Peak Mobile has passed for the majority of the ARM (ARMH) processor vendors as it has for Intel (INTC). Looking back, it was the CES Show in January 2011 that marked the Apex for companies such as Nvidia (NVDA), Marvell (MRVL), TI (TXN), MediaTek and others who were making a run on the burgeoning smartphone and tablet markets. Only Qualcomm (QCOM) can claim to have risen in value and statue with its leading edge baseband technology, while Intel continues to play a two handed game of high priced PC processors with more economical Atom chips. However, the PC and Fabless business model template that worked for years has been broken by the internal chip developments at Apple (AAPL) and Samsung (GM:SSNLF) that are combined with their vertical supply chain. What is in store the next 6-12 months is probably a series of falling dominoes that leads to unexpected outcomes. This is the first of two articles to look at today's landscape and what is likely to come.
Jen-Hsun Huang, the CEO of Nvidia was perhaps the most visible spokesperson for the vision of a mobile world driven not by the x86 performance centricity of Intel but by a power sipping ARM processor with an extreme emphasis on graphics capability. Improved visual capabilities would, after all, be more attractive to end-users and in this market be able to receive at least equal billing, especially given that it was consuming a greater portion of the overall die.
Jen-Hsun has always been in a hurry. First to beat out over 30 graphics vendors from the mid 1990s for the top spot in the PC market and then to stay a step ahead as Intel slowly integrated graphics into its platform. At first Intel combined graphics with the chipset in older process technologies in order to win the low end of the market. More recently they have moved it onto the processor die itself at the front end of the process technology curve and thus close the gap with Nvidia. Until recently, though, standalone graphics chips captured roughly 35% of the market. This would have remained true for the future if not for Intel's push into the much smaller ultrabook platforms where space, power and thermals dictate that a combined solution is preferred over raw performance.
In 2011, Nvidia's ARM based Tegra was considered to be at the head of the crowd. Jen-Hsun, in fact predicted that it could be a profitable $1B business within several years. It was a logical conclusion and based on selling 40-50M chips a year, which is nearly equivalent to what they were building for the PC market. And yet the market TAM prospects for mobile in the years ahead was estimated to be at least an order of magnitude greater than PCs. Tegra revenue peaked at $243.9M in Q3 2013 and registered in at $764M for all of FY2013.
The success from a year ago, though, has receded as Tegra sales collapsed to $52.6M this past quarter from nearly $180M a year ago. Blame it on a new crop of low cost Chinese ARM competitors and the ability of Qualcomm to leverage its 4G LTE baseband business across Apple, Samsung, Nokia and other high end mobiles. Now an existential dilemma has reared its head for Nvidia that is similar for the rest of the mobile players. Is there a way forward for the merchant vendors given several key market dynamics?
Apple's Mac Air, followed by the iPad exposed the myth that graphics had to be better than "good enough." The driving force, after all in these platforms was a thin aesthetic design that emphasized all day battery life and a low thermal, no fan infrastructure. Intel and Nvidia with its PC graphics realized too late that the new form factors had to be built with scaled down processors with graphics on board in order to not exceed the 7W max thermal power dictated by the mechanicals. Intel's Haswell processor has enabled them to close the power and thermal gap (the Haswell-ULX is 10W) but its die size at roughly 180mm is at least 50% larger than where it needs to be in order to hit the costs that would allow the MAC Air and the rest of the ultrabook market to be more competitive with tablets.
The first iterations of the Mac Air were built with a separate graphics chip and a Hard Disk Drive. Over time Apple slowly converted to SSD drives and the graphics chip was dropped as they utilized Intel's Ivy Bridge core with graphics on board. In addition, Apple tended to pick an older, slower version of an Intel CPU to gain a favorable price that allowed them to target a $999 price point. The latest generation of Mac Air features a Haswell processor with its advanced power management that Apple has skillfully controlled with its OS X to get nearly twice the battery life of similar Windows based ultrabooks (12 vs. 6.5 for a Sony Ultrabook).
The iPad, from a mechanical point of view, can be considered the smaller brother of the Mac Air. Instead of a 7W thermal threshold, the limit is around 2-3Watts. A lighter, single tasking O/S and a smaller screen has allowed Apple to get away with a processor and graphics solution that is lower in performance, yet responsive and lower power. Economically, the processors were cheap as they were built on older process technologies.
The summarization of market trends above is necessary in order to understand that Apple was running two platforms in parallel; an x86 one with MAC OS compatibility trending downward in power and performance and upward in battery life, and an ARM one trending upward in performance while maintaining all day battery life. The addition of the 64 bit A7 to the iPad 5 then is another step towards merging these platforms capability wise and Apple will let the market decide how much of each is purchased.
The launch of the iPad 5 with its 64 bit-A7 processor opens up a new front in the mobile processor wars that I will cover in another article as it sets up some interesting scenarios for Intel and the rest of the mobile merchant chip market in a world that is swimming in 32 bit processor silicon. One or two may survive if they are bold.