SAN FRANCISCO — As the semiconductor industry lurches through its worst year ever, one question arises: Just how much consolidation will occur, given that smaller companies are struggling and larger ones are looking for an opportunity to grow by acquisition? According to T.J. Rodgers, chief executive officer of Cypress Semiconductor Corp. (San Jose, Calif.), it won't be much.
Here are excerpts from an interview Rodgers conducted recently with EE Times.
"If you go back to when Cypress was founded in '83, we were, by definition, last on the list of American semiconductor companies, around the high 50s or low 60s in terms of rankings. Now there are more than 100 such companies in the United States. That was the year that Intel and Advanced Micro Devices said, here comes consolidation, and it's going to be the Big Three like in Detroit and that's going to be it. But there are tons of them out there.
"There are three factors making the semiconductor industry more diverse, not less. Tremendous specialization is going on. When we started Cypress in '83, you could take the Intel catalog and read about every kind of IC that could be made in the industry, from memory to logic. Now, there are entire industries that have been spawned by one section of the old Intel catalog. No company can possess all the intellectual property to compete in these areas.
"In addition, you're not going to get the world's best CAM content-addressable memory guy to work in a division of a big company when he can start a company and make money and have control over his own destiny.
"Second is the advent of the fabless model. That overthrows the barrier-to-entry issue.
"Third, semiconductors have been encroaching into the systems world. So you ask yourself, what's the microprocessor? It's a very complicated computer — more complicated than anything on the planet. To make that, you have to be a systems expert. The knowledge to do that is incredible. As the semiconductor industry gets smarter, it is encroaching into areas that used to be dominated by systems companies.
"There's an old joke that was trotted out at a recent conference: What does it take to get into the computer business? A screwdriver and a credit card. The idea, of course, is that Intel is the computer. Intel was a great example of how one company can take over part of an industry, and it's happening all over.
"The next industry semiconductors can take a part of is datacom. It's already happening. Systems companies are saying 'here's a chunk of our system, make a board for us.' Take line cards. We'll have an OC-48 pipe coming in, and you go through various PHY layers, framing, demodulation and into a switch matrix. They're just saying 'make that line card; give me a proposal for that solution.' We're going to nibble on the edge and make the stuff that they used to make.
"That's why there can't be consolidation. A company capable of making a card for a Sonet router is different from a Transmeta. They're totally different animals. Silicon is working its way up into the food chain and not just as a special component.
"The semiconductor guys will take over on the systems side and not vice versa. The systems guys have cost structures that aren't competitive."
