As Fall approaches, Intel (NASDAQ:INTC) is finalizing preparations for its assault on the ARM mobile device ecosystem. ARM processor makers such as Nvidia (NASDAQ:NVDA) and Qualcomm (NASDAQ:QCOM) will be in the trenches, bearing the brunt of the Intel attacks. While Qualcomm is financially strong and its diverse product portfolio shields it to some extent, Nvidia is much more vulnerable.
Starting this quarter, Intel will unleash three waves of processors intended to strike at different markets within the ARM tablet and phone ecosystem. The first wave will be ultra low power Haswell 4th Gen Core devices capable of being used in tablets and 2-in-1 convertibles, essentially tablets with removable keyboards. These tablets will have the thin and light form factor of ARM tablets but will run the full 64 bit version of Windows 8.1. The impact of the Haswell tablets will be more psychological than economic for a simple reason: the typical Haswell tablet processor will cost about 10 times as much as an ARM tablet processor.
Intel will launch the second wave of its attack when products with its Bay Trail series of Atoms ship in Q4. For the first time ever, Atom processors will get the benefit of the advanced 22nm fabrication process used in Haswell. This will allow them to consume significantly less power than current generation ARM processors, while being faster and more computationally powerful. This is mostly due to the limitations imposed by the 28-32nm fabrication processes available from ARM foundries such as TSMC, Samsung and Global Foundries.
Intel's third wave consists of the Merrifield systems-on-chip (SOCs) for smartphones. Merrifield is based on the same architecture as Bay Trail and will use the same fabrication process, but includes wireless modems suitable for phones, including LTE. Merrifield phones won't arrive in stores until Q1 2014, but will begin shipping to OEMs in Q4.
The impending arrival of Bay Trail may serve to explain the slow adoption rate for what should be highly desirable ARM tablet processors from Nvidia and Qualcomm. In Q2, Nvidia began shipping the Tegra 4, after delaying it in favor of the Tegra 4i, which incorporates an LTE modem. Nvidia books all ARM processor revenue under the Tegra processor division, and this was down 71% in Q2 compared to a year ago and down 49% compared to Q1, as disclosed in its latest earnings report. How much of this drop was due to the delayed introduction is difficult to assess, but there just aren't many devices that use the Tegra 4, while only Nvidia's Phoenix reference platforms uses the 4i.
Adoption of the Qualcomm Snapdragon 800 series of ARM processors has been even slower. So far, only "phablets" from Sony (Xperia Z Ultra) and the LG G2 use the 800 series. Even though the 800s have quad cores and about the best benchmark performance of any ARM processor, no tablet makers have adopted them.
Nvidia is the most vulnerable to Intel because Haswell and Bay Trail threaten both of Nvidia's key business areas, graphics processors and ARM processors. Haswell introduced much more capable internal graphics than Intel's previous Core generation. The long-term trend is for the external graphics processors that are a core business for Nvidia to become increasingly unnecessary for mainstream computing. With Haswell, Intel began to tout the gaming performance of its processors, traditionally a safe business haven for Nvidia. With each successive processor generation, the market for external graphics processors in consumer-oriented computing systems shrinks.
So far, Nvidia has been able to compensate for this trend by finding other uses for graphics processors in high performance computing and servers. GPU revenue was up 8% in Q2 compared to last year and 9% sequentially, but I don't expect this to last. Mobile computing has virtually taken over the consumer computing space, and in mobile computing graphics processing has moved onto the SOC.
Bay Trail will keep demand for the Tegra 4 and 4i depressed, so I don't expect much recovery in the Tegra segment. Nividia's total revenue was down 6% year-over-year, and I expect this condition to persist through Q3 and Q4.
For Qualcomm, the immediate impacts of Haswell and Bay Trail are not so dire. Qualcomm SOCs and wireless modems are in most Android phones and in every Windows Phone. Qualcomm will lose out on some tablet business in Q3 and Q4 but will hardly notice it at first, given the burgeoning Android smartphone market, which grew by 74.5% year-over-year, according to IDC. For Q2 Qualcomm posted revenue growth of 35%, operating income growth of 21% and SOC chip shipment growth of 22% compared to a year ago.
Qualcomm's problems start when Merrifield begins shipping to OEMs in Q4. This is likely to negatively impact Qualcomm's SOC unit sales and revenue, but to what extent is difficult to predict. The most likely scenario is that price competition from Merrifield will force Qualcomm to respond in kind. OEMs for the most part will not abandon Qualcomm if it stays price competitive, but will field Merrifield phones in parallel in a "let the market decide" approach. Qualcomm may hold market share, but only with erosion of its operating margin. In Q2 the operating margin of it Qualcomm CDMA Technologies (QCT) division, which produces all of its semiconductor products, was approximately 17%.
Not enough is known about Merrifield to say with certainty that it will outperform Qualcomm's best phone SOCs, but it seems likely, given Merrifield's Bay Trail heritage. Even if Qualcomm manages to fend off Merrifield, perhaps by virtue of superior wireless technology, still more challenges await it. Next year, Intel intends to introduce a yet more advanced 14 nm process for its Core and Atom processors, potentially leaving the ARM chip makers 2 full process generations behind Intel.
The ARM Foundries Respond
The response of the ARM processor foundries such as TSMC and Global Foundries has been to initiate crash development programs intended to bring 14-16 nm processes on line by the middle of next year, effectively leapfrogging Intel's current 22 nm process. Whether this will work remains to be seen. If it does, it levels the playing field with Intel, and a vigorous and healthy competition between the Intel and ARM ecosystems ensues. On the other hand, failure by the foundries to bring the new processes online would be absolutely devastating for Nvidia and Qualcomm.
I continue to take a middle view regarding the ARM foundries and the ARM ecosystem. Intel supporters like to point to the superiority of Intel's 22 nm process as proof that Intel will eventually overwhelm the ARM ecosystem, but nothing is as ephemeral as technological superiority. IBM, which has done significant work in the type of transistor, called FinFET, used by Intel, and which holds a number of FinFET patents, is in the process of sharing its knowledge with its Common Platform partners, Samsung and Global Foundries, as they try to ramp up 14-16 nm FinFET production lines. I expect that they, together with TSMC, will get their 14-16 nm production lines going by the middle of 2014.
It's not a matter of if, but when the ARM foundries will begin 14-16 nm FinFET production, but timing is critical. Most likely, they'll get into volume production by the middle of 2014, but by then Intel will have made significant inroads in tablets and smartphones. For this reason, I'm long Intel but consider Nvidia and Qualcomm to be short opportunities starting early 2014.
Disclosure: I am long INTC. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.