Twenty years ago today, I drove down from Illinois to Austin to start working at Motorola SPS in their design technology group. It was the start of a near twenty-year career in EDA (Electronic Design Automation) at Motorola, and then Freescale, where I’ve had the opportunity to work on and lead interesting technologies and projects and collaborate with some great and smart colleagues. This career has also given me the opportunity to work in the most challenging and important industry in high technology. After nearly 20 years in corporate EDA, I am now pursuing a startup in EDA.
The arc of my career traced some of the shifts in corporate CAD and vendor EDA. In the first 10 years, I was a tool developer, project lead, and manager of tools teams; my area had been in physical design, aka “backend tools”; our company had internal formal verification, floorplanning, signal integrity, and other tools (it still has a great internal Spice tool, Mica, and a layout synthesis tool, Cellerity). In the second decade, the company relied less on internal tools and we focussed more on flow development and integration of vendor tools; for example, I lead a team that was part of a “single vendor flow” migration effort, for example.
EDA has always been a difficult thing to explain to those not familiar with the industry. My own thinking is that when Rocket Scientists get together and try to tell themselves that landing a man on the moon isn’t that hard, they say: “Come on, it’s only rocket science, it’s not EDA!” After all, F=ma and gravity is a bit easier to model than the log-normal statistical distribution of the power effect resulting from the shift in threshold voltage due to atomic-level damage from NBTI. Try optimizing for that. (I haven’t read the paper but Prof Kahng of UCSD had a conference paper on “The Futility of Statistical Power Optimization” so maybe one shouldn’t bother.) No, you can’t explain that to non-EDA people without eyes glazing over, so it has been “I help support chip design teams with tools and flows.” Which is true, and helpful enough.
EDA as an industry has a challenge: Will it keep innovating, or will it get stodgy, boring, and still-capable-but-not-improving? It’s interesting that ISPD about 6 years ago started running contests in physical design tools and found – wow – that there was a lot of improvement and innovation left out there in academic research. What about the vendor tools? My observation, in looking at the innovation in other spaces – whether it’s gaming, social media, etc – is that we have some great but dated technology underlying the industry. Is this as good as it gets?
There is a challenge that the incumbent EDA vendors (Cadence, Synopsys, etc) will prefer status quo and incremental change to ‘disruptive’ change.
Now, a startup could challenge assumptions. Trends such as the rise of fab-based business model, greater IP reuse, are disaggregating the semiconductor industry, while cloud computing and other trends are disrupting business and service models for software. If you want to sum up in one word what the chip design industry needs, its agility. This provides an opening for new innovators in EDA, and my plan is to be one of them and in the process make for a better value proposition for design teams.
One fly in the ointment to an EDA startup idea is that startup attractions outside this industry are outshining EDA. Tomorrow, a ‘deal-a-day’ company called Groupon will IPO at a proposed value around $15 billion. In comparison, the total EDA industry is a $4 billion industry (revenues) and less market cap. With these comparisons, it’s no wonder the VC folks are not investing much in semiconductors or EDA these days, but are chasing the next Facebook/Zynga/Groupon.
There is irony there, in that EDA has made it all possible. No EDA means no great ARM-based chips out of NVidia and Freescale for iPad and Kindle etc. ready to consume all that cloud-based content on the web. Moore’s law has worked only because our tools have worked. Yes, EDA is the core, underlying, fundamental set of technologies that makes the entire high-technology world we live in possible. Gee, why didn’t I mention that to my non-EDA acquaintances?