut with the 22nm parts) all while probably having a lower cost of production.
This is the strategy that Intel, if it is successful at bringing its low power process nodes to the forefront and if the foundries aren't able to close the node gap, can and will use in order to attempt to gain substantial market share in the tablet and smartphone spaces. Of course, things in high tech are rarely as simple as a three-bullet-point outline (and if it were so easy, it'd be in place by now). There are many nuances here worth discussing, and it is these nuances that I am looking to see addressed at the upcoming Investor Meeting in November.
Cleaning Up The Designs
Many often comment that at the 32nm node, Intel's mobile processors were significantly behind 28nm parts from the likes of Qualcomm (QCOM) and Nvidia (NVDA). Indeed, compare a Clover Trail+ on Intel's 32 nanometer process against, say, a Snapdragon 600/800 or Tegra 4, and it will become very apparent that comparable transistor technology simply isn't enough - you need the right designs.
At the 32nm generation, Intel was basically still learning how to put together a system-on-chip and how to develop the right low power IP blocks. This stuff is a massive departure from Intel's traditionally CPU-only approach that it had in place for many years. Intel needed to develop a system-on-chip oriented process (Intel's 45nm SoC process was a disaster, but the 32nm one was good and it appears that the 22nm one is excellent), and it had to put in place the design methodologies to really get into that "SoC" mindset that folks in the mobile and embedded spaces have been used to for years.
Also note that at the 32nm generation, Intel was fighting with a very outdated CPU core ("Saltwell" was a shrink of the 45nm "Bonnell" designed for netbooks in 2008), an even more ancient system interface, and behind-the-times GPU IP. While with "Clover Trail+" Intel used a significantly better GPU from Imagination Technologies (OTCPK:IGNMF), it was still limited to the ancient Saltwell core. I am confident that if Intel had simply done a new processor core at 32 nanometers (i.e. Silvermont at a lower clock, for example), it would have been much more competitive.
However, the "bad" times on the CPU side are gone. Intel's "Silvermont" at 22 nanometers is an excellent CPU with superb performance/watt, and the company plans yearly refreshes to its cores (2014 - Airmont, 2015 - Goldmont, etc.). While some may think that Intel is pursuing a "tick-tock" strategy (i.e. "Airmont" is just a shrink of "Silvermont" without any meaningful improvements), it is actually better than that - I confirmed with the Silvermont lead architect that the "tick tock" model doesn't necessarily apply to the low power IA cores and that meaningful improvements are possible even in the first "shrink."
But there's more than just the CPU/GPU (the latter of which Intel is finally competitive on) - when it comes to phones, the modem/integration counts, too.
While I (and many others) have been pretty vocal about Intel's lack of LTE, it's clear that Intel will be the second source to Qualcomm for LTE modems at the high end. And, frankly, for a company that "missed mobile" this will be a very good position to be in. Broadcom (BRCM), too, will get out a high end discrete solution at some point, although that has been pushed off into 2H 2014 in a bid to chase the low end, integrated SoC market - I expect that it will be the #3 player, if it doesn't end up simply selling its mobile and wireless business to the likes of Samsung (OTC:SSNLF), Apple (AAPL) or even Intel itself.
However, Intel also really needs to get on the ball when it comes to integrated modem solutions. Renesas Mobile had an SoC with a built in modem, Broadcom did it in 3G, Qualcomm has it across its lineup, and even Marvell (MRVL) and Nvidia have it in their lower end solutions. Intel, of course, would need to redesign its modem on its in-house processes in order for such an integrated part to be technically feasible, which seems as though it's at least a year or two away (although I am crossing my fingers for this to not be the case). Intel, quite literally, cannot compete in the low and mid-range of the smartphone market without having a highly integrated part (this was on the roadmap given in 2012 but I have reason to believe that quite a few lower end/integrated product cancellations have happened within Intel - again, really looking forward to the investor meeting for some more color here).
Make no mistake, the higher end of the smartphone market is where the big gross margins are, but the low end of the market is quite large and growing at a much faster pace. Intel needs to play in all of these segments, and I think that the company's strategy is to do so with products built on non-leading edge process technology.
A Process Technology Caveat
Another reason outside of design that this strategy has not yet been feasible is that Intel's low power/SoC oriented processes have lagged their higher end cousins. That is, while 32nm desktop chips were in production during Q4'09, the 32nm low power process wasn't introduced until about Q4'11. The gap shrunk with 22nm with the first 22nm CPU parts hitting the market in Q2'12 and the first 22nm SoC products hitting in Q3'13. From what Intel has telegraphed, it plans to have some 14 nanometer SoCs out in late 2014, so the gap will have shrunken to, at most, a quarter.
That being said, while Intel's tablet roadmap looks to track the above, the smartphones are taking longer. The first 22 nanometer smartphone part ("Merrifield") will not ship until Q4'13 for launch in 1H 2014, and its 14 nanometer successor ("Morganfield") likely won't launch until the end of 2014 for device launches in 2015. While Intel will still have a 3-6 month window (perhaps more, but I wouldn't bet money on it) before designs built on TSMC's (TSM) and Samsung's FinFET processes hit the shelves (that is, assuming everything goes on schedule for them), the lead is not as wide as it is in the high performance space or even in the tablet space.
Of course, even a 6 month window is enough to come in and grab some serious market share if the actual chip designs properly utilize this advantage, but Intel will really have to extend its lead to at least one full generation in low power SoCs if it is to execute the premium pricing/volume strategy outlined above.
The New CEO Is Doing What It Takes
I believe that Brian Krzanich will execute on this strategy and will not let short-term focus on margins or ASPs cloud the bigger picture - Intel and Intel Architecture need to take significant share in the smartphone/tablet space in order to remain viable going forward. The PC market will, at best, grow at a snail's pace, and the server market isn't big enough to support the company's growth.
We are already seeing older "Clover Trail+" chips powering cheap Android tablets (and compared to the SoCs in the low end of the ARM-verse, Clover Trail+ is a pretty good design), and I expect that as the latest chips waterfall down the stack, Intel's competitive positioning across the board will simply become much better.
But make no mistake - this isn't going to be an easy task and I do not envy Brian Krzanich his task. While I think the tablet strategy looks pretty good, Intel faces an uphill battle in smartphones with respect to integration, cellular connectivity, and being able to build a chip good enough to get vendors to not go with their trusted supplier, Qualcomm. But I do think this new guy will do what it takes to succeed, and I sense that Wall Street thinks so, too. Or, at the very least, they're giving the new guy a chance.