The founders of EDA are retiring, and perhaps it’s time that EDA headed off in a different direction.
The EDA industry is about 50 years old, and I see the people responsible for its first generation setting their LinkedIn employment status to ‘retired, at home’ almost daily. I, for one, have a foot in that camp, but reporting/writing is different than having a full-time job because I can control the time commitment. We have seen many serial entrepreneurs who have created several successful companies over time, and they have performed an invaluable service to the industry. Unfortunately, their time is up.
Long ago, VCs realized that EDA companies do not produce the massive gains compared to other software or Internet companies. Funding almost dried up, except for companies championed by people with a track record. That reduced risk, and thus did not demand the same level of return. But it also significantly reduced the number of people who stood a chance of getting a startup going. Even then, they often had to pivot a few times before they gained traction.
In a recent roundtable, I was accused of being stuck in the last century when I suggested that AI agents would benefit from better interfaces and standards associated with the data that comes out of EDA tools. This was perceived as being a minor inconvenience and more importantly — sequential thinking. This may be true, but my motivation for the suggestion was that if AI is forced to work on unreliable data that potentially contains unnecessary interpretation, then any solution is hobbled from the start. Data cleanliness is important, and that starts at the source. Getting rid of ambiguity has got to be a step in the right direction if reliable results are to be expected.
The irony is that the people who made the accusation have been rattling around in the industry for almost as long as I have. I know I am somewhat set in my ways, and this is why it is always refreshing to hear from people who have not been doing this for 50 years and who are ready to question every aspect of what we do. One of the panelists said, ‘I love the idea because it means there are people thinking like an engineer should.’ How refreshing.
While great strides can be made by new developments, there is a lot to be gained from stepping back and correcting the mistakes that have been there for a long time. Many years ago, I was given responsibility for an acquired tool that had an enormous bug list, many of which had been there for a long time. My manager asked when those bugs were to be fixed and how I was going to prioritize them. My first step was to take a look at the overall code base, enforce some coding discipline, and to rewrite some of the foundational pieces of code in which I found errors or that might lead to performance issues. It took six months, and at the end of it, we re-ran all the bug cases, and 60% of them were gone. We also saw a 30% performance increase in the whole product. The point is that if you only ever look forward, without sometimes looking in the rear-view mirror, you sometimes do not get the whole picture.
I miss the constant feed of information that comes from startup companies. They are working on new and exciting problems that are advancing the state of the art. As soon as they get acquired, the information stream stops dead in its tracks. I can understand why. It quickly becomes wrapped in a bigger story, or product, and that tends to only be shared with their primary customers. We in the media are suddenly left out in the cold, apart from when we can extract specific pieces of information deemed suitable for the general public.
It is not only the companies that are disappearing, but the people as well. Startup CEOs have much broader industry knowledge than just their current company. In many cases, they are experts in large parts of the design or verification flow and cherry-pick the biggest problem areas for their next startup. They continue to keep their finger on the pulse of the rest of the industry because as soon as they are acquired they are probably looking forward to their next startup. That makes them highly valuable information sources, because they need visibility for the company and not just small advancements in their current tool. Regrettably, their numbers are dwindling, and I fear we have seen the last company from many of them.
But did those startups contribute to the linear thinking of the whole industry? Is it time for new people with fresh ideas to define the direction of the future? Both of the big independent EDA companies have seen a change at the top. Lots of fresh engineers are entering the ranks – hopefully with their youthful spirit and determined to make a change, to make a difference, to think like engineers. I hope those that take over from the old guard allow them to do that.
Traditionally, EDA has been about creating data that is used by humans to make decisions. This applies to design, at all levels of abstraction, verification and implementation. The new era is about how to extract knowledge from data with as much automation as possible, leaving humans to make decisions about what to build rather than how to build something. Time will tell how effective this transition will be, and where to best draw the lines between automation and human expertise.
Thoughtful piece Brian. Having been in EDA for the better part of 35 years with some pockets of time out doing other things, I have seen it from both ends of the spectrum and a little in the middle. Some great years with the #1, with frequent 2 for 1 stock splits; some fraught years with one that got into Chapter 11 difficulty and got sucked up for mere cents on the dollar; and some with start-ups. EDA is hard and EDA start-ups are even harder if you don’t get merged by one of the ‘Bigs’ in a few years, the limit of investor patience.
Many of my contemporaries have or are headed off into the sunset and increasingly the feeling left is one of past great war stories and the good old days. The industry changed a lot though and lost its way as the corporates increased their cartel like stranglehold on their pie and minds became ever more closed to new ways and leading new thinking.
Some might say this is evidence of the industry maturing and becoming the grown-ups in the room but I would demur in that view. These days, EDA vendors have to vaunt their AI credentials by being AI everything and before that signalling their cloud commitments, however, this was latterly a following paradigm, never leading that path like it once did.
EDA always was and always will at be the base of an upturned pyramid spinning top, a $20bn industry supporting $700bn semiconductor leviathan. I showed such a graphic many times over many years to investors and industry alumni alike to illustrate the vitality and importance that the EDA industry must keep spinning loud and proud, lest the whole thing begin to stall and wobble. Metaphorically, like our dear planet Mother Earth, its ability to spin is measurably slowing.