Contrary to some of the unsavory accusations that I sometimes get, I'm not here to "pump" Intel (INTC) (if you think that anything I publish has a material effect on a $115B company, then you need to hand your brokerage account to a professional). I am sharing my research and thoughts with a follower base that seems to enjoy what I write. I'm obviously bullish and the thesis is quite multifaceted, as Intel is a very complex business. Further, Intel is a business in transition, so a good deal of the work I do is in trying to determine whether this transition will succeed. If it does, shareholders will make a lot of money. If it doesn't, then at the very least I don't see the shares losing a substantial amount of value. I am long the stock because I believe that it offers a compelling risk/reward profile.
That being said, a year ago, I would have likely been ultra-bullish on Intel taking significant share in every segment of computing including smartphones by the end of calendar year 2014. Today, while I remain extremely bullish on the datacenter and tablet opportunities (the PC is in decline - I get it, although AMD's not going to make a dent in share) my enthusiasm remains muted on the smartphone side of things. In particular, I am concerned that the major merchant chip incumbent, Qualcomm (QCOM), is quite literally the most difficult competitor that Intel has faced in its 45-year history. I also worry about Samsung as the most vertically integrated smartphone vendor on the planet.
Qualcomm Cannot Lose
Financially speaking, Qualcomm cannot lose for as long as people are buying smartphones. Whether Qualcomm's processors and modems power the latest smartphones or not, the company receives a substantial royalty rate on the average price of the actual cellular-device sold. This is quite literally ARM (ARMH) on steroids.
Qualcomm has very wisely used the money that it makes from this never-ending royalty stream (that grows as 3G and 4G device shipments do) to fund its chip and modem development efforts. As a result, Qualcomm currently develops top-notch CPU, graphics, ISP, Wi-Fi, and cellular IP and is able to integrate it seamlessly into its designs. It is quite literally the "Intel" of cellular apps processors, as nobody else in the merchant chip business has a cellular modem and apps processor solution that is as good. Even Samsung (OTC:SSNLF) often eschews its own apps processors for Qualcomm's parts. They're that good.
This is a prime example of a "virtuous" cycle, and Qualcomm's management has been incredibly effective at driving leadership in this space. Intel has quite literally never had a competitor this well managed, this powerful, and this ahead of the curve.
Intel's Phone Story
What I have been trying to have my readers understand is that this isn't about "X86 vs. ARM." Android is instruction set agnostic, and Intel has done the necessary work with developers to make sure that "X86" processors run Android just as well as (if not better than). This battle really comes down to whether Intel can provide a superior solution in this space in a timely manner. The gap has been closing, for sure. The original "Medfield" system-on-chip was competitive with the dual core "Krait" parts on CPU, but the GPU was still not great and the platform didn't come with LTE.
Clover Trail+ was a significant step up - it had a very competent CPU and fairly modern Series 5 GPU from Imagination Technologies (OTCPK:IGNMF). Put bluntly, it was the first mobile SoC from Intel that didn't, well, look embarrassing. Of course, Qualcomm beat Intel to market with a quad core "Krait" paired with a better GPU. This part didn't come with an integrated LTE modem, but Qualcomm was able to supply a discrete LTE modem + RF to its customers. Again, with XMM 7160 (the LTE solution that Intel launched in August) nowhere to be found, Clover Trail+ again could not gain meaningful traction in the smartphone space in the markets that mattered (although the Lenovo K900 was a top-5 device in China).
With the recently announced "Silvermont" processor core, Intel is working feverishly on a new, smartphone and Android-tablet oriented SoC codenamed "Merrifield." This part will be a dual core "Silvermont" part paired with an Imagination Tech Series 6 "Rogue" GPU (the exact configuration is not yet known). Further, this part will actually be paired with an LTE modem (not integrated, but as Apple's iPhone and the numerous phones based on the Snapdragon 600 have shown, this is not strictly necessary). In short, this looks like a part that will finally be competitive at launch, at least in the high end.
However, there's a BIG qualifier here. Just what is Merrifield going to be going up against? If Intel is going up against the 28nm chips in this space, then it will have a very good shot of actually winning designs that matter. While this is not a chip that likely aims to do battle with the Snapdragon 800 in terms of raw performance, it will also be going into much smaller form factor devices than the S800 (notice that everything that the S800 has gone into is a "large" device, i.e. "phablet"). That being said, a very big problem arises if this part goes up against designs built on TSMC's (TSM) 20nm process.
The "good" news for Intel is that parts based on TSMC's 20nm will likely be available in devices starting in either Q3 or Q4 2014 (this is assuming volume production of parts on the node happens in Q2), which means that if Merrifield ships in Q4 2013 for device launches in 1Q 2014, then Intel actually has a pretty nice competitive window. Further, while I haven't been able to confirm this (at IDF, the best I could get was noncommittal acknowledgements that my line of reasoning made sense), Digitimes has reported that Intel plans to release an SoC codenamed "Moorefield" during 1H 2014. If I'm right, it is likely that this is a 22nm quad-core, high-end "phablet" oriented SoC that will be paired with the XMM 7260 LTE-Advanced modem (also set for launch during 1H 2014). If this is the case, then I expect Intel to stay competitive with the 20nm parts (which softens the potential risk of TSMC getting its 20nm act together sooner than expected).
How Does Intel Fight Qualcomm, Though?
Things look great for Qualcomm. It has everything - the LTE modem monopoly (which means pricing power), the never-ending patent royalty stream, top-notch silicon, best-of-class integration, and relationships with all of the major mobile players.
However, something I find very interesting is that Intel's management has continually pointed out that the smartphone players actually WANT Intel in this space. Why? Competition is great for the customers, and against Qualcomm, there is no real competition, which means that Qualcomm gets to charge whatever it pleases. While Broadcom (BRCM) will compete in a limited fashion at the low end of the mobile SoC market in China, and while NVIDIA's (NVDA) Tegra 4i is also likely to gain traction in the low/mid-range, I fully expect that Intel will be the #2 player to Qualcomm. The question, though, is how close a #2 can Intel be?
It depends on the technology and Intel's ability to execute. If Intel is able to deliver a leadership modem + RF front end as well as a top-notch applications processor, then I see no reason why Intel can't compete for - and win - the sockets at smartphone vendors that don't do their own in-house mobile processors (think LG, Motorola/Google, HTC, and Microsoft/Nokia at the high end).
However, that leads me to the next problem...Samsung.
Samsung Will Be A Problem For Both Intel And, To A Lesser Extent, Qualcomm
Samsung is the most vertically integrated smartphone vendor today. It has its own wafer fabrication plants, its own SoC integration teams, an army of software engineers, DRAM, NAND flash, its own cellular baseband group, and of course it has been aggressively hiring CPU architects, suggesting that it is currently developing its own ARM based CPU architecture. It is very tough to beat Samsung's cost structure and sheer resources, and that is a major reason why I think that Apple (AAPL), the giant that kicked off this whole mobile revolution, is in deep trouble longer term.
Intel's problem (along with Qualcomm's) is that if Samsung's in-house modem and/or SoC teams actually get their acts together, and if Samsung continues to gobble up market share in mobile, then not only does Qualcomm lose a bunch of high ASP/highly visible sockets (Galaxy S IV, Galaxy S IV Mini, Galaxy Note, various other "Galaxy" lines), but Intel will have very little shot at gaining meaningful (and high dollar content) share in the smartphone space. Now, Qualcomm is protected by its QTL division which collects a royalty on every cellular device sold, and given that Qualcomm is a foundry customer of Samsung's, there may be less of an incentive to boot Qualcomm out of sockets.
So, Intel can succeed in phones if the following happens:
- Samsung doesn't rule the world, and instead a bunch of smaller handset vendors end up taking the majority of the world's market share.
- Intel's processors are so incredibly good and better than what Samsung's internal teams can do (and better than what Qualcomm can do) that Samsung, in the interest of putting out the best device possible, uses Intel's chips.
While the jury's still out on #1, I'm not convinced that Intel will be quite there with a part that is *head and shoulders* ahead of everything else until the 14 nanometer generation (although that Moorefield chip could surprise me). From a longer-term perspective, many within Intel have given every indication that speeding up Atom development and integration is the top priority on Intel CEO Brian Krzanich's mind. If Intel can truly succeed in keeping ahead of the industry with significantly better apps processors and/or modems, then it will be in a good place. Being "good enough" will not win the prize in phones as these device vendors' internal efforts (even using off-the-shelf ARM IP if necessary) is "good enough."
This is why I am so obsessed with the performance of these chips and the timing of their release. Intel's best chance at becoming a truly leading mobile chip force (that actually makes money at it) is to have unequivocally better products than what anybody - merchant vendor or internal teams - can do today. Intel has done it before in other segments, but right now, Intel is facing unprecedented direct competition from Qualcomm and indirect competition by way of Samsung's internal chip efforts.
I remain optimistic that the rest of Intel's business from tablets on upward will thrive, and I expect that the semi-custom design/foundry initiatives that Intel President Renee James talked up will be a good driver of revenue from a longer-term basis, but I am still very nervous about Intel's smartphone plans. Medfield was meh, Clover Trail+ was OK, and Merrifield looks legitimately "good." But there are a lot of sharp CPU design teams out there all enabled by the common instruction set architecture laid out by ARM.
If Intel actually pulls this off, then Brian Krzanich could go down as Intel's greatest CEO in history (although Otellini's domination of the server and PC markets will be tough to beat). I can't wait for Intel's upcoming investor meeting to get a better sense of just what the company's plans to "win" in the smartphone space are.