Executive Insight: Charles Janac

Arteris’ CEO digs into connected cars, smarter interconnects, platform strategies, and why it’s so hard to be a CEO in the semiconductor market these days.

popularity

SE: What’s your biggest concern?

Janac: Staying current. One of the things that is really important is to understand shifts in the marketplace. In the past, we looked at whether to target our interconnect solution at digital TV or mobility. This was actually a very interesting question in 2005. The DTV market looked better, but we were not getting enough evaluations. We were doing better in mobility. If we had stuck with DTV we probably would have been out of business. The TV guys didn’t figure out how to be part of the Internet revolution. And now there are 1.3 billion phones sold a year. But nothing is static.

SE: What does that mean for opportunities—and risk?

Janac: The smartphone market is maturing. Qualcomm is taking a very big share and driving out a bunch of other companies. And the Chinese are coming up from the low end—MediaTek and Spreadtrum. With that concentration and the maturity of the market, you have to look at what’s next. The automotive segment, while not as attractive as mobility, will get quite a bit bigger than it is today. It has some very bad dynamics. It’s slow, the car guys don’t want to let the subsystem suppliers make enough margin. But the electrical car is much more sophisticated.

SE: What will this look like over the next five years?

Janac: There is the Automated Driver Assistance Systems (ADAS) with companies like Mobileye and others, which are going to become more capable. One of the real exciting things is that cars are going to communicate more effectively. The car behind will know whether the car ahead is braking or accelerating, and therefore maintain safe distances. That’s the precursor to the self-driving car. Then the next evolution will be guided lanes. If I’m driving from San Francisco to Los Angeles, I’ll go into a special lane. It may be a toll lane and more expensive, but I’m just going to let the car take me to Los Angeles. Then I’ll pick an exit and I’ll start driving again. But I’ll have five hours to do other things.

SE: How big is the market potential?

Janac: Figure there are 80 million cars made, and they’re going to have lots of chips in them. Unlike mobility, where there are only a couple interesting chips in each phone, there will be many interesting chips in a car. There are applications all the way from chassis control, which is critical and functional safety, with security applications, all the way to what’s involved in the evolution of smartphone chips. You’re probably going to have 10 really interesting chips in the car. So all of a sudden you’re up to 800 million chips. That means you have to make some bets in terms of functionality and development of those chips.

SE: But even though they’re in the same market, not all of these sub-markets move at the same speed, right? Some markets may not pan out while others are relatively mature.

Janac: Yes, and you have to place some bets. If you don’t focus on where the next growing area is, you’ll wind up like the DTV market.

SE: How do you figure out where to place your bets?

Janac: There are a lot people you can talk to and a lot of data you can analyze. You have to focus where the growth will be. You can’t abandon your existing customers, of course, but you also want to focus on the next growth area. If you go wrong, it can be extremely expensive.

SE: Arteris has made big changes internally with the sale of assets and your engineering team. That’s like a blank slate. Where do you go next?

Janac: The rebuilding of the engineering team management is largely complete. There is some basic functionality in the interconnect area you have to have. However, ultimately, our goal is to make the interconnect even more important than it is today. There are a bunch of SoC services that have to be delivered—things that allow the SoC to have pre-packaged functions. They’re highly configurable, so they don’t stifle iteration capability, but they’re essentially toolkits. There potentially will be some duplicate efforts inside the network for safety and safety controllers to get arbitration between the duplicate pieces of logic you have. It’s carrying safety-oriented signals and being able to control flaws in the electronics and make them much more reliable. And you can correct or tolerate them, depending on the use mode.

SE: This is like building in resiliency with software and what you’re programming into the interconnect rather than circuitry, right?

Janac: Yes. So there’s ECC coming out of the processor, but these signals have to be handled through the chip. You have to be able to deal with those safety-oriented concepts on an SoC-wide basis. And you have to have flexibility, because there are safety-oriented goal sections and non-safety-oriented goal sections. There are things like security.

SE: How will Arteris fit into security?

Janac: Security is such a huge area that we’re much better off partnering with companies. We view ourselves as providing the hardware foundation and connection to larger security solutions that go all the way up the stack to the content provider. We don’t have enough scale or experience or reach in that market. Security is a game for big companies. But we want to provide the connections to the hardware security stack.

SE: How far can you push that?

Janac: The interconnect can do a lot more than it’s doing. Today, it’s transferring data. Tomorrow, it’s going to facilitate the connection between hardware and software.

SE: You want to be the glue?

Janac: Yes.

SE: So you need people who speak both languages, and you need customers who know how to use that stuff, right?

Janac: That is the biggest problem. There are way too few of those people. From a mass-market perspective, they don’t exist. Software programmers have no interest in hardware. Their heroes are the Instagram and Snapchat guys. There is a need to package something that is hardware-aware that the software guys can use without changing much. It’s not an easy opportunity to exploit, but it does make the network on chip, or system-aware interconnect, more valuable than it is today.

SE: Hardware engineers are getting older, but there are very few young people coming into the market to push them into new areas. How do we fix that?

Janac: Hardware is hard. The question is how many young silicon engineers are being produced. There are very few engineers with both hardware and software knowledge. There are more hardware-savvy engineers, but the question is whether there are enough of them. Part of the reason is that it’s very hard to fund a semiconductor company now. The capital is not available because the exits are not as attractive as for other markets such as social networking.

SE: What does this mean for Arteris?

Janac: We are headed for a world that is very connected. People are talking about IoT electronics being very simple. But the useful stuff will be very complex. You see it in the Tesla. People focus on the fact that it’s an electric car. But it’s much more interesting as a connected device. The chips are complex, the software is even more complex. It does many, many things that no other system does. There will be a very important high-end component to that market. But it’s also not really an Internet of Things. It’s one automobile talking to another in a specific market. Maybe your dishwasher talking to your refrigerator will be less sophisticated. But the useful things will be very sophisticated.

SE: So where does that leave the semiconductor industry?

Janac: The future is very bright, as long as you can focus on the right segments. Without hardware, none of this will happen. The hardware needs to become increasingly sophisticated, but the hardware may look significantly different in the future than it does today. The value has shifted to the software business model. If you look at what you’re interacting with using the Tesla example, it’s the software. It’s a very sophisticated system. The future is in electronics that actually give you something of value. A lot of that is driven by new business models. It’s not driven by the hardware. It’s a new business model where the hardware is the compute engine for that business model.

SE: Does all of this lead to a rapid assembly platform approach?

Janac: Yes. You want a platform that can shift on a dime. If it’s too far, it won’t be effective. One of our customers once said a digital TV is just a giant tablet. Electronics will be a derivative business.

SE: Who drives that? Is it the chip companies or the system companies? Or do they morph into the same thing?

Janac: What I’m seeing is the system companies are going to start to do chips. You have all the way from people defining things on napkins to a billion-dollar SoC capability. The system houses are certainly going to go into the architecture phase, but the SoC design is so expensive that there are not many companies that can plunk down $1 billion for semiconductors. So it’s going to be like a pendulum. It’s going to swing from insourcing to outsourcing. The system houses will be players, but there’s room for nimble chip companies that can address these kinds of new businesses very quickly.

SE: Is this a multiple platform strategy?

Janac: Yes, but you don’t need more than three or four platforms. You need a low-end platform to add and subtract IP for a low-end chip, you need a high-end platform, and you probably need some kind of an analog-mixed signal platform.

SE: And you can put those together with an interposer in the middle, right?

Janac: Exactly, and that makes our kind of interconnect really valuable. You can build a new interconnect to add or subtract IP in a day or two that’s fully verified. That allows people to become nimble.

SE: When does the inertia end for Moore’s Law?

Janac: Moore’s Law is finished at 28nm. Once you go to double patterning, the only thing that gets better is density. You get higher dynamic power and higher cost. At 16nm or 10nm, the only thing that will make sense are the leading-edge digital dies. The rest is going to be on trailing-edge processes. Plus, you’re going to have memory. Everything is going to go 3D. People are saying it’s too expensive and asking who’s going to take responsibility. In my opinion, the packaging houses become very important in this. It’s expensive and it’s complex, but there’s no other alternative. What the actual technical implementation will be I have no idea, but people will come up with solutions to combine a trailing-edge mixed-signal die combined with a leading-edge digital die and a memory. Those stacks are going to be ubiquitous in a number of industries.

SE: This is a lot of changes for the industry.

Janac: Yes, and the biggest challenge for a CEO is managing that change—knowing when to invest, when not to invest, where to put your money. But if you’re wrong it can be frighteningly expensive. And the worst part is that at times you have to eat your children, so the product that made you successful may have to be replaced or sacrificed—and maybe even replaced by a different business model. That’s where the really hard, gut-wrenching decisions have to be made.