SAN JOSE, Calif. — In the early 1990s, semiconductor companies that didn't own their own fabs were considered runts. Today at least five of them rank among the world's largest chip vendors. Some of the credit for the shift goes to the sector's trade group, which will celebrate its 20th anniversary on December 11.
“It was generally understood, when you get to $150 to $200 million in revenues, you'd have your own fab, but I said no,” said Robert Pepper, who in 1993 was chief executive of Level One, an up-and-coming startup in communications chips.
Pepper was deeply influenced by a 1964 article from the business guru Peter Drucker, who wrote that companies should invest only in activities that really add value, outsourcing everything else. “In retrospect, I see that's what the fabless semiconductor model is,” Pepper recently told us.
At an October 1993 financial conference, Pepper buttonholed Mike Hackworth, a friend and CEO of Cirrus Logic, a much-larger chip company that also had no fab of its own. “I said we need to have some kind of organization to push the fabless model.”
He was not alone. Bill O'Meara, CEO of C-Cube Microsystems, was pushing the same idea in private conversations, as was a hard-driving investor relations consultant named Jodi Shelton.
“All these guys went to same investors' conferences, and at every event they were asked, 'When are you going to have a fab?' Not why or if, but when — there was no business model for being fabless,” said Shelton.
Even though Level One and other companies were growing rapidly, Pepper said, “we didn't quite have the respect of the financial community, because as long as didn't have a fab, we were still considered pipsqueaks.”
In mid-February 1994, O'Meara called for a dinner meeting at the Plume Horse restaurant in Campbell, Calif. Shelton heard about the dinner. “I was calling people about the same idea,” Shelton said. “So I asked if I could come to dinner, and he said, 'Absolutely not — who are you?' “
Eight CEOs — and Shelton — showed up at the first meeting. One dinner turned into another. Attendance quickly doubled.
“We had four or five dinners, and no one could agree, so I decided to hire a lawyer to write bylaws and make something happen,” said Shelton.
“She took the bull by the horns,” Pepper said. “Somewhere in that process, she said this is a real startup and we need money, so she asked for a year's dues up front, and a number of companies pitched in before we even had a real mission statement. The first time I met her, I decided this lady could get anything done.”
Momentum for the group built quickly. “These companies were growing exceedingly rapidly, but people could not imagine you would be a billion-dollar company and still not have your own fab,” Pepper said. “That was not in anyone's lexicon.”
Shelton convinced Pepper to become the first chairman of what was initially called the Fabless Semiconductor Alliance (FSA). “No one says no to Jodi,” Pepper quipped.
One of the things we did was get financial people and the press to cover us in a positive way as a new paradigm, that being fabless was a stable model that had real legs to it. We had direct competitors who could work together, including a technical committee that worked on process qualifications. That added value to being a member.
According to Shelton, the audience included “venture capitalists, bankers, analysts, and potential partners,” which were essentially trying to build an ecosystem before the concept became fashionable. “We were giving presentations to anyone even remotely thinking about creating a foundry.”
The group's first big problem was figuring out a way to get a reliable, reasonable supply of wafers. At the time, there was only one fledgling foundry. Most of the fabless companies relied on big chip makers wanting to sell unused capacity.
In lean times, that was not a problem. The giants were eager to sell wafers, rather than let their expensive fab systems stand idle. But when the industry was flush with business, fabless companies were the first to get cut — a painful experience Pepper at Level One knew firsthand.
We had a customer for one of our first big products, and our source of supply dropped out because they got a big order for their own products. It almost sunk my company. That's just one reason I have grey hair, and I think a lot of other companies had experiences like that. You have to constantly be on the lookout for where you got your products made next because foundry work was not a major thrust from anyone.
Shelton recalled the situation: “Hackworth used to say the problem with fabless companies was their upside was always limited.”
Even worse, in boom years some chip makers held the fabless companies over a barrel, demanding a license to their proprietary technologies in return for wafers. “There was a lot of intellectual property given away during that time,” Shelton said.
The group had an early champion in Morris Chang. The semiconductor veteran from Texas Instruments returned to Taipei in 1987 to help found Taiwan Semiconductor Manufacturing Co. (TSMC), one of the first big dedicated foundries. Pepper recalled that news.
When Morris Chang announced he would build a wafer fab in Taiwan, the general consensus was Morris had lost his mind. I knew he was just way ahead of everyone else, who wondered why he would build a factory to make products for other people. Thank goodness he had this vision. The size of the plan and the stature he had in the industry made people think maybe he could be right.
Before the fabless group was four years old, Shelton was lobbying to give TSMC a seat on its board. “People were worried, but I knew it was the right decision, because it was such an integral relationship that both had to be successful.”
To read the rest of this article, visit EBN sister site EETimes.