Rocky road to 300 mm?
09/01/2000
A lot is riding on how well the industry executes a shift from 200mm to 300mm (12-in.) wafers. A smooth and rapid transition will mean big payoffs for a number of hardworking players. Many toolmakers, after long waits in the green room, will finally get airtime to compete for megamillion dollar payoffs. Semiconductor fabs will be able to churn out more than twice as much product with the same effort, speeding deliveries and productivity, and they'll be making higher-performance, higher-margin chips to boot. Wafer makers may finally make a little money even after paying off the huge investments needed to build 300mm crystal pullers.
From a loftier perspective, however, there is another factor here that could prove to be a fly in the ointment. Perhaps a dozen fabs could be running 300mm wafers full-tilt by late 2002 if known plans are carried through, according to Eric Ross, an analyst with Thomas Weisel Partners. Even more such fabs could be running if more plans are unveiled. This would add a dozen "superproductivity" fabs to the industry within a short period. The resulting overcapacity, Ross suggested, could be even more severe than the last downturn, triggered partly by rapid circuit feature shrinks that allow more chips/wafer, and thus boost fab output.
Today's rapid build-up has a backdrop of electronic equipment vendors complaining of parts shortages. This is undoubtedly a result of the near stoppage of new fab building for a couple of years at the end of the '90s. Times will remain good for chipmakers and their tool and materials vendors as long as devices go for higher prices and toolmakers sell all the equipment they can make. But an overshoot in the current build-up would mean another price collapse, drying up process tool orders and repeating the cycle that has long plagued the semiconductor business.
At a panel of CEOs from tool and materials companies at Semicon West in July (organized by the Loomis Group), the consensus was that 300mm would be installed and come up to yield much more rapidly than was the case for the transitions from 6- to 8-in. or 4- to 6-in. (see "Will tools sales go South..." p. 44). Bridge tools, which can be brought to yield running 200mm wafers and then upgraded to run 300mm wafers, might help this process. But the CEOs felt that the emergence of so many growth markets, many of them new ones with much higher semiconductor content than the 20% in personal computers, will soak up all the capacity the chipmakers can bring on stream for the foreseeable future. The expansion of telecommunication infrastructure around the world, along with the rise of many Web-enabled portable devices, might keep the build-up going even if things do slow somewhat in the US, some CEOs suggested.
In discussions afterward, some executives disagreed that the 300mm transition would be as smooth as these CEOs suggested. After some four years of experience with 300mm tools, they pointed out, Infineon in Dresden was only now beginning to make 64Mb DRAMs and was struggling to boost yields. Others, like Intel, are planning to go to 300mm wafers at the 0.13mm node, probably with copper and low-k dielectrics. This will be a much tougher challenge than Infineon faced. Jim McKibben, VP of worldwide sales and marketing at Tegal and a savvy industry veteran, said he had recently overlaid the time to yield for the transitions to 6-in. and then 8-in. wafers, and they matched almost perfectly. He believes the transition to 12-in. will take just as long, which pushes out high yields until well after late 2002.
There is another factor that wasn't discussed very much: the health of the economy in the US and in other parts of the world. If there is an economic downturn and Fed Chairman Greenspan is trying to create one in the US (albeit with a "soft landing") chipmakers may slow down their plans to rush into full-production 300mm plants. Instead there may be more split factories, with the main production done on 200mm tools while 300mm pilot lines are brought up to yield. This could slow the transition and help keep the capacity buildup to a more reasonable pace.
The industry has never yet gotten it just right, keeping demand and supply in good balance without an over- or under-shoot. With the growing market complexity, forecasting is even tougher. But the greater variety of end markets might also help smooth the cycles. Let's hope!
Robert Haavind
Editor in Chief