TSIA President Assesses Upturn, Effects on Asia
-- Semiconductor International, 11/1/1999
SI interviewed Dr. Genda J. Hu, president of the Taiwan
Semiconductor Industry Association (TSIA), and general director and vice
president, respectively, of the Electronics Research & Service Organization
(ERSO) and the Industrial Technology Research Institute of Taiwan. Hu has a BSEE
from the National Cheng Kung University, and a Ph.D from Princeton. He worked at
IBM's T.J. Watson Research Center, where he pioneered CMOS technology for IBM;
and later joined Xerox's Palo Alto Research Center, where he demonstrated the
industry's first latch-up-free CMOS operation. In 1984 he participated in the
start-up of Sierra Semiconductor, and he later joined Cypress Semiconductor as
director of non-volatile technology. Before joining ERSO/ITRI in 1996, he was
vice president of technology at ISD. Hu is an IEEE Fellow.
SI: How do you view the state of the industry?
Hu: Everyone agrees we've gone through the worst of the downturn. Things are picking up, and DRAM prices reflect it. The price of 64-Mb DRAMs has rebounded nicely, although experts are cautious about whether this is a long-term effect, a short-term inventory adjustment, or a byproduct of Y2K and people are stocking up in case something happens.
SI: How do you see it?
Hu: It's difficult to say. All that is possible. However, as early as two years ago when people wondered when the DRAM slump would end, many experts predicted that supply and demand would balance out during the bottom half of 1999. Since then, I've not heard anyone say differently, and I'm therefore not surprised by the upturn. If Y2K is the driver, then I anticipate that during the first quarter of 2000 we'll have a slight drop because demand may drop slightly, and then pick up again. I remain optimistic.
SI: The downturn we've experienced could be described as catastrophic. Has it changed the semiconductor industry in Asia?
Hu:
 |
|
Dr. Genda J. Hu, president of the TSIA and general
director and vice president, respectively of ERSO and
ITRI. |
Past history has shown -- at least for the last two or three times -- that whenever we've had a slowdown it has been short: under 12 months. That had an effect, because companies worldwide tended to analyze past data and say, 'Oh, this slump is just like the one before. In 12 months everything will bounce back!' So, when things started slowing down in 1996, they predicted 1997 would be a good year and prepared to reap the expected upswing's opportunities by continuing to invest and increase capacity. In 1997, the market began with a slight upswing, then took a dive fueled by the oversupply that had been created in anticipation of the upswing. Toward the second and third quarters of 1997 everything plunged, and 1998 was disastrous.
If you graphed what happened, the market vs. year would look like a W between 1995 and 1999 instead of a V. In 1996 it dropped, and it bounced back a little in 1997, then fell again. I think that was caused by the anticipation shown by many executives that 1997 would graph like a V. People are more cautious now, and investments have slowed down globally, and that has led to what I consider a healthier balance between supply and demand. But whether people's habits have changed is hard to say -- we'll just have to wait and see.
SI: Are we repeating history because we are too concerned about the past?
Hu: (Laughing) On the contrary, I think our industry's problem is that it has a very short memory! It's all building up again -- people are beginning to pour money back in because they see an upturn coming. This is how we've always behaved.
SI: The way fabs are developing in Taiwan is truly changing the industry's face. What's your perspective on this?
Hu: All of Taiwan's industry has a common vision. We're becoming a manufacturing center for worldwide markets. Taiwan has proven over the last decade that it has the right infrastructure, talent and human resources, and the necessary capital. This combination and the industry's flexibility have made Taiwan a very successful manufacturing site for the global semiconductor market. In another 10 years we'll see a major shift in semiconductor manufacturing to Taiwan -- it's taking place now and will continue at an accelerated pace -- particularly since the government has renewed the industrial tax incentives for another 10 years.
SI: The fabs themselves seem to be taking unprecedented steps, such as TSMC's partnering with Amkor and Acer on joint ventures. Will this trend continue?
Hu: This is a must. Everybody recognizes that ours is an international industry, and you have to position yourself globally. I mentioned Taiwan is becoming an international manufacturing center, but this does not necessarily imply that the actual physical manufacturing must take place here. Many international headquarters and facilities will be here, but there'll be considerable peripheral industries established elsewhere as well. For example, TSMC is building WaferTech in the U.S.; they're forming alliances with Philips and the Singapore government and building a fab there. UMC acquired Nippon Steel's plants in Japan, and those facilities will not only be retained, but will very likely be expanded. Down the road, Canadian facilities may also be in the cards -- you have to be as close to your customer as possible.
SI: Is the increase in fabless semiconductor manufacturers stimulating this trend?
Hu: Fabless semiconductor manufacturing is a great concept. However, if you look at the industry's revenue numbers, it's evident that the fabless portion still is a relatively low percentage. The IDM still makes up the largest share of revenue generation. So fabless manufacturing still isn't a large enough segment to make a strong difference in terms of the world's changing landscape. It's an important driver and has successfully proven the contract manufacturing concept, but it isn't as yet that important.
SI: It's becoming increasingly so, however.
Hu: Yes, particularly now that IDM companies are starting to release some of their production to fabs. That's a fundamental strategic change in their business model. As president of TSIA, I've been asked to talk about Taiwan's foundry strategy, and I provide the reasoning why the foundry business model will be successful for the whole industry, rather than for just Taiwan. Most IDM companies perceive the foundry as a threat. They view it as taking away their function, their work. I ask them not to think that way and explain why fabs are successful, and why IDM companies should alter their philosophy.
Foundries are successful because they serve multiple customers rather than one. A fab is expensive to build and, unless fully used, it's difficult to recoup the investment. For example, everyone knows that a product's life cycle is getting increasingly shorter. If you're a single-customer fab and your customer comes out with a new product and a year later it is obsolete, your utilization rate falls through the floor. You've lost your customer not through any fault of your own, but because the market is gone. On the other hand, if a fab has multiple customers with multiple products it maximizes the capacity and the utilization rate becomes very high. This is why foundries are so successful.
IDM companies work the other way. If they have a product they want to sell but with insufficient capacity, they build. And after spending billions of dollars and a couple of years building, your product is gone, and you're left with an inexpensively priced product, making it hard to recoup investments. Regardless how good the product is, the IDM company always seems to be behind the wave.
A good strategy for IDMs might be to farm out, say, 30% of their manufacturing to a foundry, instead of building another facility, and keep the remaining 70%. That way they could use their manufacturing muscle to build higher-cost products and maximize returns. If the market turns sour, that 30% can be taken back, leaving the foundry to worry about maximizing its utilization, while the IDM company maximizes its own. This strategy benefits everybody, allowing IDM companies to concentrate on core competencies such as R&D and design.
SI: When you consider the industry's state of affairs in Japan, South Korea and Taiwan, things still aren't looking very good. Has the downturn changed business practices in those regions?
Hu: I believe so. Japan's certainly changing its business strategy. Many companies aren't making large fab investments; at least not as readily as before. For example, it's very clear Toshiba has changed its main strategy. They've begun forming alliances with companies in Taiwan, such as Winbond and WSMC, instead of building new fabs. They license out their technology and get royalties, while the Taiwanese companies make the investment to establish manufacturing capacity. Mitsubishi also is changing in the same way. Japan is revising its business models and pulling out of the slump.
SI: What about South Korea?
Hu: South Korea is different. Companies there haven't yet changed their business models. They want to retain control over everything they do and maintain the advantage of being large in their so-called scaled economy. But they're suffering from currency depreciation, as well as from domestic and financial problems that create obstacles to raising funds. As the upturn kicks in, they'll probably regain some of the lost momentum, but they've lost a crucial two years of sustained investment, and it'll be hard for them to come up to par with the rest of the industry. This is most noticeable in the FPD area -- during that two-year period, seven companies have jumped into it. Originally, South Korea attempted to divide this world market with Japan. Now, it'll probably end up in third place within a year or so.
SI: Some describe Japan and South Korea as 'isolationist.' Is this something that does not fit in our industry?
Hu: Definitely! I am a firm believer that isolationism must end. Isolationism vs. openness is probably one of the most fundamental factors why the U.S. has been so successful over the past two years and why Japan has suffered so much. Japan tends to be isolated, closed and protective, while America tries to be as open as possible in the correct belief that free competition stimulates progress, ingenuity and innovation, while promoting collaboration and alliances, at least in most cases.
We did a survey on collaborations and alliances of technology and investment, starting in 1996. We studied some 250 cases worldwide, divided into regions and countries. We found out that the most active country is the U.S., with the highest number of involved companies.
That explains why you're doing so well; you're open and dynamic. These days, there's really nothing you can hide. If you have anything new, you might keep it for six months, perhaps, and then someone else will reinvent and improve it. A good example is how IBM's open architecture concept beat Apple in the PC industry. Openness is crucial, and Japan is starting to realize this. Some of the major corporations are bringing in younger people with different ideas into their boards of directors.
Japan's work ethic is also changing; their lifetime employment concept has been severely challenged. I tell my people to forget about lifetime employment. The only guarantee of lifetime employment is lifetime learning; you've got to learn, keep up with the world and maintain your own value. If you do this, you will always be employed. That's lifetime employment!
SI: What advice would you give to your peers in the industry in the U.S., Taiwan, Japan and South Korea?
Hu: I think U.S. companies must get the government off their backs. The American government's exerting far too much power over the industry, and there is too much politics.
SI: Sometimes, however, it isn't the government that comes to the industry, but the other way around.
Hu: It's the same thing. Your companies need to dry their tears, blow their noses and grow up. They shouldn't have to rely on the government to solve global business problems; that's something they've got to work out on their own. Anti-dumping suits are good examples. History shows that you've been able to do that very well on your own. I've great respect for American industry. If politics could be left aside, there would be no stopping it!
SI: What about Japan?
Hu: Japan must radically change its world view. They've got to recognize that Japan is not The Sacred Market. They must realize that the global market is far more important. Unless they really open their borders (and hearts) it'll be extremely difficult for them to become a true industry leader, regardless of how superior the technology.
SI: South Korea?
Hu: South Korea is a difficult region for me to comment on, because they have a fundamentally different way of doing business. They've built a great economy, but are seriously lacking in flexibility. They must become less monolithic, allow more competition -- from within and out. The number of national major players there has gone down from four to three, and they must proceed in the opposite direction. Only competition brings out the best. Also, when one talks with their workers it becomes apparent they're not a happy bunch. That's something they must pay immediate and close attention to, or they won't go very far.
SI: And Taiwan?
Hu: We're a very open society, and it's one of our strengths. That being said, however, we've yet to learn the meaning of collaboration. For example, the competition between TSMC and UMC is so fierce that it's reached a point at which their people literally will not even talk to one another. That's one of the reasons why I respect American industry. Intel and AMD are direct competitors, yet they work together at SEMATECH. At ERSO, we attempted to establish a similar consortium. We wanted to promote collaboration and work on precompetitive technologies, to reduce everybody's risks and try to change the color of the red areas on the ITRS Roadmap, denoting technical challenges with no known solutions. No one company can solve them due to the risks, lack of resources, and economic factors. We need to pull in everybody's resources to work on the red areas. We encountered problems because Taiwan has overstressed competition and ignored collaboration. This must change.
SI: What lessons do you believe were learned by the Taiwanese semiconductor industry as a result of the recent severe earthquake there, and how do you think the rest of our global industry should profit from this experience?
Hu: Decentralization is key. Fortunately, Taiwan's IC industry is located away from the central part of the island, where the quake hit, so it was not damaged as severely as people worried. Electric power disruption turned out to be more of a problem than the earthquake. Power was restored for Science Park in less than five days after the quake, thanks to the government's support to the high-tech industry.
The importance of Taiwan in the global information industry is easily seen from the stock market responses in New York and Japan right after the quake. Decentralizing the nation's manufacturing facilities is certainly a wise strategy. Years ago, a new science park was planned in the southern part of Taiwan. The first phase of infrastructure such as roads and utilities has been completed, and companies started moving in about a year ago. This new park will certainly alleviate some of the concerns towards the 'clustering' effect of Taiwan's semiconductor industry.
SI: Look ahead to the next couple of years; what do you see?
Hu: I'm convinced the next two years will be a
boom time for the semiconductor industry. (Laughing) I just hope we do
not overheat as we did in 1994 and 1995! This is a good time to cash in on the
coming good times as well as prepare for the next downturn, because there's
always a downturn. We must
accept the fact that our industry is like a roller coaster; we cannot change
this -- we must be cautious on the top and tough when we're on the bottom.
