AMD's total revenue in Q3 was $1.396 billion, up 18 percent over the second quarter, but down 22 percent year-on-year. That last isn't surprising—Intel's own revenue was down eight percent year-on-year—and AMD hopes the gap will narrow further in Q4 as the market returns to what CEO Dirk Meyer referred to as normal seasonal trends. The company reported a net loss of $128 million, after factoring in a favorable impact of $54 million, which Sunnyvale attributed to its retirement of debt. The company's operating loss in Q3 2009 was $77 million. Both of these numbers are huge improvements over Q2, when AMD's net loss was $330 million, with an operating loss of $249 million.
Now let's switch gears and talk products.
AMD grew all three of its market segments in the third quarter—revenue was up 17 percent overall. Mobile revenue growth substantially outpaced the average at 28 percent, a somewhat surprising number given Intel's historical dominance of this arena. AMD didn't offer detailed color on exactly which mobile chipsets or price points accounted for the upswing, but did indicate that its thin-and-light/netbook competitive product series accounted for less than fifty percent (but more than 10 percent) of the total gain in mobile. In servers, the company's hexa-core Istanbul accounted for about a third of all shipments, and the company's mixture of 45nm and 65nm products is now running at approximately 50/50.
When asked why it took the company as long as it did to ramp 45nm, Meyer indicated that the switchover was delayed significantly by the reality of market conditions in 2009. The precipitous sales drop early in the year left AMD with a substantial number of 65nm processors, and delayed demand for the company's newer 45nm offerings. With GlobalFoundries fabs currently running under capacity, AMD expects no problem ramping 45nm production as consumer demand improves. By the end of the fourth quarter, the majority of the company's processors will be on 45nm, a fact that suggests the company's average cost-per-processor (already markedly lower in Q3 than Q2) will continue to fall through at least Q4. 32nm parts, unfortunately, are still at least a year away. AMD will sample both "Fusion" processors and 32nm standard chips in the first half of 2010, but doesn't expect to ramp production until the tail end of 2010. The company had no specific response to when it might be able to close the manufacturing gap between itself and Intel, but the answer, one assumes, is directly proportional to AMD/GlobalFoundries net income. The company's 32nm parts will continue to use SOI technology, and AMD has no immediate plans to transition its processors to bulk silicon production.
AMD expects CPU revenue to grow nominally in Q4, partly thanks to stronger mobile positioning and the launch of its Congo platform for the thin-and-light segment. Server revenue could also be up, thanks partly to Istanbul's drop-in compatibility with existing two socket and four socket server architectures. As for the desktop/consumer market, Sunnyvale believes it's currently underrepresented in higher-end systems, and intends to address that gap with 45nm Phenom II / Athlon II processors.
The Radeon HD 5870, AMD's current top-end DX11 solution
One interesting fact that emerged from the conference call relates to GPU gross margins. The ideal/target gross margin for these products is between 35-45 percent. AMD's not currently playing in that range, and while the company didn't provide its exact gross margin, it did note that the 5000 series is expected to eventually raise it into the lower end of that range. This isn't expected to happen in Q4, but will be a gradual transition over the course of 2010.
In addition to consumer discrete cards, AMD plans to expand its presence in the lucrative workstation graphics segment. Again, don't look for quick leaps here—it takes time to forge the proper linkages to effectively grow in that space—but its a potential source of revenue Sunnyvale has more than an eye on.
The Bottom Line:
AMD's Q3 results are quite encouraging. The company continues to move towards profitability, and its product mix across both CPUs and GPUs is strong enough that it may very well hit its fourth-quarter profitability goal (barring a major economic upset). With that said, there's nothing new on the horizon that fundamently changes the company's competitive position vis-à-vis Intel. The Core i7 product line continues to whomp on Phenom II whether we're talking about Lynnfield or Bloomfield, and AMD can't effectively counter Nehalem's HyperThreading with a quad-core processor.
AMD will pull hexa-core Istanbul (codenamed Thuban) to the desktop , which will definitely improve the company's position, but there's no magic way for AMD to jump a process generation and magically be head-to-head with Intel's current lineup. To the company's credit, it's not espousing any such strategy, and is choosing to carefully trim costs and improve margins to put itself back in the black. That's not as exciting a story as head-to-head competition in all product segments, but it's a fundamental part of remaining in business, and an absolute necessity for AMD.
It may take another product generation (and a move to 32nm) before AMD can move decisively to enter additional markets, but profitable quarters between now and then are critical. With GlobalFoundries picking up steam, the upcoming 45nm dual-core Congo platform launching, and the Radeon 5000 series in the field, there's still a lot of action to be had.
One other thing. It's become a bit fashionable to bemoan AMD's lack of a true Atom-competitor. We maintain that this is, in fact, untrue. It'd be nice if AMD just happened to have its own Atom-like architecture sitting on the shelf, but attempting to go down that road right now would be a real mistake. Additional details here.