300 mm Fabs Need to be More Competitive

Long-term capital spending patterns have changed. Yield ramps have to be faster. And a number of 300 mm fabs need to become more cost-effective and efficient. These are some of the conclusions that could have been drawn from Steve Newberry's presentation at the Industry Strategy Symposium (ISS) in January. "These days it's about technology and cost and the speed of what you can do," he said.

The president and CEO of Lam Research Corp. (Fremont, Calif.) emphasized the necessity of speed in today's consumer-driven market. "Fabs today need not just fast installs and qualifications, but also the ability to take a technology and extend it two or three generations with upgrades, as opposed to new chambers, while having the knowledge management techniques needed to produce solutions in one place in the world and replicate them in another place in the world within hours," Newberry said. "We need to understand our core competencies and take advantage of the competencies of our core partners. We also need to think creatively about cost and cycle time."

Though Newberry did not directly talk about fab and process yields, he did comment on the overall status of existing 200 and 300 mm fabs. "Lots of 200 mm logic fabs are cash cows, and they now need 90 and 65 nm solutions on critical application steps so they can continue to have those fabs be significant producers." Device manufacturers working on 200 mm platforms have long complained about the lack of advanced processing capability since, at some point, development switched over to 300 mm platforms exclusively. Perhaps this situation is changing.

Many of the first 300 mm fabs were built following the 2001–2002 dotcom crash. Because of consolidation in the marketplace and the growth of foundry business, logic IDMs started to rethink their investment strategies, and several transitioned to mixed-model hybrid, fab light operations. "In 1999, the pure-play foundries made up 25% of logic investment; by 2002, it was 41%," Newberry said. He claimed that although many companies made the early, difficult transition to 300 mm wafers, only about four logic IDMs were large enough to afford their own 300 mm fabs and compete with foundries. "The reality is, the affordability for logic companies at 300 mm is almost nonexistent," he said. During this period, the foundries benefited greatly from the expansion.

There are several indicators in the market today that industry cycles have changed for the better. "This is not your father's Oldsmobile," Newberry said. "People ask whether history will repeat itself, and I believe the present suggests things are already different." He points to a substantial change in capital investment by chip manufacturers, which has changed, on average, from 25% to 20%. This varies by product, but, long-term, memory makers have reduced their capital spending as a percentage of revenues from 40% to 37%; logic from 20% to 15%; foundry from 28% to 23%. The industry is now in its fifth straight year of revenue and unit growth; the current cycle was interrupted only once, recaps Newberry, when foundries overspent in 2004. Overall, in the past five years, we have seen very rapid adjustments of overspending, leading to less volatility in the market. And with industry consolidation, the top 10 semiconductor producers, together with their alliance partners, now account for 76% of capital spending.

"From a customer order to fab ramp, we are 50% faster today than we were in 2000," Newberry said (Figure ). "Strategies to enable current production plus three generations need to be developed at the customer site with elements of centralized R&D and also decentralized R&D. Equipment companies will continue to grapple with the reduced capital intensity, but with increased demands on technology and cost effectiveness."

Fab ramp rate has accelerated dramatically in recent years.

Like the consolidation among device manufacturers, semiconductor equipment companies have also undergone consolidation. The top five suppliers now account for 48% of the market share. Operating profits have declined overall, but especially for the smaller companies. "There is strong demand for best-in-class suppliers, with speed as the additional critical element. It's about technology infrastructure, speed and doing everything as equally as well or better at better cost points," Newberry said. He said the industry is expected to spend $150B in the next four years — 54% on memory, 21% on foundry, 10% on microprocessors and 15% on other logic.

"The gigafab is here, and if you don't have strong collaboration with your partners, the decisions made will be more impactful than before, " Newberry said.

Find more information on yield management.