Modeling the future
I applaud the Optoelectronics Industry Development Association (OIDA) for its efforts to bring a workable foundry model to photonics (see “OIDA explores optical foundry model” on page 1). While clearly such a model-whatever it may turn out to be-wouldn’t solve all of the financial problems optoelectronic component suppliers currently face, it would at least offer existing companies an alternative to shouldering the burden of an underused fab. It would also open the door for startups and other entrepreneurs to explore ways of bringing a new economic paradigm to the component space.
As the article makes clear, the obstacles before the attainment of a photonic foundry model are significant. From the technology end, the lack of standard processes for laser and other optical component production hinders foundry economics, in that it becomes well-nigh impossible to set up a production line that will handle multiple customers efficiently. As Michael Lebby points out, the photonics industry will also need a set of software tools that will enable foundry-friendly design. The question of whether a foundry model will have multiple layers-such as a group of suppliers to provide epitaxy and another to provide fabrication-remains unanswered as well.
However, like most technically driven issues, these problems should disappear with time and effort. The most daunting obstacle to a foundry model may be the attitudes of the companies that OIDA hopes may benefit. That’s because I’m not entirely sure that a significant amount of optical component companies want to go fabless.
Notice that I said “a significant amount.” One issue with the current foundry discussion is the belief in some minds that to be viable, the majority of existing companies suddenly must decide to jettison their in-house facilities. While the story points out why comparing the silicon industry to photonics represents flawed thinking, one parallel exists that I buy: Just as sources tell me that as much as 80% of silicon ICs don’t come from independent foundries, a photonics foundry model doesn’t have to capture even half of the market to form a viable alternative to the present way of doing business. The foundry model clearly won’t work for everyone, or even every product line in a company’s portfolio. Instances will always remain in which high performance, customization, or some other attribute will require specific expertise that a “standard” foundry won’t be able to provide.
However, the problem now is that the vast majority of optoelectronic component suppliers believe that everything they produce requires such unique expertise. The largest question OIDA may face is how to proceed with a photonic foundry model when so many companies view their fabrication processes as competitive differentiators. Why would they want to help create standard alternatives to their products? I can also imagine companies doubting how they would manage a shift from an in-house facility to a third-party foundry. What do they do with the in-house facilities, particularly when most other companies already have more foundry capacity than they need? Can they afford to just shut them down?
Some foundry proponents may believe that market dynamics will force a critical mass of companies to realize that if they don’t switch, they’ll go out of business. This critical mass seems to me to be a long way from forming. The problem is that if you’ve been using your in-house processes as a competitive differentiator, it’s tough to give that up without another differentiator to take its place. And Asian companies have already staked out the option of low cost, in most instances. Even if they do switch, they are likely to transfer their proprietary processes to an outside facility only with the agreement that their intellectual property remains confidential. How such moves help establish standard processes is beyond me.
Instead, I believe the impetus toward a foundry model will come from foundry companies themselves, who will agree to offer a set of standard processes, working with startups willing to take advantage of these processes. The article mentions companies such as Eblana and Syntune that have designed their products to work within a foundry model. Similar companies are probably in stealth mode. Of course, Intel and others have reported advances in the pursuit of silicon photonics. Should these efforts bear fruit, the photonics foundry model may indeed look like that of silicon after all.
I agree with OIDA that in the future, the photonics industry should and will have a viable foundry model. However, I believe it is future companies, rather than existing ones, that are likely to make it happen.