SVG Chairman/CEO Discusses Industry Health, Trends
-- Semiconductor International, 10/1/1999
Papken S. Der Torossian, the Silicon Valley Group's (SVG)
CEO and chairman of the board, spoke recently with Semiconductor International.
Torossian joined SVG as president in 1984, became CEO in 1986 and was appointed
chairman of the board in 1991. A respected industry spokesman, he has served on
the AEA executive and international committees, and as chairman of the SEMI/
SEMATECH board of directors. He is chairman of SEMI's Environmental Health and
Safety Executive Committee. Over the last three decades, Torossian has been
president and CEO of ECS Microsystems Inc., president of Plantronics' Santa Cruz
Division, and spent four years at Spectra-Physics and held a variety of
management positions during his 12 years at Hewlett-Packard Co.
Torossian has a BA in mechanical engineering from MIT and a master's degree from Stanford University.
SI: You've always been philosophical about cycles in our industry. How do you view our latest crash?
Torossian: This has been one of the worst downturns I've experienced. It's not too different from the one in 1984. Then, between October and December, bookings in the equipment industry dropped about 50%. For us, in this current cycle, between February and April, they dropped about 60% to 70%. Since we're one of the top 10 companies, I think it's indicative of the industry.
SI: The consensus seems to be we're pulling out of it. How do you see it?
Torossian: The last recession lasted some six quarters. I think this one will last about the same. We're seeing improvement since May of 1999.
SI: In your view, what were some of the leading causes of the crisis?
Torossian: It's been rightly called the
'Asian flu.' In 1984, Asia wasn't as big a factor as today. Japan was, but it's
always separated from the rest of Asia because Japan itself is such a huge
economy -- between 50% and 60% of the U.S. economy, and, in semiconductors, it
used to be almost as large as we are. This time Asia was significant because
most U.S. and European industry in general -- not just semiconductors --
depended on it for growth.
SI: What about the claim the downturn wasn't caused by industry cycles such as, for instance, DRAM overproduction, but rather by banking practices and issues unrelated to the industry?
Torossian: Certainly, that exacerbated the problem. The semiconductor and equipment industries are very capital-intensive, like the automotive and oil industries. On the average, they spend somewhere between 17% and 23% of total annual revenue in capital equipment -- that includes the front as well as the back end and facilities. For a $160B to $180B industry, depending on whose numbers you use, that's about $35B spent globally for equipment and buildings. In 1995 it was estimated that 150 fabs would be built by century's end. These are $1B to $2B installations. Now, probably that number will be 10 to 20. That's a huge change in the investment scene.
Historically, in Asian countries (unlike Europe and the U.S.), capital was always available. The industry worked on a 5:1, 10:1 debt equity ratio. Asian companies had, for example, $5B in capital and $50B in debt, because they borrowed from banks, and the government backed the banks. In the case of an international bank, foreign capital would flow into it without fear of loss because the government stood behind it.
SI: That makes quite a difference.
Torossian: Certainly! If you don't have to pay the real cost of capital -- 5% to 10% -- you can invest without worry. If you want to go into a new business you just do so. Some years ago Samsung decided to enter the automobile business because it calculated that by the 21st century, cars will be the biggest users of electronics, with 20% to 25% of their content electronic. How many U.S. electronics companies could do this? If an American or European corporation decided to do it they wouldn't have to ask the government, but they'd have to do it based on their own capital resources, marketing analyses and investment strategies. For Samsung, the only impediment was getting government permission.
SI: But there was a price.
Torossian: Yes. They didn't have to concern themselves about available capital, but then the DRAM market fueling everything dried up. They may still be going into automobiles, but now they must justify the capital. Also, Asia, including Japan, experienced great growth rates by targeted industrial policies and protective markets. It rocketed them into a substantial share of the global high-tech industry. When prices collapsed, so did they, under the huge debt load. ... The IMF and to a large extent the U.S. and Europe shoveled billions of dollars and imported tons of goods to bail out Asia and Russia and Latin America.
SI: Has the delay to 300 mm favored SVG's focus, which is more on linewidth than on wafer size?
Torossian: In 1993-94, when I was chairman of SEMI/SEMATECH, we worried about the move to 300 mm. There was much rhetoric about how the only way to keep up with Moore's Law and bring prices down was by going to a larger wafer because as device complexity increases chips get bigger, and therefore the number of chips you can pack per wafer goes down. It's estimated that the move from 100 mm to 150 mm cost the industry $4B to $5B. The move from 200 mm to 300 mm could cost $14B to $16B. Who would pay for it? For 150 mm Intel was first; for 200 mm it was IBM. Both took a financial hit.
In 1996, I was part of SEMATECH's Capital Productivity Executive Group, and one of the few who said that if the move took place at the right time (that is, three to four years later), it might cost $14B to $16B, but could be paid for just by our normal 10% to 12% R&D expenditures with adequate return on investments. But if we accelerated the introduction of 300 mm wafers to one or two years, it would cost an incremental $2B to $4B because equipment manufacturers would have to build new prototypes, and by the time semiconductor manufacturers used them in production some three years later, these machines would be obsolete. The equipment industry will pay dearly for this move, through rushed prototypes.
Whenever we've pushed technology without considering results, it has been expensive. It's best to go a step at a time, even if an intermediate step is short-lived. Technologies must mature -- 0.35 µm did not last long; everyone is on 0.25 µm now; but we did not skip over it.
SI: You've often said there's a technology shortage in the industry. What needs to be done?
Torossian: Even though there is a capacity glut, there really is no excess capacity in 0.25 µm. Device manufacturers are pushing 0.18 µm. A couple of years ago Taiwan didn't want to touch 0.25 µm step-and-scan technology. They soon realized the only way they'd get the shrinks they wanted for their DRAMs was to go to step-and-scan. You'll find that 0.18 µm step-and-scan and DUV technologies will help move companies out of the recession we have been experiencing. Already we're being asked to delay shipment of current technology and accelerate technology development for the next generation to prepare for the coming upturn.
SI: Isn't that an expensive proposition for equipment manufacturers?
Torossian: Very. But R&D must be kept up even during bad times. We also must prepare manufacturing capacity for fast ramp-up to the next generation. That's even more expensive than R&D, because when customers are ready they don't want prototypes. Today, they're unwilling to work with you in the development cycle because their cycles have shrunk from six to two years and they need a fast ramp-up. This means they expect new technology to work from the start. So while companies work with 0.18 µm, we must work on 0.15 µm and EUV for 0.10 µm and below.
SI: Considering SVG has been built in for production of advanced devices, why aren't you selling more systems?
Torossian: We were designed in for full manufacturing for 0.25 µm and sold many systems. Suddenly, customers decided to push 0.18 µm! We've given it to them but were left with an excess of 0.25 µm machines, which they're not buying.
SI: What do you view as SVG's next growth areas?
Torossian: To date, we have shipped more than 160 DUV step-and-scan systems for production. We address about 50% of this leading-edge market by about 50 machines per year. This segment of our market will grow from 100 to about 600 machines in two to three years. Over the next two or three years, we expect to take about 200 of those 600 machines. That'll give us about 25% to 30% of the market -- very healthy considering that at $5M or $10M per copy, you're looking at about $1.2B to $1.5B in sales. That's just for the lithography division!
SI: Which aspect of the industry would you like to see changed?
Torossian: Increasingly, equipment manufacturers have borne the brunt of technology investment and provided warranties and long-term service contracts, which, by the way, we aren't paid for. It's similar to the old days when IBM was the mainframe king and other software suppliers did not get paid for software that everyone expected to be included. Today, software is -- in its own right -- a unique and very successful industry. They typically charge a 10% to 15% annualized support fee. The equipment industry's warranty cost has increased from 2-3% to over 6% of sales in the last 10 years.
Much of the semiconductor industry's process know-how is shifting to the equipment manufacturer. We're expected to know a lot about lithography (which we do), etching, deposition, ion implant and, in many cases, to teach companies about it. The day is fast approaching when this ever-increasing effort we must put forth will have to be paid for.
SI: What would you advise your peers and the industry?
Torossian: (Smiling) Lower your cost of
manufacturing and reduce your cycle time. 
