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: Is there a linear relationship between the number of new customers and the number of problems that need to be addressed in complex relationship, or does it increase exponentially?

Ligthart: It’s linear, and if you have a good product that trend line should be flat. You have more customer inquiries, but the number of problems doesn’t have to increase.

Costello: It shouldn’t increase exponentially. If it does, it points to a more fundamental problem. It’s inevitable that if you do more complex things, you’ll have more interaction with customers. But it’s a relatively small number, and the impact on customers is identifiable and manageable. If you end up with an exponential problem, something has gone wrong.

Smith: We look at this differently. We’re constantly applying quality assurance to stuff we’ve already got in silicon because the nature of such a highly configurable piece of IP is that you can’t test everything. Every night we’re running thousands of constrained random networks to find things that don’t work. If we come up with something, we give our customers a script. It’s proactive support.

SE: At the existing nodes, processes and methodologies are relatively stable, but at the most advanced nodes the number of unknowns is increasing. How much of a problem does that cause for relationships with customers?

Low: We’ve increased transparency to address that. Coverage is always a problem. It may be 99.99% or 99.999%, but there is always an element of risk. There is always time-to-market pressure that all of us face, so we have to resolve it quickly. You have to create an architecture where you can still manage the risk. As long as we are transparent, we are jointly managing the risk.

SE: Are your partners as transparent as you are?

Low: It’s not just one partner initiating the transparency. And yes, they do reciprocate. But it’s still a journey. There are certain things we may not share with them. For example, there are companies that want to know fab information such as the tool speed and which chamber the wafer goes into because they think that will help with the product bring-up. So this information sharing does proceed at a cautious level, but it is increasing.

SE: Everyone here has arch-competitors who are also partners, and people move from company to company throughout the industry. How does that affect relationships?

Smith: We talk a lot. There is definitely a level of co-opetition. We never talk badly about each other, even internally. They’re a partner and that’s how we behave. But when we compete we have to do things that will give us an advantage. This is all about having relationships on multiple levels. It’s CEO to CEO, CTO to CTO, VP to VP. You need to be comfortable working with each other, and we accept that we don’t have a lot of secrets. We even share road maps.

Costello: There’s a portion of this where all the competitors in an industry work together to enable that industry. Everyone within that industry believes they will have a competitive advantage that will allow them to succeed if they have the right products. Recognizing when it’s good for the industry as a whole, and then believing that you can compete successfully within that industry, is key to success. Having an open relationship with competitors is difficult to do, but we all need to recognize that we can only be successful if we work together to make people feel good about the industry. And then we have to trust that we can compete with our own strengths.

Ligthart: Our situation is different. Many of our licensees are competitors with each other, and they will compete until death, so for them it is very important that we never talk about one to the other. We are very trustworthy there. Looking at the longevity of our company, we’ve done that successfully. But you do not talk about other customers. Our online defect tracking system never mentions company names.

SE: Can an ecosystem with tight partnerships work as well as an integrated device manufacturer?

Low: One side of that is whether you can be as efficient in driving solutions. The other side of that is to let experts do what they do best. The reason why the IDM model isn’t working is that it isn’t cost-efficient to do this anymore. The ecosystem is more efficient and it will replace the IDM model.

Ligthart: In the past, companies would have done everything themselves and spent many man years on it and come up with a solution that wouldn’t be as good as ours. They don’t even talk about it anymore.

Costello: In the past, the IDM worked well. The key to that was the complexity was something you could get your hands around and you could afford to solve it. With today’s complexity, it’s impractical to think one company can do everything as well as everyone else. We are forced into an ecosystem environment. It brings some inefficiencies with it, but the overall gains are higher. There is a little turnaround with big systems companies where they are starting to bring some things back in again, but it’s happening in a much more limited manner. While they may be doing more integration, they’re requiring on an ecosystem of software and applications to make products successful. They’re integrating more themselves, but a lot of their success is based on the ecosystem.

Smith: There are two trends that are at play here. One is whether you’re going to fab or fabless and breaking things off into many components. The other is whether you’re doing customized chips or standard parts. Systems companies can do custom development and make their money that way. At the same time, IDMs moved into the fabless world. They affect each other. It’s like looking at different waveforms.

Low: The definition of the words foundry and fabless are changeable. We were very much more loosely connected in the past. EDA would come in at a later time and our customers would come in a much later time, where we just supplied the technology and they would do whatever they needed to do. Multiple parties now talk together much earlier. It’s actually an IDM model in a foundry/fabless execution environment. The term we’ve heard in the industry is integrated fabless model.

SE: Are there different levels of partnerships, though?

Smith: It all depends on how much each customer is willing to share. That affects how much we can act like a partner and how much we can share.

Low: Partnerships are driven by our end customers. If a customer selects certain tools, we have to make it work. Eventually, with a broader customer base we have to make those tools available across a mass market. But it’s a process that’s driven by our lead customers.

Costello: In some cases, you can use your partners’ competencies out of the box. It’s more of a transactional relationship. The value the partners bring each other is by having tighter hooks. That requires advanced cooperation and sharing of information so that both can take advantage of it. That changes how much engagement you have with the customers, and it could change over time.

SE: One element that has become important in the IoT is security, and it is filtering back into other areas, as well. How do you secure your supply chain?

Smith: We have supported things in the network already. However, when you talk about security there are so many different elements to that conversation—digital rights, counterfeit parts—that we have to figure out which parts concern us. It’s probably not 100% overlap with our customers’ concerns. So our customers have to put different parts out to different partners.

Costello: There’s a lot of interest. Part of the challenge is that we’re only a chip company. The first point of attack will likely be through the software that sits on top of the chip, which requires a more integrated view of what’s going on. That ties into the partnership. We provide hooks for different types of security. We’re hearing the same kinds of issues about the supply chain. When you’re dealing with a global supply chain, how do you deal with that?

Ligthart: Our security is directly related to our partners because we ship source code. Once it has left the building, it’s on the other company’s server. It’s purely a trust thing.

Low: The elements we have introduced include electrical fuses and one-time programmable. It used to be that you could easily identify counterfeit parts, but it has gone to a new level of sophistication. There is no physical evidence. We now integrate that into our process. Working with IP companies, we can jointly promote this to our end customers. There are even more extreme solutions where they go through more rigorous tests.