Siemens-Mentor Deal Retrospective

Tony Hemmelgarn, president and CEO of Siemens PLM Software and CEO of Mentor, a Siemens Business, sat down with Semiconductor Engineering to talk about the acquisition of Mentor Graphics, the shift toward more customized design, and where AI fits into the design picture.

SE: How does a company like Siemens see the EDA industry evolving?

Hemmelgarn: Part of the reason we bought Mentor Graphics was that we saw where this market was going—not just EDA, but how that fits into the rest of what we do with software. From the beginning, on the system side, it made a lot of sense. If you think about printed circuit boards, wiring systems, wiring harnesses, electrical systems, embedded software—all of those things fit very well into the discrete side of our business. And then there was a question on the integrated circuit side and what would that mean, and that has been quite surprising for a lot of people. We see a lot of touch points that become interesting between our organizations. We also see more and more systems companies getting into designing their own integrated circuits. The Chinese market is extremely important to us, although they don’t necessarily understand what the Siemens brand means in China. But they are all over us about what we do in integrated circuits, EDA and software. We’ve had a partnership in China for 100 years. It’s how we go to market together, how our adjacent products go together, and how this evolves going forward.

SE: You started out at a high level of abstraction on your business model, and now you’re dropping it down into the IC world?

Hemmelgarn: Within Siemens, we created an integration plan. We’re pretty good about going through that document and making sure we did everything we were asked to do. Wally [Rhines, chairman emeritus of Mentor] spoke at our capital market event. We did have levels of understanding about where we wanted to go, particularly in printed circuit boards and wire harnesses. We had been working together for many years before. But in the integrated circuit space, all the touch points weren’t clear. Eleven years ago we acquired Tecnomatix for manufacturing process planning and factory layout. If you go back 11 years ago and asked how we were going to work together, you would have never thought about some of the things we’re doing today in integrated manufacturing, to the point where we’re doing process-driven product design. As a person is designing a product, they see the factory and the plan for how they’re going to be building that, and they can make design changes and alterations well before they run into problems. It’s the shift left concept. You virtually prove it all out. We’re doing things we hadn’t thought about before. The same thing is true with integrated circuits. For example, we have a solution for product lifecycle management where we could gather requirements in the process of doing integrated circuit design. And then we started looking at IoT process, the fabrication process, the yield output and these types of things. We’ll look back 10 years from now and ask, ‘Why did we not see these things 10 years ago?’

SE: EDA has been striving to take what it does at the chip level and migrate that up to a much larger system. Is it working?

Hemmelgarn: We’re starting. The real test will be as we start leveraging further levels of integration. There are some obvious things we can do right away. For example, we could be talking to a customer and they don’t even realize we’re part of the same company. You’ll start seeing that change as you look at how we integrate some of these products. It’s everything from IoT to data management and requirements management.

SE: A lot of designs at the leading edge are very customized today. There isn’t design once, sell a billion units anymore. What does this mean for how you design solutions, particularly with regard to domain knowledge?

Hemmelgarn: When you think about customized solutions, such as autonomous driving, you see more and more OEMs saying they want to build an integrated circuit because they want to optimize for the features they need. Tesla just said their IC is four times faster for autonomous driving than anyone else’s because they optimize for that. We take it one step further. If you’re really going to prove your autonomous approach and create a circuit, how do you test and pre-validate pre-silicon? When we first got together we created a program called Pave 360, where we integrate the ability to look at the design of a circuit for test and validation of the circuit with the algorithms that are going to be running on there. For example, we have a company called TASS that we acquired for verification and validation of autonomous vehicles. If you look at your algorithms, you want to prove out that the safety requirements are there. Is that vehicle going to stop when you expect it to stop? Is it going to be able to identify the things that are there? We coupled that with our emulation tools and our capabilities at Mentor so that we can go to a customer and say, ‘Not only do you want to do your integrated circuit, we want to help you design the best circuit that has already been pre-validated to prove that it will do exactly what you want to do.’ With autonomous vehicles, it’s interesting to see what we can do based upon all of the acquisitions we’ve done. Some of those were planned, some were not. In some cases, when we got together we said, ‘Look at what you’ve got here.’ You can couple the emulation space with what we’re already doing with TASS and validation of driverless programs.

SE: How soon do you think we’ll see autonomous vehicles versus assisted driving?

Hemmelgarn: There are two schools of thought. A number of major OEMs say they see Level 1 and 2 and don’t see it going any further. And then companies like GM and Waymo and Tesla are investing a fortune in solving this autonomous capability. The difference is a perspective from how you view the problem. If you’re going to do autonomous driving in a ring fenced-in area in a city, we’ll see that in the next several years. Are we going to have full autonomous driving everywhere, with all vehicles and all locations? That might be a 10- to 15-year problem, depending on where you are. There are the one-off cases. That’s part of what we do with validation software. There’s an estimate that to get to fully autonomous driving you will have to drive 8 billion to 10 billion miles. No one is going to drive 8 billion to 10 billion miles. I saw an article that said one company is in the lead because they’ve driven close to 9 million miles. I don’t believe anyone is in the lead. It’s about how you leverage software to validate these edge cases, because we can run these edge cases and start iterating to get to those billions of miles using a programmatic method. So will it happen quickly? In ring fenced-in areas where you have an average speed of 12 miles per hour, that will probably happen. Will it be where you can drive anywhere at any time? No, that will take a lot longer.

SE: When you’re talking about geo-fenced areas, will those only be autonomous vehicles, or will there be a mix of vehicles?

Hemmelgarn: It’s going to be a mix. But then it’s about how do you communicate with the infrastructure and vehicle-to-vehicle. Inside of Siemens we call it chip-to-city. It’s everything from the IC to the design of the vehicle, the algorithms, and the city, because the vehicle has to communicate with the infrastructure, as well. When we first acquired TASS, they said there is a lot of work going on in vehicle-to-vehicle communication, but if we have to wait for all the cities to retrofit all of the traffic lights and infrastructure, that will take 50 years. And by the way, since we’re already talking vehicle-to-vehicle, and we have sensors, LiDAR and radar, why do we need that? They said it’s a matter of safety. If you have to rely on LiDAR, you might have the distance of five car lengths between cars, whereas if you have vehicle-to-vehicle, you can start to shrink it down. What will happen in ring-fenced areas is that it will be vehicle-to-infrastructure as well as vehicle-to-vehicle and vehicles with traditional LiDAR and radar, which will still allow you to have non-autonomous vehicles in that area.

SE: That communication system adds its own complexity, because now you have 4G LTE, sub-6 GHz 5G and millimeter wave 5G. There is data from cameras, LiDAR, radar and other cars. Where do you see EDA playing in this?

Hemmelgarn: EDA is a big part of this, no matter where you go. But the reason this is important to Siemens is we’re one of the few companies that can put all those pieces together. If you start thinking about what it means to communicate to the city, it’s 5G communication. One of the recent acquisitions we did was 5G emulation. Will a company come to Siemens for all of that? No. We will do pieces of that. But it will continue to evolve. If you’re talking to automotive companies these days and you’re not talking about electric and autonomous vehicles, it’s a pretty short discussion.

SE: Can the tools we have today be applied to smart manufacturing, or whatever moniker you use for that, where there won’t be this mass-market kind of approach?

Hemmelgarn: Everyone has a slogan for what to call it, whether it’s smart manufacturing or Industry 4.0 or Made In China 2025. My company, SRDC, was acquired by Siemens 12 years ago. So were a number of other companies. If you asked me for the top companies that would buy us, Siemens would not have been on my list. It took awhile and we talked about automation, and there was a vision of one person who brought us in-—Anton Huber, who has since retired from Siemens. He said, ‘Look, software and automation have to come together.’ He may have been a little early, but then Industry 4.0 hit and it became clear this is where we need to be. Since then we’ve been iterating and building. We can do things like virtual commissioning of a factory, where we take weeks out of a process. We can simulate it to the nth degree, make it run, prove it out. One of the best examples of how you know it’s working is when your competition is trying to copy you. They’ve made partnerships. Two of my competitors are partnering with other automation companies. They would not be doing that unless they’re getting pressure from their customers. We think it’s really starting to hit stride for the overall production process. There’s no reason that doesn’t stretch into everything we know about EDA. We’re seeing that already with PCBs. There are things we can do with IoT with the PCB development process, and things we see with how we bring those components together. The same thing will happen with integrated circuits in the future.

SE: The software industry has done a great job of raising the level of abstraction. The hardware industry has yet to achieve that, aside from things like high-level synthesis. Where do you see these two worlds coming together?

Hemmelgarn: We haven’t done ourselves a service by making things so complex. Years ago I was sitting down with the CIO of a large manufacturer, and he said, ‘When I go to the boardroom, I see a lot of discussions about ERP, but PLM and EDA are not in the boardroom and you have more impact on this company than an ERP system. The reason we’re not there is that, particularly in the U.S., there is this idea that the engineering guys will figure it out. Part of the challenge is that we’ve made it too hard, and sometimes we forget about the value it brings. In my organization we have 22,000 software people. We have to recognize that it doesn’t matter how great our mousetrap is. It’s what it means to the customer, what value it brings, and what difference it will make in their business. We spend too much time talking about our technology instead of the impact to the business. We make it difficult because we want to show how smart we are, rather than how we can make it easier to allow it to transform the business. This is true for almost all software companies. You have to make it easier and get to the value more quickly. There are signs of some of these things coming.

SE: Isn’t part of that due to the fact that some of the customers have changed? In the past you were selling to a chipmaker that developed their own IP and tools. Now, you’re dealing with everyone from a Google or Amazon designing their own chips, as well as carmakers designing a whole system.

Hemmelgarn: Yes, but even when you’re doing components of that design, we’ve made it too hard. It’s still about value and time to market—all the basic things that are there. As you start bringing these pieces together, it’s even more important. Complexity is not going away. If you tell someone you’re going to eliminate complexity in their process, that’s a misleading statement. There will be an estimated 100 billion connected devices in the next 5 years. With all that data coming in, how are you going to leverage that data? What do you do to make decisions about that data? It’s creating complexity. We advise our customers to use complexity to their advantage. The best way to do that is to set up a digital representation of what you have so that you can make decisions fast and with confidence. That’s where the real value is.

SE: How do you see AI fitting into this picture?

Hemmelgarn: If I start off talking about data lakes, machine learning and artificial intelligence, I get a lot of attention from the press. If I talk about blockchain, I get lots of interest, too. We like to brag about technology, but the real story is that compute power needs is a part of everything you do every day. You shouldn’t have to think about it. If you pull out your phone, it’s relevant for the time you use it. It’s connected, it does all the things you expect, and then you put it in your pocket and don’t think about it. We fact the same challenges as software providers with AI and machine learning. You shouldn’t have to think about it. It should be part of how we work. We need AI built into our tools. We learn how people are using those tools, and then we start modifying and making changes to the way we present data to the user interface based upon how they are working. It should be ubiquitous and part of what we use every day.