Apple and The Road Ahead to Building an x86 Processor

A small blurb last week announced that Apple had hired Jim Mergard away from Samsung after just 15 months on the job. Previously to that he was a 16-year AMD veteran who headed up their low power x86 Brazos processor team. In near synchronicity, AMD hired Famed Apple Designer Jim Keller to be its chief microprocessor architect. When all is said and done we might be witnessing the biggest re-alignment in the processor industry as Apple embarks on an audacious path to building a family of x86 processors to power their Mac PCs and iCloud servers thereby relieving them of Intel’s high prices and putting further daylight between themselves and competitors. Both individuals may end up working together in driving the vision.

The steady stream of talent hemorrhaging from AMD’s Austin design facility since the summer of 2011 has been readily scooped up by the likes of Samsung, ARM, Apple and Altera. It is, as one friend remarked to me an incredibly deep pool of talent. And yet here, as we witness the demise of Intel’s greatest competitor, the door is opening up to a new threat whose only goal is to serve its own vast internal needs. If Apple were to form a team of several hundred engineers or perhaps collaborate with the remnants of AMD serving as the design house and legal Intel cross-licensed manufacturer, there could be a new high volume product and business model that advances Apple’s goals. Specifically, Apple is looking four years out and sees a $10B payment to Intel for processors that could be reduced dramatically. After spending $500M to buy PA Semi and develop the custom A6, would it be out of the question to drop another several hundred million in order to save billions a year?

For the past several years, Apple has been hammering Intel about the high thermal power (TDP) of the mobile x86 processors. Intel has finally responded, but only marginally. The current Ivy Bridge measures in at 17W, which is 4 times higher than what is ideal for the MAC Air formfactor. Next year, Haswell will arrive with what looks to be about a 9-10W TDP. Again, this is too high for what Apple wants, which leads to CPU throttling and performance degradation.

Moreover, what sticks in Apples craw is that the Intel ULV mobile chips sell for over $220 vs sub $50 for standard chips. If Apple could have their way and had such a chip today, they would probably build something that performs between an A6 and Intel’s Ivy Bridge but consume less than 5W TDP max. With this lower TDP chip, Apple could push the envelope on their enclosures like they have on the iPhone 5. Ultimately this design philosophy drives premium prices with no competitors in sight. Get that, the driver of premium prices today is form factor compared to processor MHz in the 1990s. This is the result of Moore’s Law over delivering for the past decade despite Intel’s attempts to price based on performance.

Intel operates on a plan that purposely doesn’t want Apple to get ahead of the pack. They rely on an even playing field and to be able to prep the market a year or more in advance with processor roadmaps shared with everyone equally so all mechanical designs are in flight simultaneously. Thus the Ultrabook campaign.

In the game that is to come, the shake out in the PC industry will result in the major software and ecommerce players owning the hardware or should we say handing out hardware like it was an AOL disk floating in mail boxes from the 1990s. Silicon, in the end wants to be free and mobile. Microsoft has already tipped their hand to a business model where Windows and Office will be preloaded on tablets selling into corporations, even at the expense of its customers. Google’s acquisition of Motorola is a precursor to a similar model, if they believe that vendors like Lenovo and Samsung are caught in the snares of Apple’s legal team.

But what will Apple do about the treadmill?

Intel has been notorious over the years of being able to minimize competition from AMD, Cyrix and others by turning the treadmill several times a year. They continue to do so with the Tick-Tock model. However Apple has shown that in mobiles it likes to run at a pace slower than the 2 year cadence of Moore’s Law so that the digital pieces (i.e. Processors, NAND, DRAM etc) eventually follow an asymptotic price curve towards $0. Treadmill slowing is death to Intel in the mobile space but perhaps an accelerator on the Cloud side as more devices lead to more servers and storage.

If we assume Intel will continue to ship x86 processor into legacy PCs and the high performance, high priced x86 treadmill continues in the cloud, then it makes sense that an Apple x86 effort would push Intel into the arms of someone like Google or Microsoft because they, in the end, would need a low cost supply of cloud and client processors and Intel could deliver based on their process technology lead.

Apple’s market share gains in the PC space have continued despite the fact that product refreshes have slowed and a strict price umbrella has been imposed at $999 and above. The calls by many ARM enthusiasts to convert the Mac Air’s processor has gone unheeded for good reasons: the maintenance of strong brand separation and the allowance of users to naturally migrate to the iPAD. The coming year will be decisive in understanding the true size of the PC market, especially within corporations. Either way, the Vertical compute model is seeking the establishment of a single company to drive both the x86 processor platform and software stack that is more economical than what Wintel + PC OEMs can deliver.

Side note: AMDs current market cap is now less than $2B and it should be interesting to see if a suitor shows up at their door soon in order to acquire the rights to build an x86.

Full Disclosure: I am Long AAPL, INTC, ALTR, QCOM