EDA is a very connected community. If you work in EDA, chances are all the people you know work in EDA. Every social gathering is exclusively of EDA folks and the discussions that start in the evening on varied topics of movies and sports by night reduce to a familiar one - the decaying state of EDA.
This has been the case for years, everyone has been talking about how 'EDA is not what it used to be' forever. Nothing wrong with that, grass is always greener on the other side, and a periodic review of your profession can only be considered healthy. But these worries over the state of EDA is no longer some stray honk at a red light that you can ignore and keep driving, it's more like getting rear ended by a pickup truck.
The last EDA vendor to go public was Magma. That was way back in 2001! A lot of bytes have flown through the net since then, but the state of EDA at best can be considered stagnant from that point onwards. The only significant product since then is perhaps the rapid prototyping tools. Not that attempts were not made to try out new things. The routing guys tried out X architecture, in synthesis people tried to raise the abstraction with various High Level Synthesis tools, verification folks dabbled in formal technologies. But nothing gained any traction beyond selling the CAD teams at semiconductor companies a few evaluation copies (okay, may be a one or two licences to the design teams but no more).
A significant indicator of the health of the industry is the annual Design Automation Conference. An interesting paper by Alberto Sangiovanni-Vincentelli summarizes the history of DAC till 2003. The graph on the second page continued to drop precipitously and by 2009 the number of attendees reduced to around 3000. At the height of things around 1999, DAC used to be a real party. Busty blondes stood in front of booths explaining "physical synthesis", giveaways meant real gizmos of the day and I used to collect my years supply of T-shirts there. Now all you get to see is engineers with droopy eyes sitting behind deserted stalls with a bowl of candies.
Profs in academia seem to have judged the situation much better than the engineers in the field. Veterans in the field, like Randy Bryant ( inventor of BDDs), have moved from EDA to "the type of computing systems used for Internet search". Most working in the field now hedge their bets by maintaining parallel interests. Venture funding for EDA startups have dropped from the 2004 "high" of around $100 million to practically zero by 2010. In comparison Internet startups continue to attract over a billion dollars per quarter in funding. Within ten years of the Internet Bubble 1.0, people find enough reasons to put millions in Twitter clones, which itself is a loss making entity with no real revenue model as of now. Still they don't want to put money in EDA anymore.
Stop! Stop! Stop! EDA serves electronic design, so as long as electronics is part of our life, EDA would flourish, right?
The answer is not very straightforward, it may even be a resounding No. The last few years have seen a slew of mega hit electronic products - smartphones, e-readers, Kinect. The revenues of semiconductor companies like Samsung, TI, Apple, ARM, Xilinx have soared. But the EDA revenues have remained frozen. The key to understanding the discrepancy is to track the number of ASIC design starts, which have steadily declined. For higher end designs, the ones beyond 65nm, the decline is even sharper. The demand for bleeding edge EDA tools is now minuscule, and the lower end design flows have stabilized. The hardware design process has evolved from building everything from scratch to platform based design, and cores like ARM have become ubiquitous. Unfortunately in this scheme of things, EDA tools are no longer the innovation bottlenecks and they are pushed to a maintenance mode role.
So what's implied here, is EDA going to die? Will all of us be on the streets, resume in hand? Of course not. Job security and job growth are two entirely different things. EDA will continue to be a vital piece of the IT ecosystem. But much like plumbing, while essential, it will be relegated to the status of a utility and the hectic days of the 90s will never return. Experts of the various tools can continue comfortably in their roles maintaining their products by making incremental changes.
But is the situation comfortable? A myth that adds to the current comfort is that EDA jobs are high paying. While that was true a decade ago, a quick glance at the salary surveys at sites like GlassDoor reveal that is certainly not the case any longer. The other myth is that EDA has above average caliber people who on an individual basis add more value to their company compared to the other software firms, and hence salaries will be high. This again doesn't add up. Microsoft generates a revenue of $70 billion with 89,000 employees, an average of $786K/employee. Google generates $30 billion with only 26,000 employees at an average of $1.15 million. In comparison Cadence's revenue of $935 million is from 4,600 employees, an average of $203K. The average for Magma comes around $197K and Synopsys $211K. The productivity of the average EDA engineer is about one-forth in comparison to the established IT names. The downward pressure on salaries can only increase from this point onwards. With few new talent coming in and trickling away of the existing talent, the notion of high caliber people will also be difficult to maintain.
Where does all this lead to? One book that comes to mind in this context is "Who moved my cheese?" Note that the number of people who like the book and hate it are about evenly split. So this is nothing earthshaking and outright pedantic for the most part. But as it happens with anything that sells over ten million copies, there's something useful in it. My takeaway was that job growth trumps over job security. The only way to stay alive is to stay hungry.
