For The Last Time, Tablets Didn't Catch AMD or Intel "flat-footed"
"Like Intel, AMD was caught flat-footed in recent years with the emergence and fast growth of mobile devices..."
This is the sort of sound byte analysis that's often spread by analysts who really ought to know better. The rationale behind it, we'll assume, is the following:
- Smartphones and tablets are really popular
- Intel and AMD don't dominate those markets
- Intel and AMD aren't going to dominate those markets in the next 12 months
- Therefore, Intel and AMD were ambushed / surprised / caught flat-footed or some other insinuation that implies neither company saw this coming.
The False Clarity of Hindsight
Intel began working on the Bonnell microarchitecture -- that's the design that would become Atom -- in 2004. The company realized early on that none of its previous parts were suitable for a handheld device and chose to develop a new part from scratch. It's true that Intel chose to preference what it called Mobile Internet Devices (MIDs) over smartphones, but the engineering challenge of building a small handheld with acceptable battery life is pretty consistent regardless of what you call the resulting product.
It's easy to forget that when Intel launched the first netbook Atoms, in early 2008, Android hadn't shipped on a single phone. It had long been obvious that traditional Windows would never run effectively on a MID, which is why Intel sank development resources into first Moblin, than Meego. Again, it's easy to look back now and say that Android was the "obvious" choice. Easy -- but incorrect.
This slide, from 2009, puts Medfield on the roadmap and again predates the tablet boom.
That's not to say that Intel hasn't made mistakes. The company was unable to ramp Atom manufacturing quickly enough to meet peak netbook demand. In retrospect, it should've shelved MeeGo altogether, moved to Android earlier, and prioritized 32nm Clover Trail development. Even if it had, there's no guarantee that it would've dislodged Apple. A number of capable companies launched a bevy of Android tablets in 2011. Mostly no one cared.
Would a 2011 Intel launch have changed this graph? We think not.
For Intel, moving into tablets and smartphones has always been about the long-term game, not any single year. It needed a Windows-compatible operating system to make a play for tablet buyers who wanted Windows compatibility, which meant waiting for Windows 8. It's true that Android and ARM were better positioned to catch the first wave of tablet designs, and that Apple has enjoyed enormous success with the iPad, but Atom was designed as a mobile / handheld processor from the very beginning.
AMD: With What Money?
Alright, so, fine. Maybe Intel saw the handheld era coming, even if it misjudged the particulars. But surely AMD fits as having been caught off guard and left gaping?
Not necessarily. Here, the situation is more straightforward; AMD never had the funds it would've needed to drive such a shift in policy. Intel started building Atom in 2004, with the design team it took off the canceled Tejas project. AMD has never had spare design teams hanging around to take on such work.
No one who could read a PDF could possibly have missed the fact that Intel was working on an x86 Everywhere strategy by 2008. AMD didn't miss it, either. Rather than trying to compete with Intel head-to-head in an extremely power-sensitive environment, AMD announced that it would target a slightly higher TDP, with the goal of driving better computing experiences.
The result of that effort was Brazos. While it launched after netbook sales peaked, it's the only one of AMD's 2011 launches that managed to hit its performance, power, and manufacturing targets. AMD may have tapped ARM for dense server designs, but its left smartphones to Intel for good reason. The company's chances of driving tablet designs will improve next year, once its next-generation Temash SoC is ready, but again, AMD has to do what it can with extremely limited resources.
A lot of what AMD has done in the past few years could be (very) fairly criticized, but the company wasn't caught off-guard by anything -- except, perhaps, the ridiculous decision to fire its CEO.
Why the framing matters
The biggest problem I have with how this issue is framed is that it impacts a reader's understanding of the current situation. Intel's attempts to move into the tablet business were stymied by the reality of 45nm power consumption, the lack of an x86-compatible OS, and the failure of its own OS initiatives. An aggressive Android schedule and prompt Atom transition to 32nm might have allowed the company to launch Clover Trail 12-15 months earlier -- right into the teeth of the iPad 2.
Putting the blame on Intel and AMD's failure to see tablets coming also obscures the complex, changing relationship between Microsoft, Intel, and the various OEMs. Apple established itself as a tablet vendor because it was willing to take enormous risk and had rich multimedia content to sell. Amazon won accolades by launching a tablet at a rock bottom price and its own content farm.
What we've seen from the traditional PC OEMs, thus far, was a whole lot of nothing -- and then a whole lot of confused SKUs in the wake of Windows 8. Microsoft made it clear that Surface was a product it built itself because it didn't feel like OEMs were willing to anything risky, different, or new with Windows 8. Some of the reports coming back from early tablet reviews make it clear that major manufacturers aren't really willing to stop installing crapware that cripples the user experience. These are the same companies that balk at ultrabook pricing rather than embracing the opportunity to build something besides $500 plastic enclosures with 37 stickers.
Right now, there is no clear winner between Windows Surface and x86 Windows 8. There's no standout, must-have product from any manufacturer; a whole lot of mud is being slung at the proverbial wall to see what sticks. This is absolutely normal. But tablets are more than just an SoC from any vendor -- they're a user experience, a price point, and the value-added capabilities that an OEM brings to the equation. Painting the issue as simply a matter of Intel or AMD being caught off-guard is a drastic oversimplification.