Despite Analyst Agitation, An AMD Atom Is No Answer to Company's Woes
A number of these criticisms are well-founded, but dinging the company for its supposed failure in the netbook market indicates that the analysts in question have a weak grasp on both CPU development and AMD's current technological position vis-à-vis Intel. In a perfect world, yes, AMD would have a netbook-class processor, be operating in the black, and Jerry Sanders would award each and every employee with their very own llama. Since the world isn't perfect, AMD has to make do with what it has and netbooks aren't really in the picture.
Intel's Atom is a small, in-order processor designed explicitly for minimal power draw. To that end, Intel sacrificed a number of features that would have increased the chip's performance at the cost of additional power usage. Atom is built on a 45nm process (something AMD is still transitioning to), and in real terms, these chips aren't all that expensive. Intel announced that it had reached gross margin percentage parity between Atom and its other products several quarters back, but matching the percentage doesn't equate to receiving the same number of real dollars per processor sold.
The point to all this is that Intel didn't just slash the cache off a Core 2 Duo, drop the clockspeed, remove complexity, and call it a new chip. Intel didn't invent the idea of a modern, low-cost, low-power x86 processor—VIA's CPU division, Centaur, (the other, other, x86 CPU manufacturer) has been building them for about 10 years—but Atom hit the market at the right place and time to turn into a sensation that not even Intel saw coming. Atom production in the latter half of 2008 was constrained while the CPU giant brought additional testing equipment online, and while Atom accounted for some $353 million in revenue for the company in the second quarter of this year, the diminutive chip can't support Intel's existing business cycle. Atom may not rise to the level of a disruptive technology, but its increasing prominence in the past 12 months is definitely changing the way OEMs do business.
At present, AMD contracts with (and is the sole customer of) GlobalFoundries to produce its processors at one of two foundries. The first fab, F1M1, is a 4.5 year-old facility that initially debuted at the 65nm node using 300mm wafers, but has since begun transitioning to 45nm. The second fab, F1M2, née Fab 38, née Fab 30) is currently producing 45nm processors using 300mm wafers; GlobalFoundries has previously stated that this facility is also working on deploying 32nm bulk silicon. Comments made during AMD's second-quarter conference call indicate that the company is still selling far more 65nm products than 45nm; AMD doesn't expect to see 45nm parts account for 60 percent or more of total shipments until the fourth quarter of this year.
Why A Netbook Makes No Sense for AMD
AMD's mobile roadmap through 2010.
Let's take a moment and consider AMD. It's been at least three years since it posted a profitable quarter, the corporation has cut jobs left and right, and has drawn its break-even point down to just $1.3 billion in revenue, down from $1.5 billion before the economic crisis tore things up. Equipment sales, the well-regarded Phenom II launch last January, and Strong performance over the last year from its ATI division have collectively helped stem the flow of red ink, but never halted it. AMD's 45nm execution has thus far been excellent—the company introduced its own hexa-core Istanbul processor on June 1, five months ahead of schedule—but Sunnyvale is walking a financial tightrope in which every decision and product introduction must be weighed against that $1.3 billion break-even point CEO Dirk Meyer has pledged to reach by the end of 2009.
Given the reality of its present position, AMD has an obligation to its shareholders to compete only in those markets where it can maximize return while minimizing cost. AMD has never done particularly well in mobile; there's no reason to expect that the arrival of a netbook product would change that. With R&D financially constrained and the company's expected capital expenditures for 2009 cut to $100 million (down from $150 million), AMD doesn't have the luxury of sitting down and designing a true Atom competitor. The best the company could hope to do in the short-term would be to build a 45nm Athlon II at minimal clockspeeds and very little cache. Such a chip might find traction in the highest-end netbooks, but it would be a kludgy Atom competitor at best.
An AMD-designed Atom-like processor would be little more than a speedbump for Intel's steamroller at this point in time, and AMD knows it. The company's solution to Atom, at least thus far, is to attempt to create a market at a price point and performance level in the space between Atom and lower-end Intel mobile Celeron parts. This approach is exemplified in the HP dv2, though CNET blogger Brooke Crothers notes that HP's new baby may not be ramping well. Nevertheless, AMD has blocked out its own price/performance segment precisely because it has no answer to Atom in the near future.
AMD's lack of a netbook processor may be a theoretical weakness, but the company's decision to focus on desktop and server markets where it's historically been strongest is the right decision. The netbook market may be burning hot right now, but Sunnyvale's gross margins are already almost 50 percent lower than Intel's and its financial position, as previously noted, is precarious. AMD is obviously hoping to strengthen its mobile position with its upcoming "Tigris" platform refresh, but it'll be desktop and server markets that make or break Dirk Meyer's profitability promise. Netbooks, however tempting, will have to wait.