Ecosystem Changes

Semiconductor Engineering sat down to discuss changes in the semiconductor ecosystem with Kelvin Low, senior director of foundry marketing at Samsung Semiconductor; John Costello, vice president of product planning at Altera; Randy Smith, vice president of marketing at Sonics, and Michiel Ligthart, president and COO of Verific. What follows are excerpts of that conversation.

SE: How are relationships changing between companies?

Costello: If you go back five years ago, everything was very insular. We did a lot of our own design IP and we rarely considered what was happening in the industry. In the last three to five years we’ve been leveraging partners both on the tools and the IP side. We need to collaborate to move forward. In the past we had manufacturing partners for back end assembly and test. But on the design side, that’s a change. And it’s for internal design, as well as to enable customers to do designs.

Low: A lot of the foundries tried to do standalone process development up until about 90nm, when the cost of R&D escalated. That’s when we partnered with IBM, Chartered, ST and a few others to form the Joint Development Alliance. At one point the alliance grew with eight partners on board. They all wanted the same thing—to develop a leadership process technology. That’s on the process side. On the customer interaction, we see a much earlier intersection point. From a engagement perspective, discussions happen three to three-and-a-half years before the first product shipment. This type of relationship becomes even more important as we move forward. We’re using finFETs now, but looking into new device architectures and new materials, so expect even tighter discussions in the future.

Smith: The baseline is changing. As an IP company we have to work with everyone. We have to work with tools companies, with customers on specialized development for them, and also with other IP companies. What is changing is that we’re used to doing the kind of development where you’re working in advance on something you anticipate will involve future needs. That has moved to support partnerships. A company is in the middle of trying to get something done and they’re not having success, and we’re impromptu gathering people from different companies and converging on the customer to solve the problem now. They have a tapeout that has to be done and you have to jump in. Companies are coming together to solve very specific problems but in a very short period of time. We’re all building time for the consumer market now. It can’t be delayed three months.

Ligthart: We have a very unique relationship with our customers. Every customer is a partner. Our IP becomes part of the product of our partners. It’s all real-time. If my customer has a problem, it’s immediately my problem, as well. We’ve been doing this for 12 years for our customer base. But for companies that have many more customers, this is new.

SE: Complexity certainly has gone up. How is that affecting relationships and how companies deal with each other?

Costello: As an FPGA company, we’re trying to do many things for many people. One of the things we’re doing is adding a lot more capability, which brings complexity. The more complexity, the harder it is to do the development and the more costly, and the harder it is to deliver a quality solution across the board. Increasingly you have to rely on partners to help solve those problems. Companies have their own domain of expertise, and you want to stay focused on those. If you try to cover too many things you dilute your effectiveness and the value of what it is that you’re trying to do. It allows us to deliver better solutions to customers, and our customers are looking for higher value, more sophistication, more integration, and moving up in the value chain.

Ligthart: It’s very important that both partners are aware they can share. Our customers share regression tests. If they share those with us, then we improve the quality of our products. The whole ecosystem then improves.

Low: Multiple parties can improve that together. There is always concern that we are both a foundry and a product company. We have to actively provide assurances to customers, showing them the system we have for protecting their IP. There is a delicate balance that takes time to build up, and we cannot break that trust. I’m confident to state that we have addressed these concerns successful with our customers.

Ligthart: I agree that intellectual property needs to be protected, but that’s also old-school thinking. If we all work together, it will all work together. If someone wants to share 1,000 test suites in the SystemVerilog area, then everyone gets improved results. Keeping that completely separate causes a problem.

Low: We need to break through this for many customers.

Smith: The fundamental problem that tools and IP companies have is that their customer doesn’t have any data. They don’t have a real live example. If someone says their next chip is going to have 400 blocks, how do you test and verify that capability? We need the depth of that relationship to go beyond what you’re going to do. We need to work with them and we need their data to find new capabilities. It can’t just be early access to requirements. We have to be there with them. Now the question is whether they’re willing to do work to allow us to help them. Sometimes in a customer relationship, as opposed to a partner relationship, they’ll give that information to everyone. Here’s the spec everyone needs to build to. Most of the people working at the leading edge are better off picking a partner. They need something and they need to pick one company to work with.

SE: What happens when you have multiple companies and you’re forced to work with them? Maybe you aren’t as pleased with their work as you are with other companies, and you’re only as good as the weakest link in a chain of partners.

Smith: You have to work that out. We have that going on. We have a yet-to-be announced partnership, and they have selected other IP partners. They’re not companies we’ve worked with in the past. It’s a shotgun marriage. We’re getting to know each other a little better and we’re going to work it out. We know the relationship a couple months in advance. We don’t get to choose.

Costello: The other part of that is to understand the value that each partner brings. As long as you have a good understanding around that, you can have confidence in what you’re bringing in, perspective on what the other partners are bringing in. Everyone is already oversubscribed. But if you understand everyone else is bringing, and you can get the customer to understand all the interdependencies and how to motivate all of us, that can be effective.

SE: What happens if you’re working on a 14nm chip with the most advanced process, finFETs and IP that isn’t fully baked? If something goes wrong, who’s to blame?

Smith: The blame part is after we fix the problem so we can go back and make sure it doesn’t happen again. But the immediate challenge is fixing the problem, no matter how we got there. That’s the way partners expect each other to behave—hard on the problem, not hard on the people.

Costello: It has to be a win-win across the board. Blame doesn’t really get you very far. Given the reality of the situation, how do we move forward so we’re all successful?

Ligthart: When we get to the post-mortem, someone will be out-engineered for the project going forward.

Smith: I have several customers who have been with us time after time. If they come to us and say, ‘I’ve got a problem,’ we could look at it and say, ‘If you had moved to the newer release like we told you to, you wouldn’t have that problem.’ But that doesn’t really solve anything. It’s not helpful.

SE: But there also can be ecosystem issues to consider. The update of the recent Apple iOS suddenly made high-end car infotainment systems glitchy. How do we move this entire ecosystem forward?

Ligthart: The whole backward compatibility issue is different in the consumer than the software and EDA business. We make everything backward compatible 15 years back. It’s difficult to tell our customers that, as of March 1, it will no longer work. But in the consumer market, that’s the way it is. You can buy a new TV and suddenly it doesn’t work with the router and speakers.

Smith: Partnerships are changing. A lot of companies we didn’t know we needed to talk to we now understand we have to talk to. These are companies providing services around EDA. They’re not tools companies. But at times we need to reach out to these guys because they’re supporting our customers. The network has gotten bigger and it’s no longer one size fits all.

Low: There are dominant companies that will try to enforce the way things are done so that their competitors can’t catch up. That’s not going to change. In the foundry arena, we are changing this by providing flexibility and choice to our customers.