Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Manufacturing/Packaging

T.J. Rodgers on politics and nano

Posted: 16 Aug 2006 ?? ?Print Version ?Bookmark and Share

Keywords:Brian Fuller? EE Times? T.J. Rodgers? Cypress? embedded systems conference?

Rodgers:programmability is the future.

Rodgers: I think programmability is the future. And once you get to programmability, software is king.

EE Times caught up with Cypress Semiconductor Corp. founder and CEO T.J. Rodgers at this year's Embedded Systems Conference.

EE Times:What's business like for the next six to 12 months?
T.J. Rodgers:I came from my staffing meeting just before this, and we were worried. We are worried that 2006 could be a repeat of 2004!with a really nice first half, followed by that sick feeling when the third quarter is weak and you know the phone won't be ringing at Christmas.

Don't take my forecasts for a given because I'm not that good at it. But our forecasting services say it's going to be a strong year. Business is really good, and the forecast I went over today said that from a macroeconomic and semiconductor-specific point of view, the world's looking good for the rest of the year. And 2007 is supposed to be a barn burner.

Some forecasters are sticking to a 17 percent to 18 percent forecast for 2006. So you're seeing strength across the world?
Yes. You're seeing the static RAM business taking off like a rocket.

There's a lot of concern about the evolution and maturation of the business!from one with high annual growth averages to something more commodity-like, with lower ASPs. Is that the industry's biggest challenge?
I think the forecasters are lagging indicators. I have a Hewlett-Packard 15 calculator, and I can put in percent growth for 2001, do function linear regression and graph any line you'd like, and of course the line perfectly fits the data!right up until the data goes dink, dink, dink and down like this.

I've seen no fundamental reason to think the semiconductor industry is going to go at a slower growth rate, especially if you define the semiconductor industry in larger terms.

For example, we're now into solar cells. That's a $3 billion per year business, taking 0.03 percent market share from a $1 trillion electricity business. If you look at that market capability for semiconductors!after all, a solar cell is a glorified diode, literally!there's no end in sight.

There's talk that the next node out, 45nm, is really where Moore's Law is going to hit the wall. We tend to think that about every next node, though. What's your take?
We're putting our 65nm technology into production right now. We've generally lagged the big guys, who have a lot of money by something like half a generation, but we've always been there all the way since 1982. I do not see technical barriers preventing Moore's Law from going forward. However, having said that, I do see an end to Moore's Law!and the end of Moore's Law will come in the boardroom, not in the wafer fabrication plant.

The newest stepper for 45nm!assuming you guess right and don't buy the wrong one!is $37 million; it costs more than a Boeing 737. Think about having three or four of them. You're looking at $150 million.

Ask yourself what you have to make and in what volume just to pay for your steppers, let alone all the other equipment. Clayton Christensen, a Harvard Business School professor and authority on disruptive innovation, has talked about technology overshoot. We're in a classic case of technology overshoot, where the return-on-investment is just going to go away. We're not going to do the 45nm technology on the Moore's Law time frame. There are fewer and fewer companies getting the benefit of Moore's Law as they move up on the technology curve. I think it can be done, but I don't think it matters that much economically any longer.

A few colleges!Stanford, MIT and others!have created interdisciplinary programs to marry electrical engineering and biology. Do you see a coming together of those disciplines in the commercial space anytime soon?
I have two data points. In gene sequencing, once you understand the sequence of genes in an organism you can go in with bioengineering and change one gene to modify the characteristic of a plant, as opposed to just doing a crude DNA swap where you put two plants together. For instance, you can engineer a grape that's more drought-tolerant than it was before, but still makes great pinot noir.

The technology needed to understand the gene sequence!that's going to go to silicon. There are startups in Silicon Valley coming into our company saying they want us to build holes so small that one DNA molecule will fit in them. They want to watch it fluoresce and find out what it is. And they want millions of chips.

On the nanotechnology front, I'll tell a story about my car. I bought a car, and I wanted to protect it. So I remembered from high school a thing called Turtle Wax that you use to prevent oxidation of the paint. I went to the hardware store and bought a $10 bottle of Turtle Wax, and the bottle said, 'Now improved with nanotechnology.' There are more con men and charlatans in the nanotech world than I've ever seen before anywhere else.

Let's talk about an area where you're trying to get some momentum, as are others!programmable silicon.
When I listened to Altera Corp. chairman and CEO John Daane talk about programmability at the ESC last year, he had it about right. As line widths go down, even a 65nm wafer is expensive. But you can cram so much stuff on a wafer, you can afford to be inefficient in terms of transistors per square millimeter. Instead of designing 10 chips for 10 customers, if you can design one chip that can be programmed for 10 customers, you can amortize a typical programmable chip, for which development is about $20 million. It's the way we're going to go.

I think programmability is the future. And once you get to programmability, software is king.

If you ran for office!say, governor of California!what would be your first piece of legislation?
Let me reject the premise of your question. I have a real job, I create real wealth, and real people work for our company. I would never, ever swap my job for a job where I make my living taking money from other people. I would rather be unemployed, literally.

Some people have said that after so many router and switch architectures went south in the crash, ternary content-addressable memory became a dumb and expensive idea. More recently, we've heard that despite the number of SRAM-like alternatives out there, there will always be a unique purpose for a ternary CAM, and it will always have a pretty good average selling price.
I don't see ternary CAMs being replaced. We do have an SRAM alternative internally in the Sahasra architecture: Sahasra is basically a large logic engine with several blocks of high-speed static RAM embedded in it, and CAM. It is twice as silicon-efficient as standard CAM, but the penalty is that the latency for the first answer is longer. It can serve in some functions, but not others. But it is cheaper by a lot than standard CAM.

Design engineers get really frightened at offshore trends. Should they be worried that there are no more U.S. design jobs?
Cypress has 17 design centers. Offshore, we have India, China and Ireland. So most of our design centers are in the United States, and we would be crippled if we didn't have them. There is a warning to the engineer, but the warning is a little bit different. It's the same warning for any company: Tell me what you're going to do when you recognize we can get something offshore.

You must change, you must move to the next level of value creation. If it's just turning the crank on me-too chips, that can get done anywhere in the world. So the challenge for our design centers is to embed software and systems knowledge to move to the next level.

Do you worry that the venture community is getting too regimented about saying, 'OK, I just want four or five guys in management here, and everybody else in your startup should be based in Bangalore and Shenzhen'?
I don't see that too much in Silicon Valley. They give you credit for having a low-cost deployment, but the fact is, Silicon Valley is still Silicon Valley. There are still thousands of engineers there, and they're very smart. They're embedded in a highly value-added matrix. You can't just go somewhere in the technical hinterland and necessarily compete. But they do have to come up with a better idea. They can't just turn the crank on something that's already there.

- Brian Fuller
EE Times

Article Comments - T.J. Rodgers on politics and nano
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top