Hey Chipheads, Here Are Some Startup Ideas

I spent my morning in a lecture hall here in Austin, Texas, listening to the heads of manufacturing at three different semiconductor fabrication plants talk about the future of their aging facilities. (Solar isn’t the only thing an older fab can be used for, but I’ll get into that in a later post.) What struck me was how many holes these guys are scrambling to fill as they retrofit their fabs, and how many of them need equipment and help that no one is providing.

The first thing they all requested was more support for older generations of equipment. Fabs are filled with multimillion-dollar tools that can be hard to trash, so finding both spare parts and people to efficiently maintain them is hard work. From a venture perspective, service businesses aren’t the greatest, but maybe someone could bootstrap it?

The other two big pain points might require some serious capital, unless they can take advantage of existing equipment and intellectual property squirreled away in another industry.

Chris Magnella, director of operations at Freescale Semiconductor’s Oak Hill Fab facility, said he’s making MEMs in his fab (among them the accelerometer that powers the guitar in GuitarHero), but lacks the equipment needed to see those chips from all sides. A typical chip is fairly flat, and inspection tools look at an overall shallow pattern etched on the chip. But MEMs are deeper, with embedded 3-D structures that are harder to inspect.

The other idea was some kind of 200mm immersion tool that can etch scan below 65 nanometers, which was suggested by Randy Blair, a VP at Spansion. There are two things to understand here, so take a deep breath:

Cramming more transistors on a single chip leads to smaller lines etched in the chips. That’s called the process node, and right now advanced chips are being made at 45nm (in 300mm fabs).

You also need to know that fabs make chips on wafers that range in diameter from 150mm to 300mm. The bigger the wafer, the more chips you can put on it, creating economies of scale. Most fabs in the U.S. are 200mm fabs that were built during the ’90s.

Without tools that can etch scan at 45 nanometers, these older 200mm fabs can’t follow Moore’s Law and keep packing more transistors on a chip. The key here will be figuring out a way to make such an advanced tool at a cost that’s appealing for older fabs. And if any of these ideas pan out for a startup, I expect scoops from them for the remainder of my journalism career.