Wafer expansion hits the buffers

By Chris Edwards |  No Comments  |  Posted: January 18, 2017
Topics/Categories: Blog - EDA, IP  |  Tags: , , , ,  | Organizations: ,

Water flows downhill. Moore’s Law continues. Wafers get bigger. The rules of the world seem simple enough. Except that the last one has gone into reverse. A resurgence in 200mm manufacturing has caused a short-term shortage in raw wafers: one that may translate into a longer-term problem for the industry. And the move to supposedly more efficient 450mm wafers is dead in the water, with only an outside prospect of flash memory saving them.

In October, SEMI reported a gradual increase in 200mm production that is seeing the wafer size introduced more than 20 years ago coming back into fashion. According to SEMI, production of 200mm wafers peaked in 2007 just before the financial crisis. Fab owners slashed their output, seeing wafers per month drop from more than 5.6 million to 4.6 million in 2009. With 300mm production being more efficient on a per-square inch basis, it seemed the days of 200mm production were passing much as they did for 150mm and smaller wafers before them. Yes, there are plenty of wafers being run in older fabs on those sizes but they are for speciality products where the economies of scale do not justify a move up to a larger wafer size. And many are on non-silicon technologies that do not offer the wafer makers a viable payback.

The return of 200mm production

SEMI expects production based on 200mm wafers to reach 5.5 million per month by 2020 on the back of IoT, power, analogue and other designs that are not dominated by digital logic. Foundries are now responsible for much of today’s production. The largest single supplier of silicon from 300mm is Samsung, according to a December IC Insights report. Including joint ventures, TSMC is now the leading supplier of completed 200mm wafers, accounting for 11 per cent of the global total. Texas Instruments is second place with 7 per cent. Foundries UMC, SMIC and Grace are also top-ten suppliers with 6 per cent, 4 per cent, 3 per cent, respectively. Production of 150mm wafers remains dominated by IDMs, although TSMC has 3 per cent of that business now.

The resurgence in 200mm production is helping to drive an unexpected shortage. Supplies of raw 300mm wafers are also expected to be thin. But Malcolm Penn, president of Future Horizons, said at the analyst’s six-monthly update in London the increase in price for 200mm wafers was less expected. For suppliers of the wafers, it’s been a long time coming.

“They’ve put prices up in the weakest quarter of the year. And the price increases have stuck. Why? Because they’ve maxed out. No-one has made an investment in 200mm wafer capacity,” Penn said. Having been forced to trim margins over many years the few suppliers left have not spent on equipment to expand supply. “We have a serious wafer-supply shortage in the first quarter of the year. What’s going to happen in the third quarter?” Penn asked, indicating the industry’s strongest period for annual sales. “The problem is only going to get worse. You have to put in capacity and nobody is. They are not mindful to do it because of the current level of pricing.”

450mm recedes

For future large-scale expansions of capacity, a shift to 450mm at the other end of the business now looks remote. “It’s over. Every single European program on 450mm has been cancelled. The real problem is not so much the existing projects being cut. It’s that, if you had to resource it for another attempt, you would have to start again because the R&D is tied to the process nodes that were anticipated when the 450mm work began,” Penn said.

Multiple patterning is one of the reasons 450mm floundered, Penn said. “You couldn’t get payback on it.”

Flash memory may yet provide an avenue for a limited 450mm migration. With the shift to 3D/vertical bulk-memory devices, the emphasis is very much on wafer-scale etching and deposition steps rather than per-reticle lithography – the factor that killed 450mm for logic devices.

“The investment in 450mm hasn’t been wasted,” Penn argued. “Spinoffs have gone into 300mm and made those 300mm factories much more efficient than they would have been under the natural roadmap,” Penn said, pointing to another incidence where continual price pressure removed the incentive to invest in a new generation of production. “It looks like the equipment industry got their way on this: they hated it.”

Comments are closed.

PLATINUM SPONSORS

Synopsys Cadence Design Systems Siemens EDA
View All Sponsors