T.J. Rodgers, Cypress Semiconductor President and CEO
-- Semiconductor International, 2/1/2001
 T.J. Rodgers (Source: Cypress Semiconductor) |  |
T.J. Rodgers is founder, president, CEO and a director of Cypress Semiconductor Corp. (San Jose). He graduated Dartmouth College with a double major in physics and chemistry, and later Stanford University, where he earned a master's and Ph.D. in electrical engineering. While at Stanford he invented, developed and patented VMOS technology, which he sold to American Microsystems Inc. (AMI). He managed the MOS memory design group at AMI from 1975 to 1980 before moving to AMD, where he ran the SRAM product group. Rodgers founded Cypress Semiconductor in 1982, building it into an international supplier that designs and manufactures hundreds of ICs in four product divisions: static memories, data communications, interface products and timing technology. Rodgers was the 2000 SIA chair and is a board member for C-Cube and Silicon Light Machines.
SI: You once said it takes a special investor to invest in high tech as it requires discipline and courage to hang on.
Rodgers: Absolutely! Our stock tanked in the crash of 1996. Then we had a wide spectrum of investors, from institutional to retail. There were a lot of upset doctors and dentists calling us. Profits and revenues were up and we'd beaten Street expectations, but they kept calling. I wrote a booklet, "Thinking About Cypress Stock," in which I analyzed our ups and downs since we went public in May 1986. I calculated, on a day-by-day basis for years' worth of trading, a parameter I called "time to money." That is, if you buy and the next day the stock dives, how long — statistically and based on real history — would you have to wait before getting back to even and begin making money. The maximum period was a little over three years — it would bounce up and down, and our curves were similar to those of other high-tech companies.
SI: What are your long-range plans?
Rodgers: I'll address the biggest change we're working on. There's a sticking point that many semiconductor companies hit — we did — at the $500M range. Both National Semiconductor and AMD were stuck in the almost $2B range for years. The transformation required us to go from a product-line organization ("We make memories, datacom chips, USB controllers, etc.") to an end-market organization ("We make chips for the Internet, for computers") and end up serving the market.
Phase one is the semiconductor company as a technology company. Let's use Intel as an example. Intel's catch phrase in the 1970s was "Intel Delivers." They were saying, "We're a high-tech company that, unlike others, actually delivers." Their mindset was "We have a hot semiconductor process and we make whatever parts our technology gives us an advantage in," and this may be microprocessors, or whatever. With us it was high-performance CMOS — we were the first to say "CMOS" to low-power technology. So we dubbed ourselves a high-performance CMOS company, meaning we made any niche product that used high performance to an advantage, and made a number of products not necessarily all that related to one another.
Then you go into phase two, away from the technology phase to the product-line phase, with niche products you make and aggregate. We made PLDs that turned into a family of high-performance devices, our RAM family expanded. We made some specialty memories — dual-ports and FIFOs — which became a family. When you realize you cannot run diverse businesses, you appoint general managers to run the divisions. We've been there for about a decade.
If you're an entrepreneurial company, those divisions are very autonomous — they share the fabs and the sales force, but other than that they do whatever is required to succeed. That whisks you from $500M to $1.5B (where we're now), and also prevents you from breaking the $2B mark, because you have divisions that aren't very cooperative — they're like tribes — and don't present a coherent picture to the customer. ...
We're moving toward phase three — communication segment-oriented product lines. We've reorganized our product lines into four autonomous centers — memory, data communications, interface (USB) and timing technology. In addition, we're going to have four others that operate orthogonally toward the market — WANs, storage networks, wireless terminals (PDAs and cell phones) and wireless infrastructures (base stations).
Some of the P&Ls for the segments will equal the sum of the P&Ls for the centers, so we're going to count our money in both directions and ask ourselves where we're investing, what products we're making, and what returns we're getting. We'll also look at market direction — are the products strategic? Over time, we'll turn into a market organization, with product portfolios, and our different divisions will be a strength because we'll be a strategic provider with innumerable ties with large vendors, such as Lucent and Nortel.
So in the third phase you're no longer making products that fit your technology, you're acquiring technology to build other products.
SI: : What geometries will you be working on over the next two years?
Rodgers: We're ramping production on 0.15 µ m. The next step is 0.12 µ m — we've launched that project already. We've partnered with Mosel Vitelic, a Taiwanese company, and they've brought in 30 extra (for us) engineers, giving us more talent than we would have had without them.
SI: : As an end user, what's your perspective on tool manufacturers?
Rodgers: They've improved. Every segment of the food chain tries to abandon valueless activity and focus on value-added activity, which means crowding your customers a bit. For example, Intel crowds computer manufacturers. We're learning system expertise and data communications, so we have engineers who can solve problems from the optical fiber up to the first few communications stack layers. That's a higher level of integration.
Similarly, equipment manufacturers have pushed well beyond selling platforms and are selling processes. This means everything works when you buy it. We used to have equipment that never worked and was taken out to the parking lot to be rained on while we were trying to get our money back. Now everything works, but if you're cranking out garden-variety CMOS, anybody can buy the equipment and do the same. They've taken over some of our value-add, pushing us to do special and different things that you cannot get from a foundry.
SI: You've been described as the spear point of Silicon Valley's effort to get the government to grant more H1-B visas, and you regularly testify before Congress about it. What are your views?
Rodgers: Immigrants are part of our basic fabric. We wouldn't be the company we are without them. We aren't atypical; as a percentage, our immigrant population is probably no different to that of most high-tech companies. Without immigrants there wouldn't be a Silicon Valley and people here wouldn't have the wealth and opportunity that they have. Immigrants built America, and continue doing so in Silicon Valley.
SI: You've made rather pointed statements about American education, particularly in the area of government funding for science instruction. How do you view the problem and how would you like to see it solved?
Rodgers: Big picture, divide education into two parts — K-12 and college. K-12 is a government-run monopoly. Colleges are part of the free market. There are government-run colleges, but there's enough variety and private education in colleges to generate real competition. ...
The reason why we lost half of the automobile market to Japan is a good analogy for our educational system — there was incompetent, arrogant management running General Motors, fighting with a hostile union that only cared about its own power base. Nobody worried about the customer, so the business vanished. ... We need vouchers — I cannot understand why people don't understand that, if the government gave parents back the money to spend in the free-market schools, like colleges, we'd be better off.
Now, the question is how the government should interact with universities. If the government is going to be involved, we should look back a couple of decades when, during the Cold War, we had DARPA.
True, we were trying to develop weapons technology, but weapons are a product of applied physics and physics is a product of basic research. Back then, government money to colleges went to research. You wouldn't give Stanford a nuclear bomb program, but you might have them investigate some aspects of nuclear physics. This money was beneficial and was managed independently from military or government control. The military didn't want to manage its budget based on political considerations but on need. This resulted in a relatively objective meritocracy to evaluate programs. ...
When DARPA changed to ARPA, politics poisoned the equation. Now they worry whether Ohio is getting as much money as California, and study politically and environmentally correct things — all the typical political BS.
SI: : The Justice Department has intervened in Microsoft, and may do the same with some Silicon Valley companies. What can we, as an industry, learn from Microsoft's travails?
Rodgers: Microsoft is another sorry chapter in the story of government intervention. In 1945, the Department of Justice broke up Alcoa Aluminum because they were taking market share. This was because they had an efficient production process, good quality, and were sufficiently ahead of the financial curve to use profits to build new plants to be ready when demand rose. The court's reasoning for declaring them a "monopoly" was that, as a company, they didn't have to anticipate demand and build capacity to fill it. According to the court, this was a conscious act performed to create a monopoly.
In terms of lines of code and functionality per dollar spent, Microsoft drives software prices worldwide to an all-time low. Then they're sued for creating a monopoly to raise prices. This is irrational! If you lower prices, like Microsoft, you're trying to drive your competition out of business by undercutting them. If you raise or maintain them and make a healthy profit, then you must be colluding with your competitors not to lower prices — catch-22!
SI: : You're one of the few leading chipmakers that owns fabs. How do you view the fabless model?
Rodgers: It's doubtless concentrating power in the fabs. TSMC and some others exert enormous control over many semiconductor companies. Last year they told us our prices were going up and that the new price is X, and X is not negotiable and, by the way, would you like some wafers? Translation — Would you like to stay in business? TSMC has been pretty blatant about its leading position, and this may cause other foundries to appear and compete for its 40% profit margin.
We use them and others, as well as make our own silicon. To me, whether or not you are in the fab business is the same as whether or not you want to keep your business free from outside influences. The day we stop making silicon, Cypress Semiconductor will be looking for a new CEO.
SI: : What industry trends should we prepare for?
Rodgers: In addition to the shortage of college graduates, the proportion going into electrical engineering — within that shortage — is dropping. The human resources shortfall can only get worse.
Some old trends are dying out. Japan and Korea were scary for free-market capitalist companies, with their model of selling at a loss or not at a profit, and not funding future growth. They used low-rate loans — essentially free capital — to hammer the industry. This model collapsed the Japanese and Korean economies. It's still practiced in Japan and they're still hurting. That model's spectacular failure will keep most from trying it again. There'll be exceptions — Malaysia and other countries in the Far East will take a shot at it — but I don't believe the trend will continue.
Moore's Law won't go away, but the SIA Roadmap's barriers are made out of papier mache. I remember when 2.0 µ m was considered a fundamental physical limit due to concerns over "hot electron injection," where the transistor supposedly got so hot that electrons zoomed out of the gate oxide, damaging it. That was an order of magnitude ago in gate length. We're now below 0.20 µ m and headed beyond 0.10 µ m.
Another big trend is software as value-add. Most everything we make today has software in it. Cypress has dozens of software engineers working on software we sell to customers to enable our products. The ratio of software designers to hardware designers will surpass unity in the near future.
SI: People are talking about another downturn. What should be done to ameliorate this sine wave the industry is on?
Rodgers: Nothing. The last thing I want is someone like Al Gore meddling with our market. The market will handle the cycles. One development will go far in reducing them — the partnering between vendors and customers. Customers used to have an SOB running procurement and you'd have a bigger one doing qualification. You'd fight over price, customers would shift from one vendor to another to save a nickel, and they would overbook. There was hostility between system houses and semiconductor manufacturers, as well as manufacturers and equipment vendors.
Now, our strategic accounts have Cypress inventory on their sites. We have a pre-agreed pricing mechanism both sides can live with, which tracks the market. They no longer try to screw us out of a quarter to make a quarter because they know that if they make us uncompetitive, it will cascade on them. They don't jump from one vendor to another because the product proprietary software and technology content are so high they'd have to start from scratch to get an identical deal someplace else. We automatically enter orders into our database, which they adjust on a daily basis, and we watch the profile and move to make what they want — it's a virtual integration. Eventually these oscillations will get better because everyone is working to make them so.
— Alexander E. Braun
