THE ANNUAL SEMICON WEST conference and exhibition has opened in a bullish mood with analysts and industry figures predicting record growth rates for the year but fears remain about a fragile world economy and longer term technology issues.
Stanley Myers, president of industry association SEMI, reported that semiconductor sales went above the pre-recession high of July 2008 for the first time in May. A basket of eight analyst predictions for the year’s sales rose from 14.9 per cent at the start of the year to 28 per cent now, and FAB spending rose 117 per cent in 2009 to support growing demand.
“This year will be a record year for semiconductor sales and unit shipments, something we’ve never seen before,” said Myers. “The industry is clearly back on track.”
The PC and mobile phone markets are expected to drive demand, and Gartner has predicted growth in these areas of 22 and 14 per cent respectively this year.
Samsung is also ramping up its flat-screen production significantly, spurring further demand for chips.
Stable prices are also a factor, and the industry’s FABs are now working at 93 per cent capacity.
“Talk about falling off a cliff. We really did last year from the peak of July 2008,” said Myers.
“We hit bottom by April 2009. The slope of the drop-off only hints at the extreme measures the industry was taking at that time to preserve capital and stay alive, to keep shipping and keep staff on the payroll.”
However, Myers warned that the future growth of the industry is not assured. The economy could falter in the short term.
And the chip industry is already forgetting the lessons from the past slump in demand, according to executives at Semicon. While demand is strong several senior managers expressed concerns at an executive roundtable that the industry had not learned from past experience.
“My fear is that we’re going to unlearn the lessons of the downturn,” said Rich Wallace, chief executive of KLA-Tencor.
“The challenge needs to be how to meet demands without overheating the market. I suspect the industry has got better at working leaner, but old habits creep in.”
Bernard Meyerson, IBM’s VP of innovation, likened the situation to an engineer slamming his hand in a door and doing it again with the other hand to gather additional data.
Chip manufacturers are highly sensitive to demand, but the industry’s tendency to oversupply in pursuit of market share could cause long-term problems.
Stephen Newberry, president of Lam Research, was also somewhat pessimistic about the industry’s maturity.
“Everything is different, and everything will stay the same,” he said, pointing out that there were 16 major DRAM manufacturers in 2000 but less than half that number now, and that access to capital for startups is much more difficult.
“The good thing for the industry was that less players could search for demand, but as we go forward it’s only a matter of time before we oversupply demand,” he said.
There is hope, however. Memory manufacturers got together and cut supply in 2007 to prevent the market being oversupplied, but were caught out by the world recession.
“[Manufacturers] scaled down investment to get a soft landing in 2008, but the economy fell off a cliff and demand fell beyond all expectations,” said Newberry.
Keith Barnes, president of Verigy, commented that the industry still had not learned the lessons from the 1980s downturn, that is, that when the economy dives IT planning has to take this into account.
“A lot of companies need to look at the macroeconomics. If there’s another Asian crisis we have to be able to respond,” he said. “In 2008 things were starting to turn, but we were in denial. That period in 2009 when it fell off the cliff came as a surprise to some people in the industry.”
Cliffs or no cliffs the macroeconomics are not stopping industry spending 40-45nm scale manufacturing or its movement to develop smaller transistor sizes. Thomas Sonderman, vice president of manufacturing systems and technology at Global Foundries, suggested that the next “crux point” for the industry will be at the 28nm level.
“The battleground is going to be at 28nm with our competition,” he said. “We are focusing on low-power handheld devices there.”
But while some think that any investment is dangerous with a fragile world economy Myers is very concerned at the lack of R&D and test spending for the longer term. The industry traditionally invests 13 to 15 per cent in this area, but this is now in single figures. “The semiconductor industry needs funding for innovation,” he said. “I’m concerned that we are not investing enough to build new innovative processes.”
That investment is needed because silicon technology is fast reaching the limits of atomic engineering, and the industry is going to need a revolution in design in the coming years, delegates at Semicon West 2010 have been told.
Bernard Meyerson, IBM’s vice president of innovation, said during his keynote address that silicon is being engineered on such a small scale that a whole new way of building chips is needed.
For example, scientists have already been able to build 2.6nm carbon nanotubes that function as a transistor, but there is no way to manufacture them in the kind of bulk needed to be useful to the wider industry.
“Silicon doesn’t scale to these dimensions,” he said. “There’s a challenge. The industry is up to it, but at what cost? That’s the question.”
The problems are too big and too expensive for any one company to handle, according to Meyerson.
The average return on of R&D spending in the 1950s was in the region of , but this has fallen to now because chips have become more complex.
Companies need therefore to collaborate or consolidate. There will be a lot of mergers in the years ahead, Meyerson said, but IBM is working on collaborative efforts between companies.
FinFET gating technology is an example of how this could work. AMD, IBM and Motorola had collaborated with academics to develop the double-gate transistor technology in a way that would have been too big a financial risk for one player.
Another area that needs addressing is chip proximity. Costs need to come down as optical networking grows, but chips have to be physically integrated to reduce latency. “I never thought I’d say this, but the speed of light is woefully slow,” Meyerson told delegates.
Another challenge is data analysis. Information is now being analysed in real time, and it is the companies that can build these kind of systems that will see the strongest growth in the future.
Is this therefore the future, not in ever diminishing scales of nanometric manufacture but in the creation of good computing systems that can handle the vast data crunching needs of the world economy? Data crunching services that don’t need the capital investment hikes new foundries need could perhaps adapt far more easily to a weak and volatile world market. Only time will tell. µ
Corsair’s Performance Pro 128GB and 256GB drives are the latest from the firm that has been bringing its enthusiast oriented brand to the SSD market. Rather than going with Sandforce controllers, Corsair decided to slot in Marvell units, claiming a sequential read data transfer rate of 515MB/s and a write rate of 440MB/s with 60,000 and 65,000 IOPS for the 128GB and 256GB units, respectively.