The theory is already stirring debate, much as the now-notorious "end of history" thesis did when advanced by public-policy analyst Francis Fukuyama in 1992 — just before the world entered yet another period of tumultuous change.

"I'm not suggesting that we have consciously fixed the problems that have caused the semiconductor cycle," Jelinek explained. "Rather, circumstances have changed to dampen the cycle without our consciously intending it."

The first such "circumstance" is the growing financial conservatism in the semiconductor industry. "Many of the companies that own fabs are public and are watching their capital expenditures very closely," the analyst said. "You always have the controller watching you."

Jelinek acknowledged that this is less the case in China, where national priorities and an interesting banking system grease the capex wheels for favored companies. "But SMIC and Grace are now talking about public offerings on global markets," he added. "That will enforce on them a degree of fiscal responsibility that companies in their startup phase might not have."

The second circumstance, according to Jelinek, centers on fab equipment. "With earlier processes, the economics favored building a fab shell and completely filling it with equipment as soon as possible," he said. "The building and support were the expensive part, and the equipment was relatively less expensive. Also, it was tremendously disruptive to add equipment to an operating fab.

"But now, the equipment for a 12-inch, 90-nm fab is vastly more expensive than the shell and is quite modular. Companies tend to build the shell and then add only as much equipment as demand justifies, knowing it will be easy to incrementally increase capacity by adding machines later. This permits companies to tie their capacity plans much more closely to demand projections, rather than bet on huge lumps of capacity two years in advance."

The combined circumstances of greater fiscal restraint and the ability to quickly add incremental capacity mean that companies no longer need risk overshooting demand by a large margin; thus the pendulum mechanism that caused the industry to swing from rapid demand growth to massive overcapacity no longer ticks. "There will still be cycles, but not the huge gaps we have seen in the past," Jelinek said.

This is not the first time such a prediction has been offered. As early as February 1997, Alan Greenspan told Congress that information technology would result in a smoothing out of the business cycle. And in one of those quotes that the author probably wishes would disappear, In-Stat analyst Grant Johnson stated that because of incremental capacity expansion and concentration of production in the hands of a few companies, the traditional boom/bust cycles of the semiconductor industry could start to look "more like rolling hills." Unfortunately for Johnson, he made this prognostication at the beginning of 2001.

Sumit Sadana, director of strategy for IBM Technology Group, said he could make a case for Jelinek's scenario, but along a different line of reasoning. "The investment for a new 300-mm fab is huge — $2.5 billion to $3 billion," Sadana said. "This is forcing the concentration of advanced-process capacity in the hands of a few companies. And a market served by a few companies is inherently more stable in pric-ing than when you have many companies all trying to be the biggest at once. This is a muting force, and it's helping to restore some rationality to investments."

Beyond that qualified nod, though, response to Jelinek's proposal was largely skeptical last week. "Have the economics of the industry really changed? No," stated Ed Ross, president of TSMC USA. Even with modular equipment, Ross said, capacity must be added in large chunks if the expansion is to be profitable. "If you build a big shell and only facilitize a small portion, you are going to lose money," he said.

Richard Tobias, vice president of the ASIC and foundry strategic business unit at Toshiba America Electronic Components, was similarly skeptical. "It's true that you can add equipment in modules, but it's not that easy," Tobias said. "There's still a very significant lead time for equipment — you can't just bring in a piece in six months — and equipment lead times help drive the cycles. You have to be judicious, but you still have to order equipment ahead."

Tobias said the current situation is typical of the ambiguity planners face. "We are all trying to fill our fabs right now," he said, "but at the same time we are watching a lot of equipment going on line, particularly in China. And things are getting tighter. I'd say that just about everybody with a leading-edge process will be full within three months."

With demand approaching capacity, the potential exists for another boom in capacity investment. "Unfortunately, there is a natural tendency for companies to invest in capacity when they are seeing growth in profitability," Sadana observed. "That means that in practice, companies end up adding capacity near the peak of a demand cycle. It is very difficult for companies, even though they recognize this, to change and invest when they are losing money, and to restrain themselves when they see shortages and increasing prices."

Indeed, he said, that very temptation is returning, in part because global capital markets are currently awash in liquidity.

Jelinek and Sadana also cited the capacity growth in China. Jelinek predicts that as the Chinese fabs move from local debt funding in the notoriously creative Chinese banking system to outside equity funding, they will also move from a rapid piling up of capacity to more circumspect investment.

But Sadana is not sure just when — or if — that will happen. "When you have a country with a national priority to gain share in a business, that distorts investment decisions," he said. "But I think the impact of Chinese fab investment will be confined to trailing-edge technology nodes. There are significant technical challenges in making leading-edge processes accessible to designers, independent of how much investment money you have."

Greg Frazier, executive vice president of supply chain services at Avnet, invoked an earlier prediction about the semiconductor boom/bust pattern. "In 1995, I think it was, someone said that the industry was no longer cyclic," he quipped. But "it's still impossible to forecast the arrival and impact of new products, and it's still easy to see a ramp and slip into an ordering frenzy."

Even so, changes in supply chain administration since the catastrophe of 2000 may have moderated the swings, Frazier said. "Back then there were supply chains, but there were no checks and balances on forecasts," he said. "The recession has caused us, as an industry, to look at the credibility of, and the liability for, forecast numbers. The kinds of contracts you see today make it advantageous for a purchasing manager to be more realistic all the time.

"There's more collaboration now," Frazier continued. "If we are looking at the cell phone market, for instance, we are not going to order enough parts for each of our customers to build the entire global demand forecast this year. We may order enough parts to meet the demand, but we will delay committing volume to a particular company until they actually need it. Customers are working with us on this as well; they have a lot more sensitivity to their own balance sheets now, and that's an incentive for them to manage their supply chains well."

Fiscal restraint, more concentration of resources in a few companies, better supply chain dynamics — all smooth the semiconductor cycle. But the industrial policy in China is an uncertainty. And human behavior under pressure, while certain, is not easily rationalized. So the question is open.

Like Fukuyama's challenging thesis, Jelinek's proposal contains important insights. But insight is not prophecy.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved