DAC reflections: who's going to pay for all that EDA R/D?

Funding for research and development is always a problem for public companies. But for the EDA industry, where shrinking user bases, declining design starts, and the perception of dwindling differentiation have all eaten away at revenues, the problem has become acute. Shauh-Teh Juang, senior director of design infrastructure marketing at TSMC, worries about this.

"The EDA business model is questionable," Juang said during one of those conversations that tend to happen at DAC. "If you plot tape-outs, wafer consumption, and EDA revenue, EDA revenue tends to track tape-outs. They may need to change that. It’s probably true that as an industry, EDA right now is not generating enough earnings to fund another major round of research and development."

Of course no EDA executive would say that his company in particular faces an R/D crisis. But there is no question that there is a pinch. "There is still a lot of innovation taking place in EDA," counters Lip-Bu Tan, president and CEO of Cadence. "And we are still putting over $450 million into R/D here. But in the big picture, there’s never enough money for R/D."

Part of the problem is an historical shift, according to University of California Berkeley professor and Cadence chief technology advisor Alberto Sangiovanni-Vincentelli. "The big issue is, where did all the research money go? If you look at history, some of the really big ideas, such as synthesis of synchronous circuits and automated place-and-route algorithms, were funded by DARPA (the Defense Advanced Research Projects Agency.)" Dwindling of DARPA funding for non-military purposes under George Bush—though a subject of some dispute in political circles—undermined much of the research funding in areas such as fundamental algorithms. But that was just one problem. "The disappearance of the big corporate laboratories with broad charters—like IBM Watson Labs and Bell Labs—has also left a vacuum,"  Sangiovanni-Vincentelli continued. "Now there may be no industrial research lab that can produce a quantum leap."

Nor is the venture-capital community–which traditionally has paid the bills for EDA-industry R/D by funding start-ups, which the large companies later buy—going to help. Juang said that there is essentially no VC money for EDA start-ups now. "They cost too much, and they take too long to reach an exit," he explained. Better, I guess, to pour the funds down the rat-hole of yet another photovoltaic start-up, where at least the exit will come sooner.

That leaves universities, which are dependent on government and industry grants to fund research. As government money dwindles—and Federal stimulus funds appear to toss only a very small bone to research—the question turns just to industry. The good news is that some of the funding flowing into university research in EDA is coming not from EDA companies, but from their prospective customers. For example, Tan said that a substantial portion of the research money coming into Berkeley and MIT is now coming not from US EDA companies, but from Taiwan.

So who does the R/D? Tan, for one, likes what he sees in the European research consortium IMEC. IMEC combines some local government funding from its native Flanders with some EU funds and significant fees from research partners to perform pre-competitive research in microelectronics. A typical program might find companies who are fierce competitors in the market working together with each other and IMEC staff researchers in joint basic research on, say, EUV lithography. Tan said he would be very interested in any proposal for such an organization in the US.

In the meantime, that leaves the challenge of funding R/D up to industry. Tan suggested that customers—the companies who will benefit, some say disproportionately, from the R/D–are increasingly willing to share in the development costs. "Some customers are already working with us," he said. And there is low-hanging fruit in this orchard. "There are some major customers out there doing their own tool development. Does it make sense for us to do joint R/D with them? After all, even big companies need for their tools to fit into an ecosystem." It seems clear that Tan is willing to reach out in a way that would have been quite foreign to previous Cadence management.

In the long run, sharing R/D costs with customers could evolve beyond simple gestures such as the loan of researchers or sharing expenses. Tan suggests that the long-term solution, as R/D becomes increasingly expensive on a per-tape-out basis, might be for regular customers to share their revenue with EDA suppliers, rather than paying license fees up-front. Moving from one model to the other would either have to be done very gradually, or it would require a significant injection of capital into the EDA industry. But it seems an important alternative.

And it’s not altogether implausible—in fact, it’s already happening in another part of the industry, according to Juang. He said that TSMC had made some selective investments in small EDA companies when the foundry thought the little companies’ technology was essential to a new flow. "We have taken an interest in some, and moved them more toward a wafer-based revenue model," Juang said. But he emphasized that even though TSMC now has about 600 engineers working on EDA, they do not have the management infrastructure to take the place of the VC in nurturing small companies. "We don’t have the right expertise to become investors in EDA," he said.

EDA majors team with their customers. Small players team with foundries. Is there a pattern here? "I think the industry is moving toward vertical reintegration," Juang suggested. "But this time it will not be integration through ownership. It will be through relationships—starting maybe with technology relationships, but eventually relationships involving cash flow as well."