Richard Templeton, president and CEO of Texas Instruments , sat down with Electronic News to discuss his company’s growth prospects, its position on foundries, and its stance on cooperative technology development. Excerpts:
Electronic News: What’s up and what’s down in TI’s business?
Templeton: From an intermediate and longer-term perspective, we just couldn’t feel better about where we are. The world is driven by communications and entertainment. We’ve certainly got a good lineup of products with our focus on DSP and analog, which has been going on for well over 20 years. Our Digital Light Processing is front and center in a world of communications and entertainment. In our opinion, those are going to be front and center for the next 15 or 20 years.
Electronic News: Why?
Templeton: Just as the mainframe gave way to the minicomputer, the PC is giving way to that sector right now. Pick an area and you can draw the analogy. The PC is not the driver of technology anymore. It’s usually communications or entertainment, or a combination of those two. From that perspective, we think we have the right markets, the right products and the right customers. It doesn’t mean anything’s a given. You have a lot of hard work to translate that from opportunity to results. Over the past three years we’ve been able to grow at rates above most of the significantly sized semiconductor companies. When you look at 2004, it was a year that began very strong and ended very weak. I think most people will agree that the inventory situation was straightened out by the end of March of 2005. It was primarily an inventory correction.
Electronic News: TI historically has operated as a company that does everything itself. You’re nowusing foundries to augment your existing capacity. What’s driving that change?
Templeton: Three years ago we publicly started discussing where we were going. We wanted to have a more responsive manufacturing machine. The unpredictable markets we have tend to go along with a digital consumer/ entertainment or communications environment. They grow very fast. They don’t announce themselves. As a result, we wanted a more responsive system. We put very deliberate plans in place back at the 130 nanometer node and it really served us well. In the second half of 2003 when things turned up very strong, we were early to market with 130 nanometers and we were able to outgrow all of our semiconductor competitors because we had an augmented foundry model. We turned around, took advantage of that as things slowed down in the second half of last year, and at the end of 2004 we were ramping that up on 90 nanometers. We have found the ability to integrate a foundry element to have a more responsive machine without compromising our time to market for new technology.
Electronic News: How much of your manufacturing is being done by third-party foundries?
Templeton: It’s less than 20 percentof the total wafers.
Electronic News: Is that oldergeneration manufacturing?
Templeton: No, in fact it’s the opposite.
Electronic News: Does that mean when you begin pushing into 65 nanometers, you’ll do that outside?
Templeton: We’re in the process of ramping up 90 with a good mix of external, and you’ll see us do the same at 65. It’s a very conscious choice that we made back five years ago. We’ve talked about it for well over three years. There’s no magic to those percentages of what there are inside those nodes. If we believe we can get enough 90 nanometer or 65 nanometer outside and at competitive pricing, we don’t mind letting a higher percentage on the outside. If we don’t think we’re going to get the support for customers that we need on the outside, we’ve got the ability to build it internally.
Electronic News: Is TI doing more make-or-buy decisions than in the past?
Templeton: Compared to 10 years ago, when we were doing none on the outside, we’re clearly doing more of that.
Electronic News: But is that a new approach?
Templeton: We have no intention of putting any high-performance analog on the outside because you have to tool it once, you want to run it in one place, and the cost model of just being effective to do it once and do it right says do it internally. It’s also the case that if you give us four or five years to get use of the assets, we can make good returns on those. At the other extreme is advanced CMOS. The reason you’ll do that on the outside is that if we ramp up a wireless node on 130 nanometers, we’ll bring it down in two or three years. We use it inside the life of the equipment, so don’t put a bunch of fixed costs in place that you’re not going to be able to utilize for a long time. That’s really the cost decision we’re making. We can’t get asset life out of 130 nanometers because we’re going to ramp it up and bring it down in two years. Don’t build to your short-term demands, build it to your long-term demands. How much ASIC will have 5- and 10-year life?
Electronic News: If you do that type of cost analysis, do you also consider whether to design internally, as well?
Templeton: The real challenges right now—power dissipation and power utilization of the overall chip—are probably becoming bigger than raw performance. TI is probably ahead of most companies on these because we’ve worried about power a lot longer because of the market we’re in. At 90 nanometers, the benefit of circuit meets process meets architecture inside one company where it can be done quickly and can evolve quickly is a competitive advantage. Why did TI ship 5 to 6 million units of 90 nanometer chips when very good fabless companies in the systems business are just ramping 130? It’s because when they’re that separated from the basic modeling and basic process development, it’s going to be longer before they can take advantage of it in a cost-effective and a power-effective way. Our model is to control process development and circuit design and development. The foundry becomes an augmented role. The foundries have to match our electrical specs. We don’t use their libraries. They use ours. We need to have the ability to rapidly ramp these things up. That’s not a ‘going-to-do,’ it’s a ‘has-done.’
Electronic News: Is your R&D as a percentage of revenue holding steady or rising?
Templeton: Compared to 10 years ago, it’s up. We ran as a company about 15 percent last year. That number probably peaked in 2001 in the low 20s. That’s because our revenue declined 40-something percent. But our R&D only declined modestly. It’s been coming down by percentage for four years, but only because 2001 was such an aberration because of the market backtracking so rapidly.
Electronic News: Are you planning on doing research with partners, similar to what Philips and IBM have done?
Templeton: No. Technology development is not a team sport. I’ve been there and sat at the table where the goal was to get the best of this company and the best of that company. What you got was the least common denominator because of the complexities and the speed at which those companies have to work. We’re just not big believers that these big R&D consortia are a great solution. What it really reflects is that the number of companies that can afford to be a big CMOS operator are fewer and fewer. A lot of people fell out during the 2001 downturn.
Electronic News: I think a lot of that was very early development.
Templeton: There are areas of precompetitive R&D. IMEC is a great example. We work with them and the SIA and Sematech and all the nano centers. There are examples of pre-competitive R&D where you can push new structures, new film, new materials. The opportunities to do R&D in consortia and universities is very high in that case. But when you talk about applying it to products where you’ve got a good product and on time and it needs to be competitively driven, consortia are not a great answer. History says it’s not a great answer. ▫
The author is editor of Electronic News, a sister publication of EDN Asia.
