Don’t stick a fork in AMD

With all the vigor and exactness of stock market analysts explaining a one-point shift in the Dow Jones Industrial Average, pundits are penning obits for AMD in the aftermath of Sun Microsystems’ recent decision to buy chips from Intel. Poor AMD: first Core microarchitecture, the looming doom of quad-core Core, and now the defection of its sole first-tier monogamous mate. Talk about your slings and arrows of outrageous fortune.

I invite my colleagues in the media to run their AMD cover stories with Titanic metaphors now and get it out of their systems. As they say in my business, this whole Intel-eating-AMD’s lunch angle doesn’t have legs.

[ Talkback: Should AMD be worried? ]

The Sun flap doesn’t call for much commentary; I don’t recall seeing headlines of gloom for Matsushita when some OEM started integrating Philips DVD burners in some of its models. If you’re willing, deliver these words, very slowly and with sharp enunciation, to the person nearest you who is sticking a fork in AMD: Sun is an AMD customer. Have them say it three times, slowly.

Sun did not dump AMD; AMD will continue to supply Sun Microsystems with enterprise-grade x86 server CPUs. Sun found in Intel a cheap second source of supply for x86 microprocessors. There is a place in the market for a chipmaker that can jump the buzzword bar without actually meeting the requirements of the technology that the buzzword was meant to describe. For example, to Intel, multicore means pasting multiple discrete CPUs onto a chip-shaped carrier, and a server CPU is a desktop CPU with bigger cache. There is a place in the market for a cut-rate player, and it’s not like Intel’s CPUs are junk. As I’ve said, Core is Intel’s best work to date. But Intel has to keep current x86 owners coming back for Core 2.01 and Core 2.1.

So Intel has decided to change the rules of the x86 game. Innovation be damned, it’s stamp ‘em and ship ‘em as fast as you can. In a price/volume war, the company that makes the greatest number of widgets with the lowest manufacturing costs wins, yes?

That’s Intel’s logic, and it leaks like a baby. It relies on the assumption that the market will absorb every part that it pushes out the door. As Intel shifts manufacturing gears to put out models with higher clocks and bigger caches, it assumes that the market will come along. The trouble with this strategy is that the introduction of new CPUs no longer spurs a fresh cycle of buying. The market isn’t in saturation, but it is in the very state that Intel hopes to exploit with technology that’s in technological second-place to AMD: I believe that many x86 buyers are willing to settle for good enough computers. But that includes the ones that buyers already own. They’re saving their pennies and investing instead in consolidation, services and strategic objectives. In other words, I see Intel sitting on a lot of inventory as a choosy market isn’t brought back to “click to buy” by a 200 MHz kick in clock speed.

Without the inventory problem, AMD has more relevant new technology to bring to the table than Intel’s road map portends. And where speed bumps are concerned, I’ll tell you a secret: AMD isn’t worried about speed. It’ll have no trouble keeping pace with Intel’s road map, and doing it without burdening distributors and OEMs with inventories of the prior generation of CPUs.

Enough about economics. Let’s talk quad core next week.