Amid AMD's mounting losses market rumors could signal shift in HPC chips

Last week we reported on Intel’s Q4 (reminder: they made almost $2B but their stock tanked; details here and a million other palces). AMD’s news was much more on the “leap out of a window” side of things.

AMD reported recently that they lost $1.722B in Q4, compared to $576M in Q4 a year ago. For the year AMD lost $3.4B, compared to only $166M in 2006. You have to admire a guy that can get that kind of news and then say this

“We were close to break-even operationally for the quarter, reducing our fourth quarter non-GAAP operating loss to $9 million. We improved gross margin by three points sequentially, driven by increased shipments of new products, higher average selling prices and cost containment actions,” said AMD CFO Robert Rivet. “We shipped a record number of microprocessor units in the quarter, including nearly four hundred thousand quad-core processors.”

I want some of what he’s taking. He’s right in the theoretical sense that the chip business did better than the numbers reflect — the loss was driven by a $1.68B write down from its ATI purchase. But that takeover of ATI is a symptom of AMD’s problems: execution, execution, and execution.

Then last week came news of rumors in the markets. AMD shares were up 10% on speculation that it would deepen its relationship with IBM. From the story in Dow Jones MarketWatch

Analyst Krishna Shankar of JMP Securities said in a research note that the stock was “reacting positively to industry speculation regarding a more formal partnership between AMD and IBM, which may involve a strategic equity infusion or even an outright merger between AMD and IBM Microelectronics.””

IBM and AMD have relationships on chip development going all the way back to 2002 when they teamed up on new 65 and 45nm processes. That relationship was renewed in 2004 through at least 2008.

All of which is relevant to HPC in the biggest possible way. Right now HPC machines are dominated by chips from Intel and AMD. There are other players, notably Sun (Niagara) and IBM (with Power and, to a much lesser extent, Cell), but chips from the majors are nearly ubiquitous. I would be concerned about the pace of innovation in the chip market if Intel lost its AMD pressure. On the other hand, if IBM can bring its ability to execute to help AMD get its gear out the door, that could definitely be a good thing for HPC.