The news that Synopsys wants to bring together the R&D teams and EDA software products from its recent and proposed acquisitions of Magma, Ciranova and SpringSoft puts me in mind of an intriguing and market-changing perspective.
It looks like EDA is going back to where it came from – but with a twist. Synopsys has said it is making its consolidation move, in part, to offer higher levels of automation in custom tools. Custom in-house tools is where EDA started about 50 years ago.
Back in the 1960s and 1970s both chip making, and the EDA tools that enabled it, were the internal R&D activities of the large OEMs; companies like IBM, AT&T, Motorola, Siemens, Philips, Hitachi, NEC and so on.
With the creation of specialist chip firms like Intel and National Semiconductor, a first level of disaggregation occurred. For a long time that disaggregation was only partial and some OEMs kept on competing with the chip firms, gradually spinning out their chip operations while fabbed and then fabless chip companies sprung up like weeds. The fabless trend, starting in the 1980s, was another disaggregation that separated chip manufacture from design.
Meanwhile, and slightly behind this trend, EDA was separating from chip fabrication, giving rise to a series of EDA leaders to sell to those chip makers. Of course just as some OEMs never disaggregated, some chip companies continued to keep a lot of EDA tool development in-house seeing it as a point of differentiation and added value. Intel is perhaps the best known example. But most chip companies, especially the fabless startups, were happy to use third-party EDA software to get chips designed and get to market quickly.
More recently the fact that relatively fewer companies are prepared to go to the leading edge in IC design at each node has been a growing problem for the EDA companies. It means fewer design starts, fewer big ticket customers, less seats and potentially less revenue to pay for R&D.
And fewer EDA tool buyers means more powerful buyers, perhaps arguing that they don't want insights they are sharing with EDA tool vendors to immediately benefit their rivals. This is what I suspect is driving custom EDA implementation. Synopsys seems to be catching the trend, presumably hoping to turn a problem into an opportunity.
We are starting to see hints that fabless chip companies are thinking about putting money down to at least wrest some control of manufacturing back from the foundries. With fewer players in the field access to leading-edge manufacturing is starting to become an important point of differentiation in some markets.
Whether EDA R&D will ever go back in-house at the chip makers is another matter. It is a low enough cost item – essentially software – that it could, but it does need to be linked to very high capital cost manufacturing and things never quite go back to where they were.
We are living in the ecosystem era now but not all ecosystem members are created equal. Custom collaborations between narrower groups, down to those as small as one foundry, one chip developer and one EDA provider would represent the partial return of chip making to its roots.