Edward Ross, TSMC President
-- Semiconductor International, 11/1/2000
 |
| Edward C. Ross (Source: TSMC) |
SI: BusinessWeek recently named TSMC as the world's fifth leading infotech company, and Red Herring sets you among the Top 100 companies. This must be gratifying.
Ross: Immensely! We do view ourselves as a world-class company and are extremely happy to receive the recognition we have been given.
SI: Based on all this, what are your plans and strategies for TSMC?Ross: We're the largest foundry in the world, as well as the leading technology foundry. Because of this, our plans and strategies are pretty consistent, which is to continue building the capacity that our customers need, and doing it at the cutting edge of technology.
SI: Not exactly an inexpensive undertaking.Ross: No. We're undergoing a very massive capital expansion right now, and investing slightly under $5B this year. This is a huge investment for a company of our revenue size, as well as an indication that we're bullish on the industry.
SI: Where is this money going?Ross: We're investing in large capacity increments. For example, the capacity increase from this year to the next — the delta capacity — is about 1.5M 200 mm equivalents. That's a large increment; it's a number that's difficult to get your head around.
SI: How is this spread out?Ross: We're building out our existing fabs' capacity, so there's a number of 200 mm fabs where equipment is being installed as we build out the capacity. We're adding capacity to our joint-venture fab in Washington. We're bringing up a joint-venture fab with Philips in Singapore, and we are in the process of building two 300 mm fabs in Taiwan — one in Hsinchu and one in Tainan. Our pilot line for 300 mm is in Fab 6 and that is nearly finished — the equipment set is complete.
SI: Obviously, these strategic alliances are an important part of your plans.Ross: No question. The key point is that we anticipate that by 2001 we'll be the largest suppliers of silicon in the world — including all the major IDMs.
SI: How far will you be above your closest competitor?Ross: (Smiling) We'll be just above the water to the closest IDM, a Japanese that manufactures primarily DRAMs. We will have a clear lead of about 4.8M 200 mm equivalent wafers, by 2001, over our closest foundry competitor. Afterward, we expect this growth trend to accelerate.
SI: Will this growth rate require you to change some of your procedures?Ross: We don't expect too much change in that respect. The company is well positioned to handle this — we're well structured, well organized. It's just a question of scaling up, and I don't expect that there'll be any great changes in the way we do business. We've been growing rapidly already, and have a system in place to move skilled people from one fab to another to facilitate ramp-ups. It's a practice that we've gotten pretty good at. One of the advantages of size is that we have a dozen or so fabs with a pool of skilled people at all levels. So as we bring up an incremental fab, we can put a team in place that's completely tuned up, skilled and ready to run it like a hose. That's part of the "secret sauce," the infrastructure that we bring.
SI: In short, you seed your fabs with experienced people.Ross: That is exactly what we do — we take our most experienced people and put them in the new fab.
SI: Within all this activity, which are the major things that you believe you are doing differently from your major competitors?Ross: Probably what really characterizes us is that we are single-mindedly focused on being a foundry supplier. We've not diverged from this. This is how the company was formed, and that's the way we are today. We've never ventured into product spaces, or any of the things others have done. The other thing is, of course, that we have intercepted the SIA Roadmap. We're at a world-class technology level developed internally, by ourselves. A truly unique strength that we bring to the table is our internal R&D organization, which is capable of developing the specific technologies that our customers require.
SI: That certainly must be the case, since you're now licensing your technology.Ross: Yes. As you know, we recently licensed our technology to National Semiconductor. That was the first time ever that a foundry has licensed technology to an IDM. In the past, it has always worked the other way, and I believe this gives a clear indication of where we are.
SI: What is your R&D expenditure?Ross: It exceeds $128.4M.
SI: Obviously, much of your growth comes from adding capacity. However, within that, what do you see as your next growth areas?Ross: The first thrust is that we're continuing to add capacity at the cutting edge, so that all the fabs that we are putting in right now are capable of going to 0.13 µ m and below. We don't fill new capacity with old technology. The second thrust is that we've probably built our last 200 mm fab — it's all 300 mm from here on. Our strategy and intention is to remain the dominant player in our space.
SI: Does it seem to you that growth is becoming increasingly difficult to achieve?Ross: Revenue growth is very dependent on the customer base. In the past, we've been driven largely by the fabless semiconductor segment, but always had a reasonable sprinkling of IDMs as well. Now there's a great movement on the part of the IDMs to move to an outsourcing arrangement. Se we're now getting healthy growth from both segments.
SI: At what level do you see this trend in, say, five years?Ross: Some predict that in seven to 10 years foundries will account for more than 50% of the world's silicon. This appears to be an irreversible trend.
SI: Speaking of trends, which industry trends do you think we ought to be paying more attention to?Ross: Clearly, the industry right now is being driven by wireless communications and the Internet infrastructure. It's anybody's guess how long this'll last, but it certainly looks like it will be a sustainable source of growth for quite some time. There probably will always be bumps in the road, but the world is still running at full tilt down these two paths. This is not to discount the PC market, which is still very big, but it has become more of a commodity market.
SI: What would you like to see fixed in our industry?Ross: The biggest need — for IDMs as well as for us — is accurate forecasting of need. Not only by quantity, but by mix. As you know, many of these cutting-edge technologies are really not fungible one to the other. So you install a certain amount of capacity of A technology and a certain amount of capacity of B technology, and if you don't get the correct mix you end up with underutilized assets — and our whole function is asset utilization. (Laughing) Other than that, ours is a really good business to be in.
SI: When will you finish the facilities that you have under construction at present?Ross: The last one to be finished will be the 300 mm fab in Tainan. It's due to come on line at the end of 2001.
SI: Once the dust is cleared, what do you estimate your wafer capacity will be?Ross: It's difficult to say, because in the meantime we'll be building other installations. We've put some numbers out for guidance. For example, for 1999 we had 1.8M 200 mm equivalent wafers, for 2000 3.4M, and for 2001 we estimate 4.7M.
SI: How do you plan to incorporate up-and-coming technologies — what is TSMC's technology roadmap?Ross: Basically, it's the SIA Roadmap accelerated. When we do a particular technology node, such as 0.18 µ m as an example, we bring out a "vanilla" 0.18 µ m logic technology that is basically right on the SIA Roadmap in terms of technology requirements, but ahead in terms of product schedule. Then add specific modules to it — as demanded by our customer base, such as a low-voltage module, a low-power module, embedded flash modules and embedded static RAM modules. We bring on a baseline process and then bolt modules onto it to meet our customer base's demands. All of this takes place simultaneously in the same fabs, on the same lines.
SI: So the R&D is done on a production line?Ross: Correct.
SI: Any acquisitions on the horizon?Ross: (Smiling) If there were I could not talk about them.
SI: How do you pitch an IDM on going fabless?Ross: We don't have to pitch them. They pitch us. The reality is that there's an optimum balance — for the big guys — of owned fabs vs. outsourced. First, what you do is lay off the risk of asset underutilization by going outside. Would you go outside if you paid more than when you build it inside yourself if you are fully utilized? So if you think it through, there's an optimal balance of internal vs. external. Secondly, one of the things some of the IDMs are doing now is outsourcing specific technologies, so that they don't have to develop and maintain them. This enables them to focus on a more narrow range of production technologies. Of course, when you are talking about a large IDM, they're looking for lots of silicon, so you have to be a quantity provider — such as we are — to get their attention because some are outsourcing very large numbers of wafers.
SI: What about the smaller IDMs?Ross: Mid-size IDMs come to us primarily for advanced technology, because many of them cannot afford to develop it themselves. They have considerable legacy or specialty technologies, and the mainstream, cutting-edge technologies they outsource to us. It's a trend that, barring a catastrophic change in the business environment, will continue.
SI: It would appear that, from your perspective, fabs are becoming the industry's technology drivers.Ross: Unquestionably, foundries are the basic technology drivers. Don't forget that the customers tell us that they need a low-power module, or a specific transistor for a specific function, so it's a combination of both. That is, we get a lot of important technology-related feedback from our customers while we're in development with a process.
SI: Do you think the time may come when the fabs, rather than the IDMs, decide which processes are to be implemented? Some smaller companies seem concerned over the possibility that the giant fabs may decide sometimes to postpone the implementation of a process requiring new capital equipment, until the time when they have gotten their ROI on the current equipment.Ross: This isn't a concern for us because we build out to a particular level. A couple of years ago we were putting in a lot of 0.25 µ m technology. We don't go back and retrofit those fabs to work on another leading-edge technology; we just build new fabs and continue to run those old fabs at capacity. We have plenty of customers with legacy products, such as image sensors, that are on 1.0 µ m technology — old stuff.
SI: So you don't see a time when some of the giant fabs might say, "We don't believe we ought to have one generation every 18 months; we think it should be once every 36 months."?Ross: That might happen. But if it does, it won't be an arbitrary decision. Remember, we keep building new fabs so if we need another generation we tend to put it in a new fab. We don't have the problem of having to have a whole sequence of technology nodes in the same fab.
SI: Understood. However, what's the average fab cost these days?Ross: Two billion dollars.
SI: How many of those can you keep building?Ross: (Laughing) As many as the customer base will support.
SI: You mentioned that, although you have considerable resources, TSMC is not going to go on the product side. Can you amplify on this?Ross: That was Morris Chang's basic philosophy when he founded the company. The simple and direct point is that we don't intend to compete with our customers. If we maintain that position — and we've been scrupulous about it — it obviously gives us a deeper insight into what our customers are doing and want. They'll open up to us more broadly in terms of their plans because they view us as a partner, not a competitor.
SI: What should the industry do to try to ameliorate the cycles we undergo regularly?Ross: These cycles are the biggest issue facing a capital-intensive industry. Of course, if in the downcycle you are not using a substantial proportion of your capacity, it begins to hurt financially. Two of the things that are moving towards ameliorating this (although I'd be the last to call it a paradigm and say that nothing negative is going to happen), is that we have these large industries that are growing very quickly — wireless communications and the Internet infrastructure — which seem to be very strong. Whether they'll continue growing at the same rate nobody can tell. However, it's unlikely that they will stop growing anytime soon. The second is that when you get to be our size, with a dozen or so fabs, the increment represented by the next fab you build is on the order of about 10% of your capacity. This means you have flexibility on how to outfit it and how the equipment is brought in, and so forth. On the other extreme would be a one-fab company, for whom it's all or nothing. If there is a downturn, it can really hurt you then.
SI: Care to guess when the downturn may come?Ross: (Smiling) I'm extremely bad at that sort of prediction. All I will say is that demand is very strong right now. Over the macroview, we're expecting our business to grow 28% or so over the next decade. Whether that will be 50% on one year and 10% the next is difficult to factor, but given the dynamics underneath the growth of the customer base, the outsourcing of the IDMs, our position in the marketplace, we think we have a good future ahead of us.
SI: If you had your peers in front of you, what advice would you give them?Ross: (Laughing) Get out of the business! Seriously, I don't think they're in much need of advice. This industry is pretty well understood now. We've demonstrated a model that works very well, which is build your capacity prudently. Few of our competitors can go head to head with us. The others we believe are going to become more specialized either in the technologies or industries that they serve. They're all smart and know what they're doing, and need no advice — they know how to play the game. •
— Alexander E. Braun
