In case you missed it, there was a fascinating thread on Olivier Coudert’s Blog over the last month or so. Olivier struck a chord with his post entitled “Why Service Companies Will Eat Up EDA“, which put forward the case that the design services industry would soon emerge as the primary implementers of complex SoC designs.
Olivier’s vision is that SoC design will become a commodity to systems companies whose differentiation would be in their system-level design and algorithms. Implementation from RTL or netlist to GDSII would be performed by these design services companies that know the methodologies, tools, and target technologies better. These design services companies could also leverage economies of scale by sharing licenses among multiple projects and selecting from a large pool of accessible IP. Olivier pointed out that even today, the design services industry is twice the size of EDA and that is likely to grow.
According to some of the comments, competition to pure design services companies could come from the EDA companies or the foundries. The large EDA companies have the advantage of free tools at their disposal and so can gain a price advantage over independent companies. Their intimate knowledge of the tools and flows and access to tools experts in R&D also enhance their attractiveness. Meanwhile, foundries could step into this space as well, amortizing the cost of design with the revenue from production volume. UMC has been doing this for quite some time with Faraday Technology Corporation and TSMC with Global Unichip.
There is a third set of players that is worth watching as well. Last week, OpenSilicon, a “fabless ASIC vendor” acquired Silicon Logic Engineering, a “front-end design house”. Together, they can potentially offer “spec to parts” services, a much broader offering than most other players.
As a coincidence, I had a conversation this week with someone whose been in the semiconductor biz for close to 3 decades. He’s got an idea, and also some technology already developed, that would enable companies to design SoCs at a very high-level and very quickly. Although the key to this capability would be some nifty tool, perhaps he’s better off using the capability to do design services rather than enter the EDA space. It seems Olivier would give him that advice as well.
What do I think?
When the automobile first came out, you had to be a mechanic to own one. There were no auto mechanic shops or service centers. If something went wrong or if you wanted to tune the performance, you had to be able to lift the hood, take out your tools chest and get to work. Over time, the auto service industry emerged until today when only the true automobile enthusiast or race car driver does much more than change his own oil. For almost all of us, we get much better service at a lower price than if we did the work ourselves.
Of course, chip design is not the same as car servicing, but I think the same principles apply. For the majority of the market, they will be able to get better results with less cost using a design services company. After all, how much do you really need to optimize your design and silicon when silicon is so abundant. For the rest that absolutely need the highest performance and only trust themselves, they will continue to buy the tools and do the work. What will be the split and how long it will take to get there is anybody’s guess. And when we get there, will there be enough business to support an EDA industry? We’ll see.
harry the ASIC guy