IP Market Shifts Direction

Experts at the table, part 1: Even standard IP is getting customized by large vendors as consolidation continues to change the fundamental premise for third-party IP.

popularity

Semiconductor Engineering sat down to discuss intellectual property changes and challenges with Patrick Soheili, vice president of product management and corporate development at eSilicon; Navraj Nandra, senior director of marketing for DesignWare analog and MSIP at Synopsys; Kurt Shuler, vice president of marketing at Arteris; and Charlie Cheng, CEO of Kilopass. What follows are excerpts of that discussion.

SE: What are the problems in IP these days?

Nandra: Customers are starting to look for differentiation in IP. That breaks the whole idea behind IP because you want to build an IP company that is scalable with IP that works everywhere. Leading customers are not just starting to outsource IP—they’re outsourcing the entire team. We made a recent acquisition involving AMD’s IP team. We’re starting to deliver a whole bunch of IP to customers. But how do you maintain an investment and margin that can support a standard IP model and still give the customer some kind of differentiation out in the market?

Soheili: We’re an ASIC supplier and an IP developer. We don’t make money until the chip is in production. It’s all about the quality of the IP. To make your PPA (power, performance, area) the IP has to do what it’s supposed to do. From a development perspective, the same challenges we see in our IP we have ourselves. The customer wants to see their differentiation all the way down to the physical level, and then some. They go to the foundries and have their own PDKs. They have their own compilers or their own special I/Os. We see that as a huge challenge. And then the cost of development, between tools, labor, competencies and the number of hours that it now takes for double patterning, and then all the PDKs—as things get more complicated there is version 1.1, 1.2, 1.3—and with all of those your design margins are getting impacted. We see it both ways.

Cheng: The big issue for us is consolidation. The need for differentiation is a byproduct of that. IP reuse used to be the way to justify development costs of IP. You could simply amortize over multiple customers. But when the industry starts to consolidate, and it’s not difficult to see sub-billion companies getting sucked up, there isn’t as much reuse as we like to think there is. Those are the ones who are licensing IP—there aren’t that many players—and to compete you have to differentiate. The amount of differentiation needed for every different customer is monumental. The customers are consolidating, and the fabs are consolidating, so both ends of the supply chain are getting squeezed, with us in between.

Shuler: We are starting to see camps forming because of consolidation. For us, the challenge is trying to remain independent of all of these camps. What we’re also seeing is that if you have a company providing GPU cores and memory, often they supply a low-level interconnect. It may not be the best, but it’s free. So the challenge is you have to supply a better deal than someone getting something for free. If they’re paying money, what you’re offering has to be much, much better than what they’re already getting.

SE: This all runs contrary to everything we’ve heard about IP, which is that you want to develop once and re-use IP many times. Is this new model even sustainable?

Nandra: We are finding ways to sustain it. If you’re looking at a high-speed memory interface or USB or PCI Express, architecturally and electrically they are the same. Where you differentiate, as Patrick said, is in the PPA. Why would a customer come to us? It’s because what you’re offering is smaller and lower in area. We used to talk about standard interfaces. This is now definitely a requirement in the mobile market, and it’s the same technology in the enterprise market. But the PPA—the packaging, the lane configuration—they are different. You have to gear up for all that differentiation.

Soheili: You push the customization out as far as you can.

SE: So you have a core of the IP, and then you modify other parts?

Soheili: Yes, and the farther you can push it out from an engineering execution point of view, the more similar it is to an off-the-shelf part from a business standpoint. It’s do-able, but it’s not as profitable. But to Charlie’s point, if there is a consolidated market and the volumes make up for a fragmented market, then you’re making more money.

Shuler: For us, everything is 100% configurable. We can create a custom interconnect. We deal with that all the time, and it’s becoming easier as a lot of companies are moving away from internally developed protocols. So it’s certainly easier from that aspect. But when you think about integration, such as with IoT sensors, there is now a lot of strange stuff on the chip. The bigger customers are not getting a version of IP that is off-the-shelf. Some customers will use off-the-shelf IP, but with others they will license cores and then add in a ton of custom IP.”

SE: If you customize IP for each customer, your characterization is going to be for that customer, right?

Nandra: There is more characterization. Whoever is supplying the IP has the infrastructure in place to do testing to their satisfaction. But in the automotive market, for example, there is driver assist. When you bring that down into qualification requirements for IP, that’s a completely different infrastructure. It includes temperature ranges, ESD levels, electromigration. If you have ODP (OpenDataPlane), can you get it to work at 175 degrees Celsius in an environment where it’s going to be subject to electromigration? These things do have to be tested.

Cheng: We have two distinctive R&D areas. One is our fundamental bit cell. We have to be able to only characterize that once. Periodically we have changes, and it’s usually not just for one customer. If it’s just for one customer it’s extraordinarily expensive. Getting it to 175 degrees Celsius is very costly. On the edges, a wider bus, more robust design, faster, lower power—that kind of characterization thanks to Cadence, Synopsys and Mentor Graphics, is very doable. Our limit is our engineering team’s ability to absorb new technology coming from the EDA industry—enough money to buy all the latest tools. But it’s not a limiting factor to our business.

SE: In the past, there was value in knowing which IP worked best for a particular chip. If it’s going custom, that’s not as important, right?

Soheili: That’s correct. It’s all about ROI. If the customer is willing pay in risk to get differentiation then we can take that risk with them. But it has to be understood and it has to be quantified. If the customer is risk-averse, and in exchange give up some differentiation, then we don’t do it.

SE: How much of this is affected by the process node?

Cheng: The dominant factor is the consolidation. Once in a while we are a customer, as well. When we go out to find partners, it’s amazing how much the field has thinned out. We just have one really viable vendor that can do all the things we need to do. We didn’t really have a negotiation like we did 10 years ago, where the CEO storms out of a negotiation and the vendor comes scrambling out trying to get a deal done. It doesn’t happen anymore. Customers have thinned out and the vendors have thinned out. If you take a deal for low ROI, it probably means you don’t have a very good market position or value proposition and you’re taking a deal that’s not healthy for the business.

Soheili: The risk is both higher and more measurable because it’s consolidating.

Cheng: That’s right. It’s very hard to say, ‘I’m going to go to a startup and not Synopsys for a particular product line,’ where Synopsys has a well-established position because there’s no reason to take that risk—especially for a consolidated field where most customers are vying for multiple billions of dollars of market valuation.

SE: So the market valuation is coming down and forcing more consolidation, right?

Cheng: I’m not sure it forces further consolidation. All the analog/mixed signal companies seem to be merging. For an IP company, there are just fewer customers and there is much more chunky business to win. If you don’t win several of them, you’re out of business.

Nandra: It’s an opportunity, as well. Sometimes you are working with a smaller startup that wants to buy IP because they can’t develop internally. We’ve seen this many times over. The bigger company they’re trying to get the IP from convinces management that their IP is better. We do see an opportunity to go in and present ourselves. And this is the opposite of what I said at the beginning of this discussion—you go in with a standard set of good. And then we start peeling away how we do things.