I've always tried to keep an open and honest relationship with my readers. I always try to carefully lay out my assumptions, and I'm always happy to listen to arguments that run counter to my conclusions and assumptions. I don't purport to know everything (far from it), despite what I believe to be extensive due diligence, but I do my homework and believe that my understanding of both the business as well as the technological side of the Intel (INTC) story allows me to offer up unique insights.
Very recently, the sell-side firm Piper Jaffray reiterated its "underperform" rating and $20 price target on Intel and cited, what I believe, to be unfair assumptions to justify the downgrade. I certainly understand the bearish argument(s) - and there are certainly valid ones - but I believe that the arguments presented by the team at Piper Jaffray simply do not hold water.
Intel Isn't Fab'ing Apple's Chips - Can We Just STOP?
It has been well known that Taiwan Semiconductor (TSM) would get Apple's (AAPL) iPhone chip business. While I believe that Intel has a significant manufacturing lead (this is well known), Intel is, first and foremost, an IDM. In the semiconductor world, profit & loss - whether you're fabless or run your own fabs - is determined by how much money you can make per wafer. If you're an IDM like Intel, not only do you get to keep the margin that a traditional foundry would keep, but you receive a return on the investment in the design and development of an actual product.
This means, quite plainly, that Intel would much rather sell its own, in-house designed chips and get a return not on just the wafer fabrication, but also on the design of the products. However, there are a number of reasons that I believe that the notion of Intel running a selective foundry for strategic customers is a smart idea.
Sweet, Sweet Margins
The trend in many sectors of the semiconductor space is consolidation to a few key players, particularly as products become more complex and the costs to compete just keep skyrocketing. The majority of semiconductor firms have their chips built at Taiwan Semiconductor, which means that the playing field is level. The idea is that since Intel has higher performing and lower power transistors than Taiwan Semiconductor (yes, transistors are about much more than simply size), it can - in the case of a duopoly - serve as a "tie breaker" to give one firm an advantage over another. With better transistors, design teams can extract higher performance and feature integration. Altera (ALTR), a firm locked in such a duopoly, is a perfect example of the kind of "strategic" customer that is willing to pay up for an "edge".
No, Intel Isn't Going To Fab For Its Competitors...
So, you're probably wondering, "well, why doesn't Intel fab for the likes of Qualcomm and Nvidia?"
It comes down to the following. See, I think the majority of the market is still incredibly confused about what's going on in the low power space. There's this notion that Intel's "X86 can't be low power", but it's an absolutely misguided one. First of all, "X86" isn't a "design" - it's an instruction set. This means that it's just a set of commands that any processor built around it has to be able to run. The micro-architecture determines how well these instructions can be run, how much power it takes to run them, and so on. Until now, Intel's designs haven't been targeted at the power levels (and didn't have the deep idle power states to keep the processor from consuming power when it wasn't doing anything).
What I'm saying is that there's nothing inherently "good" or "bad" about either X86 processors or ARM processors for any particular market segment. Someone could spend the money and time to build an ARM chip that goes toe-to-toe with the best Xeon processors in high performance, or Intel could build a low power X86 chip that can fight honorably in the low power space. What matters, at the end of the day, when we're talking about power efficiency of a chip are the following:
- Process Technology: There's no way around it; transistors are the building blocks of all semiconductors. The lower power and higher performance your transistors are, the better your chip has the potential to be.
- Micro-architecture: There are thousands and thousands of design choices that go to determining what the actual chip design looks like (i.e. how it actually runs the ARM or X86 instruction set), and this is, given equal process technology, what determines power efficiency and performance. A design targeted to go into supercomputers is obviously going to have a different set of trade-offs and choices than a design targeted for a smartphone. Intel's designs up until now have not been for the smartphone/tablet markets. (The original Atom platform was for MIDs and netbooks.)
- Physical Design: With good transistors and micro-architecture, an inexperienced design team could botch up the physical implementation of the micro-architecture. Typically, though, a better physical design usually means higher clock speeds at a given process generation and micro-architecture. This is why Apple bought Intrinsity - to help it optimize the physical design and layout of the chip for maximum power efficiency.
So, there's no real "black magic" that prevents Intel from making a power efficient design, and this view has slowly permeated its way to the markets. Now, in light of this, ask yourself the following: does it make sense for Intel to build chips for its competitors and share the profits when it could keep everything under one roof and keep the profits all to itself?
See, if Intel were to build, say, a Snapdragon for Qualcomm, then for each of these Snapdragons shipped, a potential Intel socket would disappear which means less margins for Intel, and less return on all the money spent designing those shiny new Atom processors. Make no mistake; there's no socket for Qualcomm (the world's leading mobile SoC vendor that sells more chips than Apple and Samsung combined) that Intel can't potentially win. This means that the majority of the mobile market is open to Intel for the taking if it provides a superior product and is able to market it properly.
But what about Apple? Well, this one's thorny. On one hand, Apple very much fits into the following definition of "Strategic Customer" as laid out by Brian Krzanich,
If there was a great customer that we had a great relationship with laptops and other mobile devices, and they said look, we'd really love you to build our ARM-based product, we'd consider it. It depends on how strategic they are,
But on the other hand, Intel's own home-grown chips are being sold to tablet and phone vendors and it is in Intel's best interest to sell its own products, so there's some real conflict of interest. Further, given that Intel's foundry strategy is selective and based on finding companies desperate for an edge (and Apple's problems right now have NOTHING to do with silicon technology), this seems an inappropriate relationship. TSMC, as a neutral party, is much more appropriate for this task, and while being one generation behind Intel in terms of process technology isn't great, it's not a show stopper given that Apple lives and dies by its software ecosystem and brand awareness, not its silicon.
...But Who Cares?
So, Piper Jaffray made a big deal about Intel not winning this business as though it were a surprise; it's not and anybody seriously following Intel would know that an Apple-Intel foundry deal was never likely, no matter how much short term traders wanted it to happen.
Now, Piper Jaffray believes that since Intel didn't win the Apple foundry deal, and since it's not going to win every Samsung device ever, that it's shut out of the mobile market for CY2014 and CY2015. Are these guys serious?
Do they forget that the tablet market is just beginning and that many companies are playing to win big time? While Intel has a good relationship on the PC side of things with Samsung (and again, Intel is in both the Galaxy Tab 3 10.1" and the Ativ Tab - two flagship designs based on older Atom silicon, not the new good stuff coming later this year), these analysts forget that there's an army of players - all of whom have excellent relationships with Intel - coming to market in force with both Windows and Android tablets (and Intel can participate in both). These firms are,
Of course, it helps that every major Taiwanese ODM will be building around the upcoming "Bay Trail" platform, so you will see the market full of Intel based tablets this Holiday season from all of the aforementioned OEMs.
So, with a huge army of Android and Windows tablets coming with Intel silicon ranging from 7" to 11" and from sub $199 to $499, does it really matter that Intel isn't building Apple's A-chips? No, and in fact, each one of these Windows and/or Android tablets that gets sold is a lost sale for Apple and a market share shift in Intel's favor...and Intel gets more than just foundry margins for these chips.
Smartphones May Take Longer To Penetrate
Intel's problem in smartphones at this point has nothing to do with the apps processor and everything to do with the modem/connectivity. While Qualcomm and Broadcom (BRCM) have everything from low power WiFi to 4G LTE, Intel still doesn't really have that in-house. I'm hearing rumblings that Intel is planning to develop its own in-house low-power WiFi, and Intel's 4G LTE appears to be on-track for the 1Q 2014 "Merrifield" platform launch, but this - and not the processor - is what holds Intel back here.
I expect that Intel will eventually have all of the IP needed to really fight in this space, and the upcoming "Merrifield" platform and its integrated-modem counterpart "6331" should be able to penetrate the smartphone markets in a more meaningful way than the "Clover Trail" and "Medfield" processors were able to as a result of a much better apps processor built on a leading-edge node as well as an LTE solution. I remain skeptical, however, that Intel's 22nm smartphone SoCs will demonstrate a significant enough lead in performance/watt over its competition to immediately displace Qualcomm in key sockets.
I expect, however, that at the 14nm generation with the "Airmont" based SoCs (which I hope Intel is pulling in to 2014 rather than 1Q 2015 if the 1 year cadence holds) Intel has a chance at a sizable lead, but TSMC's 16nm process should be ramped in volume by 2Q 2015 if the firm stays on schedule, which could diminish Intel's lead in this space significantly. I need more data on Intel's upcoming 14nm process before I can come to any meaningful conclusions, however, and such data has been elusive.
Realistically, I think I need to see data on Merrifield (I have seen none, unfortunately) before I would know if the solution is credible, but I certainly don't think that there's anything structural preventing Intel from putting out a competitive solution; it's just a matter of spending the money and making it a priority, which Intel is doing.