It's hard to imagine 450-mm wafers-heck, the sight of 300-mm wafers today can still cause eyeballs to bug-but they're not far down the road.
It's hard to imagine 450-mm wafers-heck, the sight of 300-mm wafers today can still cause eyeballs to bug-but they're not far down the road. Our man in Austin, David Lammers, reported recently that industry experts are looking at 2012 for the 18-inch behemoths to start rolling into fabs around the world. That's not far off.
A decade ago, I flew to Geneva for Semicon Europa and attended a special panel the gear industry was hosting on the coming 300-mm wafers. I had prepped for the event by visiting Sematech in Austin (Was 300 mm the right size? Was 250 mm a better intermediate step, given the mysteries surrounding 300-mm manufacturing at the time?) and Intel in Santa Clara (Did they know how much a boatload of 300 mm would weigh?).
EE Times published a special section examining the challenges of the era, from pulling polished wafers that wouldn't warp to designing lithography and deposition machines that could accommodate the large form factor.
At the Semicon session, everyone not battling jet lag had agreed 300 mm was the way to go. But a funny thing happened on the way to the fab. The usually conflicted semiconductor equipment gang balked 300 mm. And why not? At the time, 200 mm was in its prime and amounted to a highly profitable business. Then along came a couple of downturns in the semiconductor market. The schedule went into the shredder. Only in the past 18 months have 300-mm fabs come online. At a 300-mm Intel fab in Oregon earlier this month, Lammers and I saw the fruits of that effort-overhead conveyors dipping wafer boats into machine bays with the grace of a tuxedoed waiter. Everything hummed along in ultraclean-room splendor.
If there's one thing I'll guarantee, though, it's that 450 mm will not even be close to production-ready in eight years. It's not that the issues are overly technical; it's that industry economics have changed fundamentally from Geneva in 1993.
In the 1990s, a fair amount of IDMs could still afford to help with development costs. Today, you can count them on the fingers of one hand. And the foundries? The cost pressures on that crew now and in the coming years will make them blanch at being at the bleeding edge of 450 mm.
Control over masks and even machines edges closer back into the hands of the remaining IDMs. Reaggregation marches on.