Executive Insight: Aart de Geus

SE: What worries you most?

De Geus: Everything I do is with high intensity, and what is of super high intensity right now—and there are challenges and opportunities in it—is that we have the confluence of some very big changes right now happening at the same time. On the technology side, there are multiple intersections. One is the intersection of another 10 years of Moore’s Law—finFETs and maybe FD-SOI. On the economic side the cost per transistor will continue going down, but not as fast as before. But the number of transistors and the complexity of those is substantial.

SE: Just so we’re on the same page, what node is a decade from now?

De Geus: 7nm. We’re at 28nm now for production design. One sees roughly a decade where the people who are knowledgeable and work in it can see it. From the implementation creation point of view, the chips will be possible. A whole new generation of implementation tools is being introduced as we speak. The second aspect to that, however, is that the complexity of what can be created is amazing. Now truly everything is one or more supercomputers. The verification challenge is growing much more rapidly than the Moore’s Law challenge, and that is amplified by the fact that everything is loaded with software. So the complexity is not just hardware. It’s hardware and software, and they’re no longer independent. They’re highly interdependent. That means massive emphasis and development on the verification side of the hardware, the hardware and the software, and the software.

SE: Is it really separation of physical devices, or are we moving into a mesh network for data?

De Geus: If you look at a slightly higher level of abstraction, which is the telephone system or the Internet, it’s already a massive mesh. Every so often some bad player does havoc in that mesh, and the bad players are the bad guys. As we move more to the application domain, it’s not only fight the challenge of complexity up from physics. There are proactive bad players here whose objective is to hack something. It’s not just battling complexity. It’s battling complexity with bad players.

SE: What does this mean for semiconductor design?

De Geus: If you look at the application domain, which will blossom over the next 15 years—many people call it the Internet of Things, I prefer to call it Smart Everything—these become another 10 to 100X more entry points for a massive mesh of things that are connected.

SE: So we had a skeleton outline and filled in the pieces?

De Geus: Yes, and we added five more dimensions, some of which include your skin. Jules Verne in the 1890s said that cannons ultimately will break shields. Bullets will ultimately pierce steel. You do better steel and new bullets will be created. That’s exactly what hackers and viruses are all about. It’s finding an entry point. They will continually find it and we will continually do battle against it. But now there are so many places. What in hardware is call verification in software is called quality test and security.

SE: What does Coverity bring to Synopsys in this area?

De Geus: It’s a very strategic move for us. On one hand it’s completely adjacent and on the other it’s completely unique. We’ve looked for years for this kind of opportunity. It’s completely adjacent because half their business is in embedded software. On the other hand, if you go to applications, we’re now selling to banks and oil companies and health care companies and trading platforms and shipping companies. These are all people who rely on software for decision-making and execution.

SE: How much of an opportunity is there for current EDA tools in those markets?

De Geus: The hope is that as these markets become more sophisticated—more software, higher penalties per bug—the value of applying the same type of formal and static techniques we use in hardware will apply to the software world. The proof for that was the $75 million price we paid. This is not a fly-by-night startup. It has repeat and growing customers. It’s still to be proven, but for many years the price of a hardware bug was far higher than software. With software you do a patch. With hardware you do a mask set. But patch on top of patch and then plugging in some USB—it becomes very expensive. Everything on your home computer worked great until you plugged in that thing. And now you don’t even know which thing it is because it was when you plugged in two things together. That is why Coverity is so important to us.

SE: Is the impetus for this based on the fact that EDA can only grow so much, particularly with startups disappearing?

De Geus: Any company at any point in time tries to augment their business. We’ve followed the strategy of trying to be better than the other guys and grow with the product we have, and we’ve been explicit about following a strategy of adjacency. Within the core EDA we’re mostly there, although we may want to add something in the future.

SE: You’ve been in photonics for awhile. Where is that?

De Geus: Photonics is a little bit down the street rather than right next door.

SE: So that may be more lucrative when 2.5D becomes mainstream?

De Geus: Exactly. Many of things are more obvious when it’s at your door. The Coverity is one where part of it is adjacent today, but it’s much bigger in terms of the number of customers. This is not comparable to any other field we’re in.

SE: Fabless chipmakers have been a major source of tooling for EDA, but they’re under serious pressure these days compared with systems companies. Who wins and what does that mean for Synopsys?

De Geus: The winners are very agile to find combinations of things to have differentiation in the market, and that differentiation is not always predictable. At the end of the day it comes down to whether you’re in the right market at the right time with something that is both differentiated and cost-effective. Now there are rumors that new companies are building chips. These are deep-pocketed, highly motivated companies.

SE: We’re now measuring companies in Internet time and they rise and fall quickly.

De Geus: This is an exponential age, meaning for the same unit of time the difference is exponential. That has big impact on the present leaders in semiconductors, because they’ve had very big exponentials. Mobility has led to more phones than have ever been produced.

SE: But aren’t we also heading to bigger markets sliced into smaller pieces?

De Geus: These are normal cycles in a market. You have a new idea, and then there are two ideas—someone likes it in blue and someone else likes it in green. In mobility, we have the wave of cheaper smart phones, and right away instead of two or three providers it’s 10. And inside it’s a different brand than what’s on the outside. They are very active, fast-moving markets. And then from differentiation it goes to cost.

SE: What does that mean for your ecosystem and customer base? Does it stay the same or change quickly and you just hope that in aggregate it stays at the same level or grows?

De Geus: It has continually evolved. In semiconductors we’ve seen a lot of M&A. Companies are either trying to reach for critical mass or economic efficiency. Two companies that are sort of doing the same thing can do it more efficiently if they get together. Or they commandeer a bigger piece of the market and therefore be more attractive to customers because of lower risk. That happens constantly. Every few years, if you look at the makeup of the top companies, it changes 20% to 25%.

SE: It seems to be more pronounced lately.

De Geus: That is because it’s not only an economic challenge. There’s also a technical challenge.

SE: Companies can fail just by missing one window, right?

De Geus: And they can succeed by the same token. But in some ways, that is the characteristic of high-tech. The industry lives on exponential, and the most profound exponential has been Moore’s Law. That continues. The next incarnation is smart everything.

SE: Why smart everything instead of the Internet of Things?

De Geus: It’s a given that everything will be connected. But if you’re connected and you have some actions to take, do you take them here or on the central computer. That is completely changing. If it’s just sensing data and doing some manipulation, that’s the old way. The new way is there is sufficient compute power to locally decide many things. If you put cameras on the building to determine who is allowed to come in, we can certainly send pictures to the main computer to do facial recognition and then say yes or no. But we also can do facial recognition in the local camera. The storage isn’t that large and the algorithms are fast enough and the compute power is there. And then you’re not dependent on the link, and you don’t have to send a boatload of data up and down the link. That’s where a lot of this is going to go—as much distributed computation as possible. That is why everything is going to be smart one way or another. There was a big push for artificial intelligence, and that’s coming back in some form. It’s fundamentally inference from many pieces of data.

SE: That’s been talked about for years, but without success because it required too much memory and processing power. We seem to have overcome that limitation.

De Geus: Yes. The starting point is now. It’s not that big a step to also add, ‘Here are 20 things that if we get them correct, we can determine your mood.’ These are understood in many fields. The question is whether you can model them. If you can, they’ll be a lot better understood. Now you can create feedback loops.

SE: This is like a marketing nightmare for consumers, isn’t it?

De Geus: Yes, but if you’re a good marketer you will not tell your clients what you know. You will happen to have exactly what they want at the right moment. They will know more than we know about ourselves at some point in time. It’s not that long ago that the Dick Tracy watch was an impossible dream.

SE: That would have far-reaching impacts.

De Geus: In 10 or 15 years, anything that costs more than $10 will have silicon smarts. You’ll be able to tell if a bottle of wine was shaken, whether it stayed in the trunk of some hot car for a week—that will allow you to enough margin to have an extra 50 cents of smarts. You see the opening all of these opportunities, and the question for Synopsys is how we balance our investments in all of those. We are really one of the ultimate deep high-tech companies. Arguably, we build among the most sophisticated software in the world today, and it’s all in this boiler room of this incredible infrastructure. It’s the semiconductors and the systems and the software and so on. We’re almost at the atomic level, then we have the chip, the system around that, and we have the software.

SE: What are the new technologies you see coming that are interesting? FinFETs, nanowires and maybe biotechnology.

De Geus: FinFETs, nanowires and biological components are three very different domains. It’s like comparing horses, cars and teleportation. There are still benefits to gallium arsenide, but it will never catch up to silicon because the investment has been so deep. These new areas, such as biological systems, have a different outlook. There are many other directions to take before that. And there are some variations. Organic materials—carbon nanowires—will be interesting some day. But organic materials in displays are very interesting because you can fold them.