Executive Insight: Wally Rhines

Mentor’s CEO looks at consolidation, what’s behind the Siemens acquisition, and the big opportunity in system design and integration.

popularity

Wally Rhines, president and CEO of Mentor, a Siemens Business, sat down with Semiconductor Engineering to talk about industry consolidation, a shift in emphasis from chips to systems, and what the recent acquisition by Siemens will mean for Mentor. What follows are excerpts of that conversation.

SE: A year ago it looked as if the entire industry was going to be boiled down to a handful of companies. Consolidation has slowed down considerably—is this a pause or a trend?

Rhines: The data is not quite so overwhelming as the emotion associated with what’s called ‘semiconductor consolidation.’ This is due to several factors. First, even with the $100 billion per year of market value acquisitions that have occurred with almost 70 acquisitions over 2015 and 2016, it’s still a relatively small consolidation. The top 10 semiconductor companies now have a combined market share 2 points great than they did before this big wave. They’re still only about 2 points higher than they were 40 years ago, so it’s not that overwhelming. Second, it looks like they’re a bigger part of semiconductor revenue than is actually the case, because increasingly in recent years we stopped measuring semiconductor revenue to the same degree of accuracy we used to. Why? Because companies like Apple, Samsung, Huawei now make their own application processors. When you look at the semiconductor data, the growth year to year, you see the semiconductor revenue from TSMC and other foundries but you don’t see the other 50% of the revenue that occurs. When you add that all up, in a single year, that’s been more than $10 billion of revenue subtracted out of the semiconductor TAM (total available market).


Photo credit: Paul Cohen/ESD Alliance

SE: What’s the third element?

Rhines: The actual combined market share of the 50 largest semiconductor companies has decreased over the past 10 years. Market share is not consolidating. It’s actually decreasing as it always has. The semiconductor industry is an industry where you climb to the top based on a specific capability, build a large market share, and then you become big enough and you grow at about the same rate as the overall industry. All the new startups and new companies are entering based upon a new wave of growth, so they’re growing much faster than the overall semiconductor industry. Over a period of time, they displace companies in the top 10 and become part of the top 10 infrastructure. They, too, have to be innovative and nimble if they want to stay in the top 10, because two-thirds of the companies that have been in the top 10 are no longer in the top 10. One of the signs of maturity of the semiconductor industry that’s always cited is that just like any mature industry—like steel, cars, or whatever—the semiconductor industry has hit a plateau, growth rate is low, and it needs to consolidate for efficiency. But it’s not really consolidating. It’s actually growing a lot more. People also point out there are hardly any fabless startups anymore. But there were reported 800 additional fabless startups in the last two years in China alone. Because of the Chinese government stimulus, the estimates are that the Chinese semiconductor revenue will grow $350 billion over the next eight years, which is more than 100% of the semiconductor revenue in the world today. A lot of factors say, ‘Don’t take that consolidation too seriously because 80% of it was six mergers.’ Even though there were a lot of smaller ones, almost all the dollars were in just six.

SE: Isn’t part of what is going on is that the industry is getting set for the next waves of growth? In the past, we’ve gone from PCs to mobile, and while mobile processors are still a huge portion of the industry the growth rate is flattening. There are a lot of new things coming that will drive a a whole different wave of startups, growth, and different ways of doing things.

Rhines: Yes, and this happens about ever 10 to 15 years in history, sometimes less and sometimes more. In my early days in the semi industry, the success in the military and aerospace was really what drove the largest growth. Then it was mainframes, minicomputers, PCs and laptops. In the past 10 to 15 years, the biggest part of our growth has been wireless. Communications as a segment has gone from 15% to 20% up to 35%, which is even greater than computers as a segment. Once again, like every other time, the winners consolidated out so there are a lot fewer companies that are generating a large share of the wireless revenue. As a result, it’s now a big semiconductor business and can’t grow in double digits anymore, so we’re waiting for the next wave.

SE: What a lot of people found surprising in the last consolidation wave was the size of the mergers.

Rhines: There are several different types. There are mergers driven by efficiencies, where you simply eliminate duplication, reduce costs, and if you have overlap you gain market share in segments or you just bulk up. That is more common than not. With 65 of the mergers in the last two years, the average synergies—a euphemism for cost reductions—were 25% reductions in operating expenses as a result of the mergers. Some mergers were more aggressive than others. Avago was clearly very aggressive in their projections of synergies. Some are less so, but the average is 25%. The interesting part is that people thought they were going to cut R&D or G&A (general and administrative expenses) by 25%. So what happened to the R&D? The R&D from the semiconductor industry grew in 2015 and 2016. Consolidation didn’t end up taking away the R&D. If it did, new competitors would enter the market and take advantage of the fact that the establishment was not investing.

SE: Mentor is part of this as well. You see if from the inside, with Mentor being bought by Siemens. What’s changed in the company internally?

Rhines: In the case of Siemens/Mentor—and also the Softbank/Arm mergers—they were a different kind of acquisition. In both those cases, you read the press release and there is not a word about cost synergies or reductions because there is no planned cost reduction. What you see in the Siemens press release is revenue synergies. It’s Siemens’ belief that Mentor, by spending more on R&D and in customer support, can grow its revenue faster than if we continue the way we were before. That was a factor our board had to consider. How fast did we grow? Can we maintain or grow our market share as a standalone company versus if we had someone with very deep pockets with a slower growth rate that was looking for new areas to invest in, new areas to expand in? That’s what Siemens presented to us. ‘We want you to invest more and as a result we want you to grow faster.’ We responded with, ‘The areas we will invest in are the areas we have number one market positions or have the likelihood of getting the number one market position.’ For Mentor that means things like the Calibre family of 40 products, design for test, high-level synthesis, emulation and things like that, plus some where we are gaining market share very rapidly and doubling revenue, like SPICE simulation. On the system side, in areas like cabling, wire harness, specific areas of automotive, embedded software, it’s a totally different kind of thing than the typical merger and acquisition you read about.

SE: In the past one would create a startup, get bought or go public. What does EDA look like going forward now that Siemens has purchased Mentor?

Rhines: Mentor, Cadence and Synopsys have had a combined market share of 75%, plus or minus 8%, for more than 20 years. Now we’ve disrupted that stability.

SE: So what effect will the Siemens acquisition on Mentor’s future direction?

Rhines: In design methodology and tools, we grew about 2X the market rate. Why wasn’t our overall revenue growing much faster than the market as well? The answer is that even though we started the IP business, we eventually dropped out of it. In 1994/1995, we did 29 acquisitions of IP companies, we built the largest market share, and everything was going great. Then Synopsys discovered what was going on and got aggressive. We simply didn’t have the money to counterbid. They had deeper pockets. For every new startup and new technology, we would go to bid and we would lose. Mentor had a strategy that said ‘you’re either number one, you’re going to be number one, or don’t clutter the world with yet another competitor.’ The world doesn’t need three suppliers or more of everything. But going forward, we’ve got very deep pockets with a parent that does $90 billion a year in revenue with positive cash flow, a desire to grow faster, and a belief that virtual design of electronics and of systems is where the biggest growth will occur. Digitalization—the ability to virtually verify, to go from concept through design, development, manufacturing, and support, and do this all on computer and a company—that has quite a strong foundation on data management, mechanical CAD, device modeling, computational fluid dynamics, and so on. It’s now a division that has a strong level of competence in electronic design, both system and IC. So what happens moving forward? One thing is the money is there for us to compete acquisition-wise. Second, the resources are there to grow into new areas that we have not been investing in as fast as we could. And third, it offers the opportunity to penetrate this interface between system-design and IC design. That is a big opportunity I’ve talked about at almost any meeting I’ve been at before. In IC design, we’re pretty well automated and penetrated through the entire flow. In system-design, the world is in the stone age, so a big growth opportunity exists if you can virtually design and verify a car, plane, or train instead of building prototypes and testing.

SE: We’ve been discussing this for the past decade, at least. Why is it going forward now while it didn’t in the past?

Rhines: There have been people promoting the theory for more than a decade, maybe 20 years. Then they’ve come to Mentor in the past and said, ‘We think all this mechanical/CAD has to come together with electrical and you’re going to have one great terminal, and every engineer is going to use the same terminal so they can get access to all the data.’ I have been one of the greatest critics of this. And I still am. I don’t believe electrical design engineers want to sit at a terminal that is general purpose enough that you can do mechanical CAD and electrical and so on. What they do need is access to pieces of information in the other domains. Why are startups successful when they are competing against big companies? My conclusion is a new problem comes along, companies try to develop a solution, the big companies have the software development in one group, the hardware development elsewhere, IC design elsewhere, and they’re all working and they never get an optimum system solution. The startup’s software developer is the same person as the hardware developer who is the same person as the marketeer, the salesman, and so on. They look at the whole problem and put it together.

SE: What has to happen for system-design to automate the same way that IC design has done?

Rhines: They have to get tools that allow the electrical designer to get access to enough of the embedded software and the mechanical design that you can develop solutions that are optimum. Today, Mentor has the leading market share in tools to develop and design wire harnesses for cars, planes, and trains. You may look at this and wonder how difficult this can be—it’s a wire harness, not an integrated circuit. The answer is that it can be very difficult. It’s a 3D wiring problem, which your IC is becoming, but is not yet. Second, there are thousands of constraints that have been built up over the years. One car company will say, ‘No more than five wires in a bundle in a wet zone in a car.’ Some other company has a different set of rules. Those constraints are just like the constraints you have that are built into your router and your verification tools. Now you get the problem where you have 150 ECUs and 100 million lines of software code. No human can develop and verify the wiring, so now you have to go to automated wiring. How do you know that the wire bundle will fit through the hole in the doorframe? How do you know the wire is long enough to go up through the roof of the car and back down to the other side? That’s in the mechanical database. How do you get access to that? With great difficulty. But over the past 25 years we’ve been doing this, we’ve found ways to get that data. I can tell you that it hasn’t been easy. The suppliers of the mechanical are the owners of the mechanical data and are somewhat reluctant to give access to that data. Over the years, we managed to get by, but when you have a partner that now gives you total access, it now motivates your other suppliers to do the same so you can provide an efficient solution. Now, all of a sudden, you have a car wiring system that can take into account all aspects of the mechanical design, optimized with the electrical design, and produce a lower-cost, shorter-wire-length, more-efficient, more manufacturable system. And you can do all sorts of things you couldn’t do before.

SE: We’ve seen a push toward much more heterogeneous systems, that includes more kinds of components, accelerators on a single chip, more chips in a package, and more components on a board. What does this mean for designers?

Rhines: This is what is great about EDA right now. If you distinguish what is different about a system and IC designer, traditionally it has been that one designs components while the other integrates disparate technologies into system solutions. System suppliers typically run higher margins than component suppliers. In SoCs, people who do more complex integration run higher margins on their ICs than people who sell commodity components, like memory components. We’re moving into the age of design automation for systems, which means we can add more value, but it requires new expertise. If you look in IoT modules, now suddenly we have to worry about Zigbee, Bluetooth, RF, digital, analog, MEMS, and photonics, and all of this working together on this little IoT module. How many people have experience in simulating analog and digital together, much less analog, digital, RF, MEMS, photonics? This is a system design problem. The new innovative designers who take advantage of this and who develop that system-level verification and design capability, are going to be well rewarded as are their companies because it’s not easy. It’s a whole new opportunity to develop expertise that designers need to get into and spend time. Digital has gotten very big, but it’s not going to grow at nearly the rate that startups and new designs are going to grow in mixed technologies.

SE: So what becomes the next challenge—technology, business, or ecosystems?

Rhines: There are problems in all dimensions. Every time we get a new node, we get a new set of physics problems, new complexity problems, and a lot of the tools break and we have to develop new generations of them. Let me separate the evolutionary from the revolutionary. Traditionally, EDA grows not by evolution but by revolution. If you go back and look at the segments of EDA, the mature ones—even though they encounter new problems with every generation—the revenue doesn’t grow much. Event-driven simulation has grown 1% to 2% per year over the last 10 to 15 years. Place and route has grown about 2% over the last 10 years. PCB design tools have grown about 1% to 1.5% for the last 25 years. Each of those technologies had to dramatically improve its capability to manage the complexity that’s occurred as electronics increase, but you don’t get rewarded for it very much. People expect to buy the new set of tools for basically the same number of total dollars, plus a little, each year. In our event-driven simulation, we increase the number lines of code by about an order of magnitude per decade. So where does the growth occur? It occurs when a new problem comes along for which there are no tools. In 1999, optimal proximity correction accounted for zero dollars for EDA. Today, optimal proximity correction provides more than $100 million in revenue for Mentor, and the industry is approaching $200 million in revenue and growing very rapidly. It didn’t exist 15 years ago. It’s now a fast-growing and big part of EDA. The same is true for these other new technologies that come in. Someday photonics will be one of those, but it’s not today. In the last decade, analog/mixed-signal in various forms has grown. We’ve had substantial growth in other new technologies. High-level simulation and synthesis has grown from zero dollars in the 1990s to a $150 million market today. Our high-level synthesis and power optimization tools we have with the Calypto family doubled its revenue last year. The new technologies that are solving the new problems are where the growth occurs. The grand-daddy of them all is system-design, the ability to apply the fundamentals of IC design automation to designing systems on a chip and systems that are bigger systems, and integrating that with embedded software where the verification problem has increased dramatically and where the revenue growth for EDA companies is growing substantially.



  • Karl Stevens

    The closing statement, ” The grand-daddy of them all is system-design, the ability to apply the fundamentals of IC design automation to designing systems on a chip and systems that are bigger systems, and integrating that with embedded software where the verification problem has increased dramatically and where the revenue growth for EDA companies is growing substantially.” implies that system design is an “extension” of IC design, but system design is about partitioning functions into functional blocks and connecting the blocks logically to perform that function.
    Existing EDA tools are based on HDLs that do absolutely nothing in the way of interconnecting modules other than instantiation of modules.
    Meanwhile in the programming world OOP compilers and debuggers have been developed that do a great job with classes that perform functions and logically connecting those classes.
    As a retired computer systems guy, it would have been great to have had this capability.