No matter whether Nintendo Co., Microsoft Corp. or Sony Corp. wins the video game console war, there already is one huge victor: IBM Corp., which designed and makes the microprocessors for all three units.
Not very long ago, some analysts were suggesting that IBM sell its microelectronics division, which produces processors for IBM servers and a wide range of other devices, including automotive and telecommunications gear.
As recently as 2004, IBM was struggling to attract enough outside chip-making work to cover the enormous cost of manufacturing. Sales of chips to other companies remained flat that year and chip profitability was marginal at best. (The unit that includes chips lost $252 million in 2003; IBM stopped releasing detailed figures after that).
Now, because of the video game deals, IBM is expected to see about $3.7 billion in sales of chips and associated design services this year, up from $2.9 billion last year and $2.5 billion in 2004. Analysts estimate the unit is profitable.
But even those gains don't capture how much game chips have galvanized IBM.
Using the engineering consulting work it did for Microsoft, Nintendo and Sony as a model, IBM has formed a new "technology collaboration solutions" unit that's expected to post $4 billion in revenue this year. Internal projections call for that division to hit $10 billion by 2010 and $20 billion by 2015.
Those targets may sound high for a $91 billion company that is barely able to grow overall revenue. But hardware division chief William Zeitler hopes to achieve them by replicating IBM's video game collaborations in such industries as telecom, defense and medicine.
"I think of this at the same stage of development that outsourcing was in the early '90s," he said. "There were a lot of people skeptical and now it represents a large part of IBM's portfolio. We had exactly the right proposition at exactly the right time."
IBM executives credit their deals for the three game consoles to key decisions that date to the 1990s.
One was a plan to aggressively redesign IBM chips.
While previous generations of chips were always tuned to generate faster and faster processing cycles (known as clock speed, measured in megahertz and now gigahertz), IBM pushed a "multicore" design that stresses the efficiency of multiple factors, such as power consumption or graphics rendering. The setup also enables more flexibility to customize chips for particular customers' needs.
The customized "Power" chips that IBM made for the Microsoft and Nintendo consoles, as well as the "Cell" chip that IBM, Sony and Toshiba Corp. jointly created for Sony's PlayStation, all use multiple cores. Cell has nine such processing engines.
"Ten years ago it would have been inconceivable that all the game processors would have used IBM product," Zeitler said. "Nobody would have thought of it because all we had was standard things."
Secondly, while IBM poured money into upgrading its chip-making capabilities — including improvements at its factory in East Fishkill, N.Y., that has gotten $3 billion from company coffers — it also formed chip-investment partnerships with other companies. Sony, for example, put $325 million into Fishkill in 2004.
The partnerships let IBM leverage a bigger pool of research and development money than it could have spent on its own. And the deals helped insulate the company from the expensive volatility that plagues the chip business. Some of the IBM-designed game chips are actually made by Sony, and others are handled by Chartered Semiconductor Manufacturing Ltd., lessening IBM's exposure to swings in production cycles.
The financial benefits from these changes allowed IBM to shrug off last year's decision by Apple Computer Inc. to stop using IBM chips in its laptops and buy processors instead from the world's top manufacturer, Intel Corp.
Even more importantly, executives say, the company's efforts to design chips for the very high demands of video games have given IBM a portfolio of chip technologies it can apply to many other systems, including computers for medical imaging, defense and complex simulations.
"From a technological perspective, we still have a tremendous amount of upside," said Bernie Meyerson, chief technologist for IBM's systems group. "We've only gone down this road one turn."
IBM is already working with outside companies, such as Mercury Computing Systems Inc., to adapt Cell and other chips for their own new devices. Cell now powers a line of IBM servers and is being used in a supercomputer for the Los Alamos nuclear lab. Toshiba plans to use the chip in TVs.
However, even with Cell's performance boosts, it could have limited paths into other systems unless IBM can encourage many software developers to create applications that take advantage of Cell's unusual architecture.
And it's unclear how many customers need the performance that the game consoles demanded, especially when industry-standard chips — which have an enormous pool of software developers ready to write programs — also are delivering at high levels.
Although IBM won bragging rights in winning the Xbox chip business away from the previous supplier, Intel — Microsoft's closest corporate partner — Intel executives argue that focusing on video game chips hasn't given IBM an unmatchable edge.
David Tuhy, general manager of Intel's desktop products division, points out that his company's latest chips — which rely on four cores — support complex "massively multiplayer" games with stunning graphics on personal computers.
"We're not losing any sleep," Tuhy said. "We're very happy where the PC gaming market is."
To be sure, new chips don't benefit only games. Microsoft will take advantage of Intel's quad-core chips to speed sophisticated financial modeling simulations in the Excel spreadsheet program.
But games and the stunning detail of their virtual worlds provide some of the most telling evidence of microprocessor performance. Which is why no matter what happens from here, IBM can claim a significant achievement with the three game consoles.
"Most folks are not aware," said IBM's Meyerson, "that gaming chips drive the frontiers of the industry, not general-purpose computing."