Problems Ahead For EDA

You may have discovered that the Semiconductor Engineering Knowledge Center (KC) provides various ways in which data can be viewed. One way is to see what events happened in a given year. During the 1990s, company activity in terms of new startups and acquisitions reached a peak, and in 1997 there were at least 29 startups that the KC contains and 25 companies acquired (let us know if there were others that we are not aware of).

Compare that to 2014 where there was a single new startup and 15 acquisitions. This is not indicative of a healthy industry. Semiconductor Engineering sat down with , chief technology officer at Carbon Design Systems; , chief executive officer for ; Randy Smith, vice president of marketing for Sonics; and Michel Courtoy, vice president of marketing and business development for Kilopass Technology, to discuss the health of EDA. What follows are excerpts of that conversation.

SE: Is the EDA industry healthy and does it rely on startups for innovation?

Courtoy: EDA is healthy. It serves its customer base. We have to separate the technical delivery and solutions that EDA provides from the investment side. On the latter, the investment side, it is dry.

Smith: Startups in the ’80s and ’90s required about $10 million, and that is still the case. But the fund that those monies would come from are now looking at putting at least $15 million to work, and if you seek balance and want to have two VCs, that is $30 million. Nobody in EDA needs that much. If you look at the exits, we are looking at companies that have been around for a dozen or more years. That is way too long a time for product adoption to interest VCs. They can put their money into a dot-com and be in and out in three to five years. The problem is that it is not attractive to the investors. That doesn’t mean that there cannot be startups, but they require more bootstrapping, using angels and people willing to take more chances with their own economic situation.

Neifert: EDA is a difficult market to break into. The industry has been around $4B for as long as I have been in it and is not growing much. When it does grow, it happens by bringing in other stuff. Synopsys gets a lot of money from IP these days and their latest acquisition was Coverity, which is not even in the space. [Since this panel, Synopsys also acquired Codenomicon.] They are attempting to go after more money by growing into other industries. Startups are a reflection of that. It is also a reflection of chip design in general. There aren’t as many chip starts taking place, and many of those are around just a few SoC architectures.

Davidmann: The goal of the electronic design automation industry is to help people build electronic products. If we look at the people around the table, we have an EDA company, two IP companies and we (Imperas) are really a software company. The industry started off with computer-aided design, and then over time started trying to build automation. We developed technology to help make designers more efficient. The industry trajectory is moving away from design automation. I got involved in an IP company in 1997 because I saw the opportunity to make designers more productive by bringing in pre-verified IP. The best way to improve verification is to bring in something that has already been verified. Now you have a different problem with integrating it. If we consider verification, there has been a dramatic change in this over the past 15 years. It is changing again and will begin to morph into the software space. Ten years ago, I started moving out of traditional EDA, then into IP, then verification and most recently into software. EDA has to find ways to use the skills we have and the methodologies that have been developed and move that into software.

Courtoy: For EDA to be successful, the customer base needs to grow. If you look at the traditional customer base, which is primarily the semiconductor companies doing big chips, it is consolidating quickly. Over the past year there have been about 10 big acquisitions, Qualcomm and CSR, Freescale and NXP etc. For EDA to grow, we cannot focus on that tip of the pyramid. We have to build it up, and the is a perfect example. There will be lots of designs, but those people may require different sets of tools. They don’t want to do 14nm finFETs, they don’t want double patterning, but they are putting together a complex system that contains software.

Davidmann: I am excited about the way the EDA industry is moving forward and the way it is evolving. It is no longer about place and route, or SPICE. It is moving and evolving into the areas where the biggest challenges are. This whole idea of IoT, a tiny device that sits out in the field, talks to the web – this is a complex system problem and yes there are bits of electronics, lots of computers and lots of software. EDA is moving to the point where it will encompass software and there is a lot happening in that space.

Courtoy: Consider the Fitbit. Most of the semiconductor content is off-the-shelf, manufactured in 55nm or 65nm. It could probably be designed using free software from Berkeley. The industry has to evolve and grow it at the base of the pyramid. The top is well served by the Big Three. There has actually been an uptick in venture funds for semiconductors. Venture investment for 2014, compared to 2013, has doubled for semiconductor, although it is still small. Kleiner Perkins, the biggest and best known VC in the world, put $15M into Ambiq. The last time they made an investment in semiconductor was a long time ago. This was all about low power.

Smith: The difference between IP and EDA doesn’t matter much anymore. It is all about design acceleration. The way they are monetized is different, but if you look at ARM or Cadence/Tensilica – we deliver as much tools as we do IP. They are all configurable and we have to build test environments for people. EDAC now contains companies such as ARM. We have to deliver tools along with IP to make it work and a lot of IP now has to be delivered with software stacks as well. It all starts to blur together. The question really comes down to – where is innovation going to happen? Where can we make a difference? The physical design flow has been commoditized at this point. It is hard to come up with anything that will make a big enough difference that you can add value. For system-level design there is still plenty of room, but you will have to come up with some smart ideas. This is because there is no standard flow. There are many ways that people build their chips. Do they use ? How do they define a platform? At what level of abstraction do they define the platform? There are still many differences. Where there is no standardization, there is lots of room for innovation. After they have been standardized, they get progressively improved but it is hard for anyone to make a leapfrog breakthrough.

In part 2 of this roundtable, the discussion examines some of the other difficulties associated with starting new EDA companies.