Size doesn't matter

One of the commonly accepted measures of progress in the semiconductor business is wafer diameter, which now is limited to 300mm, or about 12 inches. The push is on to go to a 450mm wafer, which would be about 18 inches in diameter — hence the perception that IBM Burlington's fab is obsolete, because it is stuck with 200mm wafers. Tooling up for bigger wafers would be prohibitively expensive.

When mass-producing chips in the billions for something like a CPU in a laptop, wafer size matters, because the more chips that fit on a single wafer, the more cost-efficient the process is going to be. But when the goal is innovating new approaches to semiconductor design — as it is at IBM Burlington — those efficiencies become irrelevant, Chief Technical Executive Stephen Luce said.

"There's a high-level impression in the semiconductor world that the only way to be relevant is for things to get smaller, equipment to get more expensive, and have new factories," Luce said. "That's actually not the full market.

"There's a portion of the semiconductor market where geometries are more stable, and what they need is a team to innovate new features," he continued. "The advantage of that approach, and that's what we're doing, is you can keep most of the equipment the same, continue to use a factory that's been around a lot of years, and yet stay relevant."

Luce said IBM Burlington changed its approach to the fab in a "very big way," favoring innovation over scaling, starting in about 2005.

"At this point, we are almost entirely out of the businesses that scale, and we're into things that require innovation on these very unique structures and features," Luce said.