Venture Panel Debates Opportunities, Exits in Fabless Semi Roadmap
A panel held Tuesday evening of four prominent venture capitalists and one dean of engineering took on the question of whether there was financially meaningful innovation left in the semiconductor industry.
The panel, sponsored by angel investor group Silicom Ventures, surveyed the current state of process technology and its implications for fabless companies, integrated device manufacturers, and especially for early-stage investors. The conclusions were far from predictable.
Laying out the scenario from a technology perspective, Stanford University dean of engineering Jim Plummer sketched in a process technology roadmap in which the technical problems keep getting harder, the cost of new designs keeps escalating, and no end to these trends is in sight. “Today a fabless company really only has two alternatives,” Plummer said. “Either they must use a lagging process that is affordable, or they must have really huge market volumes to amortize the design cost.”
Lip-Bu Tan, chairman of venture firm Walden International, echoed the cost concern. “In the U.S. today, it costs a new venture at least $15 million to reach tape-out,” he lamented. Tan went on to observe that such high initial investment was too steep for a venture investor unless there was a good probability of a really huge return. There was no value to the venture investor in putting $25 million into a company to achieve a $50 million exit.
But within that apparently dark scenario, all the panelists agreed that there were still deals to be found. “There are still opportunities for angel investors,” insisted Andy Rappaport, partner at August Capital. “Increasingly, they are in companies that find clever ways to push out the curves a little as process technology gets harder and harder. But they are there.”
Chris Moran, VP and general manager at Applied Ventures, agreed, pointing in particular to the fact that as process technology gets more restrictive and successive nodes yield fewer gains in raw performance, thinking outside the architectural box becomes a potentially valuable skill. “There are new applications coming in areas like multicore computing and parallel processing—often things we haven’t thought about yet,” he said.
John Oxaal general partner at the Sevin Rosen Funds, cited potential applications at the intersection of computing with entertainment and medical applications. “We are getting many more transistors to work with from advanced processes,” he said. “Video processing applications, for example, may very well be able to absorb all those new transistors.”
The panelists generally agreed that good investment opportunities would be tied not just to good basic technology but to command of a particular application area. And they warned that even for start-ups, working silicon was no longer enough—a product today was a complex assembly of silicon, software, design support and reference designs. Further, Rappaport added, most of the semiconductor companies he had invested in recently have had intimate relationships with their foundries—working together on process tweaks that would help differentiate their products. Such areas as new kinds of non-volatile memory cells were mentioned.
Tan pointed out that the opportunity in fabless semiconductor companies varied considerably by region. While in the U.S. the cost of getting to tape-out was high, in China, Tan said, he had companies that had taped out for less than $6 million, and it could be done for even less. “China is supposed to consume 21 percent of the world’s semiconductors now,” Tan observed. “But it has very few fabless semiconductor companies to serve that demand. There has to be opportunity there.” Tan was much less optimistic, though, about the opportunities to invest in expensive fabless start-ups in the U.S.
Addressing the question of whether nanotechnology would create a breakthrough that could change the playing field, Stanford’s Plummer sounded a pessimistic note. “The CMOS switching element has been highly refined,” he said. “There is nothing on the horizon right now that could replace it. I don’t mean that we don’t know enough about any of the nano-scale devices to be sure they could succeed. It’s that we now know enough about them to prove that they cannot be better than CMOS switches. If nanotechnology is going to create a breakthrough, it will require a complete reorganization of the way we create systems, in order to base them on something other than the logic switch.”
Another area in which the panel was pessimistic was EDA. It was pointed out that no one EDA company controlled enough market share in any one area to protect their margins, and so EDA had become locked into low-margin businesses. Further, the panelists agreed, new, specialized tools that could improve the margin situation were increasingly developed by either design teams, foundries or equipment vendors, not by the EDA industry. “It’s not that hard to develop a specialized tool that can help differentiate your chip,” Rappaport said. “There’s no barrier to keep a semiconductor company from doing it for themselves. In fact most of the chip companies I’ve invested in had some proprietary tools.”
The panel was asked if the increasing activity of private equity companies would mean that the only possible exit for a successful startup was to sell at bargain prices to a private equity firm. The venture capitalists on the panel said emphatically not. “The goal for our investments in the electronics industry is still a public offering,” Rappaport said. Successful IPOs are not common now, but they are still happening, and they are still the right goal.”