Recently, the ESD Alliance organized the annual CEO Outlook panel with Simon, Wally, Grant, and Dean, moderated by Ed. OK, maybe first names aren't quite enough in EDA (is there another Wally in EDA?) but they are: Simon Segars, CEO of Arm (which is a Softbank company but doesn't go out of its way to advertise it) Wally Rhines, CEO of Mentor-a-Siemens-company (actually, I'm not sure if his title is still officially CEO, or whether he is now something like Donaudampfschiffahrtsgesellschaftskapitän) Grant Pierce, CEO of Sonics Dean Drako, CEO of IC Manage Ed Sperling, Editor-in-chief of Semiconductor Engineering Bob Smith welcomed us and pointed out two things that the ESD Alliance has organized: First, a half-day workshop on Digital Marketing. They are flying in OneSpin's Nicolas Athanasopoulos, Head of Digital Strategy. He comes from outside EDA, having worked with Fiat/Chrysler, Ubisoft and Barclays. So there is an opportunity to learn how other people are leveraging social media and so on. Second, there is an "infrastructure alley" at DAC. Microsoft, Amazon, and Google will all be in the alley. So by "infrastructure" think "infrastructure as a service" which is a trendy way to say cloud. Ed started with a softball question to let each of the CEOs give a vision for what they see in the next few years. In each case, he didn't let them get away with just a view through rose-tinted glasses, he asked them what negatives they saw. They started from the right in the above picture (for once the CEOs did something that I wished, as a blogger, all panels would do—they sat under their names in the right order). Simon says it is an exciting time. Everything was focused on mobile, but while that's big it's flattened. Coming next are big designs in the cloud and lots of innovation at the edge. Cloud, in particular, is exploding and he sees no end in sight to growth. It is also changing who is doing the leading-edge designs, with big system companies wanting to build their own chips: Apple, Google, Amazon and so on. The big negative Simon sees is the cost of design. "The cost of a 7nm design is not for the faint-hearted." Managing compexity is going up and up. "We need some good EDA tools." Wally pointed out that Simon buys a lot of tools so he should know ("a lot more than I used to" Simon pointed out). Wally, being Wally, had data. He sees now as one of those times when things come together. We had 4 years of 2% semiconductor growth but last year it was 20%. Yes, half was memory pricing, but if you take that out the rest still grew 9.5%, and the forecast for this year is about the same. Wally also pointed to the fact that new people like Amazon, Google and Facebook are entering the market doing their own chips. Furthermore, they get big fast. Wally has been doing a little investigation about consolidation in markets (not just EDA and semiconductor) and gave me a sneak preview of some of what will be in his new keynote (he's going to try it out off-Broadway in Israel soon). But it is clear that a lot of companies are doing autonomous vehicles, and many will not be around in 10 years, they will either fail or get consolidated like the early 20th-century automobile companies got rolled up into three. But it rarely gets stable, and once an industry is down to three, it blows up again (once GM, Chrysler and Ford were it...suddenly the Japanese and the Europeans arrived in the US, never mind Tesla decades later). Wally sees "this AI thing" as driving a big discontinuity. VC-funded semiconductor startups had been in decline for 10-15 years, but it has resumed with $900M last quarter. "Before that everyone had to depend on Lip-Bu." If that remark doesn't make any sense, then I should point out that Lip-Bu is also Chairman of Walden International and has been one of the few VCs who would even consider investing in fabless semiconductor startups. For EDA, the ESD Alliance market statistics had just come out. The quarter grew 10.5% frrom 4th quarter 2016, and 9.5% for the whole year. It was very evenly distributed across the categories the report uses: CAE, PCB, IC Physical. IP is 14% growth, the rest is 7-8%. The outlook is rosy. To Ed's downside followup, Wally said "lawyers". Nasty things are going on in trade negotiations, which don't seem to affect semiconductor since it would hurt everone too much, but might affect the whole economy. Grant said he's going to change the name of his Alexa to Wally, so he can say "Hey Wally, tell me something I don't know." He said that today feels like earlier in his career when he was in the embryonic MIPS (which, of course, just won the Turing Award for its founder John Hennessey). It felt then that it would change how people wrote software, and now everything is RISC. If you fast forward to toaday, then it's all congitive computing, AI, machine learning. We have big datasets in the cloud and resources, so you can develop without needing compute resources, just AWS. He thinks it will be exciting as an IP provider, since there will be different specialized architectures running different neural nets everywhere. "It's really exciting when we read the newspaper," Grant said. I have to point out that I don't think many of the people designing those neural nets are reading newspapers, not on paper anyway. The negative is that these people are software developers or architects, not chip designers. But that makes it a great opportunity. The level of support that will be demanded will be high and it is tough for smaller companies like Sonics to address that. Dean said that he decided to pontificate (his word) on the EDA industry, not semi in general. His big point was: I think one thing that will affect us all is the movement towards cloud computing. Something that hasn’t happened yet, but people are starting programs to do it. SAAS will be $75B by 2021, and Infrastructure as a service (IAS) will be $72B by 2021. Infrastructure ally at DAC has Amazon, Google, and Microsoft touting IAS, and the move to the cloud in a big way. The buzz about cloud computing, Dean said, started 10 years ago, but this time it seems to be real. Barracuda Networks (where Dean was the founder and CEO) moved spam filtering to the cloud, so he has some experience. He's obviously read all the advice about social media that people love listicles, so he had 5 points about EDA that he thinks will be relevant: EDA in the cloud is on the cusp and there will be tremendous growth. Customers want peak computing (known as "burst to the cloud") for regression tests, simulation runs and so on. They wen to exit the compute farm business, since they are not experts and Amazon and Google are cheaper. One of Dean's customers reckoned that it was roughly an order of magnitude more expensive than AWS to have their compute farms in-house, which is hard for a CFO to ignore. Security has been a big concern, but all those same customers have moved their ERP and email and everything else into the cloud already. Their IT departments have done due diligence on the cloud companies to leap through the security hoop and that box is now checked. Business models for EDA will change, like when we did the transition from permanent licenses to TBL. Amazon & Co operate on a per-minute or per-second basis. "It's going to be fun since there's lots of worry on the EDA side that they may lose revenue." But Dean thinks it will go up. He thinks users will just run more simulations in parallel. IC Manage uses the cloud extensively, and their bill just keeps going up. Cloud model, for the foreseeable future, will be hybrid. Computer farms are not going to get thrown out, and there is a huge investment in scripts, and giant makefiles and so on. Those scripts pretty much require shared storage and are designed to run like that, whereas cloud-based systems don't really have a good shared-storage model. For EDA developers, software development for cloud-based systems is significantly different. The databases used, the methodology for spinning up machines and spinning them down, the scalability. These are very different, with hurdles to be overcome, but it is a big opportunity. To be continued...the rest of the evening became more of a discussion. I'll cover that in tomorrow's post. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
↧