Executive Insight: Jack Harding

eSilicon’s president and CEO sounds off on consolidation, new business models, and why there’s still lots of life left in Moore’s Law.

popularity

SE: What’s changed over the past 12 months?

Harding: My starting point these days is around consolidation. At last count there were about 85 companies in the semiconductor industry. My bet is that at this time next year there will be about 70. The size of deal will not matter. Nothing will be too big. The strategic question is whether you’re playing musical chairs and when the music stops, you have no place to sit, or whether you’re the last man standing and have some independence. Clearly, when industries aggregate, size starts to matter. But if you can maintain your value proposition and you have a focus on a particular market, that focus tends to get valued more highly than if you’re buried inside of a huge organization. So is it all bad news because everyone else is getting bought, or is it good news because your competitors are getting swallowed up into a large corporation with the inherent inefficiencies?

SE: How are business models changing as a result?

Harding: There are businesses where they are in the same business serving the same customers, and there is more efficiency in the channel. That’s common sense. But in the M&A world you hear the word ‘transformational’ in almost every sentence. It’s a nice way of saying, ‘The industry isn’t growing anymore so you can’t bank on that. Let’s do something off to the right that is transformational.’ I’m curious to watch how many of these transactions are linear in their strategy—one plus one equals three, because you’re selling twice as much to the same guy—or whether you can give more breadth and a hedge to your business because you’re too concentrated in your market. The ones where you’re rolling up lookalikes into a bigger company are obviously good because you’re purging inefficiencies and leveraging scale. The transformational deals require real strategic thought and more risk, because you’re not sure if what you become is better than what you were before. Those are the deals to watch. They are the highest risk, but they potentially can combine marketplaces into a single institution.

SE: What’s your view on Moore’s Law, which seems to be a hot topic these days?

Harding: At a minimum we have another 20 to 25 years of this. From a technical standpoint we’re good down to 1nm. We’ve historically solved the business problems coincidentally with the technical ones, so if there is a way to get us back to a lower cost per acre, we’ll find a commercial way to make it work. But we’re talking at least four or five generations spread over the next 20 to 25 years. It’s a strategic discussion everyone should be considering, and with organizations like Imec out there, we’re in good hands. But the one thing I’ve been amazed by is our industry’s ability to extend and substitute so the average person doesn’t miss a beat.

SE: And that involves preparing all the possible options before we get to that point, right?

Harding: That’s correct. We’ve been thinking about design virtualization, where we provide ecosystem feedback to the EDA recipe iteratively throughout the entire process. So one decision made a year ago doesn’t propagate into a less-optimal chip once you spend $250 million to develop it. That’s our contribution. If you look at what’s going to happen at the single-digit-nanometer geometries, the area and mask costs will be important, but the bigger cost will be the cost of getting it wrong.

SE: We’re hearing a lot more about bet-your-business types of discussions. How real is that?

Harding: I have that conversation with our customers all the time. In the past six months I’ve had three conversations involving hundreds or thousands of miles of air travel just to have an executive sit down and look me in the eye and say, ‘We’ve given you this chip to develop and we’re betting the company on it.’ I used to have those meetings once a year. Now I have at least one a quarter.

SE: Is it the leading-designs or at the IoT level?

Harding: For us it’s the leading-edge companies. But the term ‘leading edge’ is confusing. It includes the finFET-class chips, but it also has to do with the relative complexity based on what you’ve done before. So you can be going from 40nm to 28nm and the electronics world in general feels, ‘We’ve got this.’ But as a CEO you’re scared to death because you’re betting everything on that one chip. It’s not just the absolute risk. It’s relative risk.

SE: What’s happening on the IoT side?

Harding: There are two dimensions to that. Big companies will build products they will sell to the world as platform-class products. Like every other standard product, there will be an approximation of what that platform will need. There will be too much area, too many features, and too much of everything for what people need. That level of overkill will be exacerbated by the industry because the user of platforms won’t be just other companies. It might be individuals or groups of individuals, ranging from academia to inventors with an idea to put a device into a toothbrush or a bedroom slipper. Those people will be the least able to deal with overkill in a standard product. That will create a bottom-up market for IoT that will be point-and-click IP, reasonably competent stacks of software and open-source hardware. There is the so-called Internet of Things, which is the connection of things that are electronically enabled. But there also is the Internet of Everything, which includes all those things that are not electronically enabled, as well, but can be or should be. It’s a superset of IoT. Those markets are going to be large and ubiquitous. They will shrink the world. And those companies that can provide infrastructure in terms of useful intellectual property or the mechanics to build these products and provide information to optimize them can do well over the next decade.

SE: You’re referring to the excess guard-banding in chips and correct by construction?

Harding: Yes. If you use the term hobbyist, for example, the notions of standards and security can’t be on the tip of their tongues. It has to be innovation and infrastructure that many companies like us will provide. One company has launched a library of Web-enabled black boxes. They want to control the access of those black boxes through the Internet, because that’s their business. That’s a very good start for someone trying to get into the market. If you develop a unique and very simple application, do you put it in the cloud so that someone can pull it down and use it and modify? All that is a very likely outcome. But it’s not the Internet of Things, it’s the Internet of Everything where you add interfaces to things that are not connected yet.

SE: Large system vendors are now building their own chips. How does that impact the market?

Harding: All of that is part of the idea that there are fewer design starts. On the other hand, when these major OEMs make electronics the centerpiece of their strategy, that extends the electronics culture to new places that are otherwise disconnected and creates opportunities for innovation. It brings a recognition and awareness about the electronics to the world. If a big social media company starts talking about electronics, that can affect 1 billion people, of which 950 million may never have thought about electronics. It plants the idea about what else they need or what they can do. In 10-year trends, it’s a good thing for the electronics industry. Of course, we’d love for one of these companies to say they want us to develop 950 million chips, which could happen as this stuff gets so complicated that most people don’t want to do it anymore.

SE: Changing topics, how is your business doing?

Harding: We’re doing extremely well on the booking side, fueled by our custom IP and virtualization tools and 15-year-history making chips. The issue we have now is too much volatility, which is commonplace in semiconductors these days. We’re much bigger than people think, although I can’t give you the exact number.

SE: And eSilicon is still private, right?

Harding: Yes. We’ve filed all the paperwork for an IPO but haven’t done anything more yet.

SE: Business models are changing everywhere as we started thinking about how things go together. How does that affect your operation?

Harding: I read an article recently in Harvard Business Review. The premise was that business models are becoming marketing-centric, not factory-centric. What more can we do for our customers? The value doesn’t lie in the capital assets in your firm. It’s in your ability to source assets globally. I agree with that. I was the first guy who said you don’t need CapEx to make semiconductors. For the first year or two people laughed us out of the room. By year three or four all but a couple guys had dumped their factories and were saying, ‘It’s all about what you’re doing for your customers, not what you own.’

SE: You’re referring to Jerry Sanders’ famous comment, ‘Real men own fabs?’

Harding: Yes, and there was Keith Lobo’s retort, ‘Real men make profits.’ That’s the hub of this evolution you’re alluding to. It’s a matter of, ‘What can I do for my customers to make things better for them,’ versus ‘This is what I make, who wants to buy one?’