By Phil LoPiccolo, Editor-in-Chief
Just when the debate over transitioning to a new, 450mm wafer size was thought to be put on ice for the foreseeable future, pronouncements at the recent SEMI Industry Strategy Symposium (ISS) in Half Moon Bay, CA, indicate that the controversy is as contentious as ever.
“People have very different perspectives on this issue, and virtually no one in the industry will acknowledge anybody else’s views,” explained VLSI Research CEO Dan Hutcheson, in his presentation. “We’re about as divisive as a theological unification meeting.”
One year earlier, at the 2006 ISS meeting, Hutcheson asserted that the International Technology Roadmap for Semiconductors’ (ITRS) timetable for a shift from 300mm to 450mm wafer production by the year 2012 would be virtually impossible to meet. Not only would the development time required for the wafer-size change push implementation out to somewhere between 2020 and 2025 — to build a new lithography platform alone typically takes up to 10 years, he argued — but equipment suppliers would also be unwilling to bear the staggering development costs, especially after getting burned by footing the bill for the previous transition from 200mm to 300mm wafers and seeing little return on their investments.
Today, Hutcheson is surprised that the 450mm wafer-size transition is still a front-of-mind issue. “I thought that this topic had died, but it hasn’t, and it’s costing the industry a lot of money,” he said, adding that “it’s all stalled in committee” amid debate among four opposing groups.
The first group is comprised of those who interpret the Roadmap’s 2012 transition as a commandment, and are “willing to fit any data to it, even very complex models, like the industry economic model, which nobody understands,” Hutcheson said. He noted there were similar zealots whose careers were ruined by staunchly trying to adhere to a 1997 deadline for the 300mm transition, and advised the 2012 camp “that you just can’t make it happen. You have to look at real data.”
Next are the 300 Primetime people, represented mostly by equipment manufacturers. “They’re pretty reasonable,” said Hutcheson, but are also “molding data to fit their own conclusions, and most of it is historically inaccurate and doesn’t make sense.” He accused this group of “trying to create a fog around 300 Prime, so that 450 blurs into the background, and everybody forgets about it — but of course that never happens.”
At the far end of the spectrum are the 450mm agnostics, who simply believe the 450mm transition will never happen. “They tend to be loners, and I don’t find them very often. But when I do, they’re usually hanging out with the 300 Primetime group,” Hutcheson noted.
Last are the Moore’s Law aficionados — with whom Hutcheson aligns himself — that don’t really care what happens, as long as the industry continues to keep driving down the cost/transistor. “That’s the miracle of this industry over the last 40-50 years,” he claimed. “This is what we’ve brought the world, and the wealth effect has been phenomenal, so we’ve got to fight for that.”
Bottom line: who’ll pay the bills?
Probably the only consensus today is that serious cost issues are likely to be encountered in a transition to 450mm wafers. Hutcheson showed how equipment industry spending to develop new wafer size platforms exploded during the last wafer size change versus previous transitions — from tool costs roughly doubling during each switch from 125mm-150mm and 150mm-200mm, to leaping more than tenfold from 200mm to 300mm wafers, at a cost of nearly $1.5 billion (see chart, above). If this level of spending were extrapolated linearly, the cost of moving to 450mm would be about $100 billion, Hutcheson said, far exceeding industry resources — and more alarmingly, at the same rate, equipment development spending would reach nearly $1 trillion.
Comparing relative areal cost of capacity over the past three wafer-size changes shows a similar surge for 300mm wafers. While equipment development costs rose incrementally from about $16/cm2 for 150mm wafers to about $17/cm2 for 200mm wafers, they doubled to $34/cm2 for 300mm wafers. Meanwhile, fab costs rose modestly from about $7/cm2 for 150mm wafers to $12/cm2 for 200mm wafers, but dropped precipitously to $9/cm2 for 300mm wafers (see chart below).
“What this says is that the semiconductor makers got a huge deal,” Hutcheson asserted, pointing out that all being equal, today’s $3 billion fabs should in fact cost as much as $9 billion. That’s obviously not economically feasible, he acknowledged, but pointed out that “it was not economical to spend so much on R&D for 300mm,” especially since a whole generation of tools became unusable because suppliers were required to prove they worked on prior-generation technology, Hutcheson said. “The result was that we built in a very expensive development cycle.”
If the industry learns from the experience, it could actually make 450mm a pretty economic cycle where everybody wins, Hutcheson said. But the first place to start is dismissing 2012 as unrealistic, and ceasing to worry about any timeline, he concluded, and instead focus on continuing to decrease cost/transistor by the use of new tools, technologies, wafer sizes — or any means possible. — P.L.