Challenges At Advanced Nodes

Experts at the table, part 3: Will 22nm FD-SOI fill a void or just be one more of many competing technologies?

popularity

Semiconductor Engineering sat down to discuss finFETs, 22nm FD-SOI and how the how the market will segment over the next few years with Marie Semeria, CEO of Leti; Patrick Soheili, vice president of product management and corporate development at eSilicon; Paul Boudre, CEO of Soitec; and Subramani Kengeri, vice president of global design solutions at GlobalFoundries. What follows are excerpts of that conversation. To view part one, click here. To view part two, click here

SE: There are a number of different flavors of every process these days—probably more than will hit mainstream adoption. How does this get sorted out? Is it just a price war?

Soheili: It’s not just a price war that clears them out. It’s also lack of investment or lack of ROI. If your margins are high but your ROI doesn’t meet your corporate acceptance rate, you’re going to get out of that market. If it’s not returning the type of money you’re expecting from a process or piece of innovation that you’re working on, it’s going to die because the market just didn’t adopt.

Kengeri: There are two parts to that. One is the baseline platform, the ULP (ultra low power), which is the mainstream workhouse. RFA (integrated RF and analog) will come in a little later. Those are the two major nodes.

SE: How about beyond FD-SOI?

Kengeri: We don’t have too many flavors like some of our competitors. There’s 28nm SLP. There’s one at 20nm and 14nm, and we will come up with one more.

SE: Does Leti believe one approach will win out over another or will they all work? Will it be vertical stacking, FD-SOI, or are you just placing bets across the board?

Semeria: There is a new opportunity regarding fully depleted SOI. Regarding 3D integration, it’s a way to integrate more and more functionality into a system. The end of Moore’s Law opens up a very large space. We started working on SOI more than 30 years ago. It’s a long way from there until today, which is a turning point for this technology. So all of this takes time. Over the next 20 years or 30 years we will see more changes in the industry.

Soheili: You’re not alone in not being able to call the future. The semiconductor industry is one of the few that has a negative ROIC (return on invested capital) over the last 30 years. The industry has made many bets for many industries at the wrong time and for the wrong reasons. We all think of markets that should exist. On paper and in presentation slides they look very good. But in actuality a lot of them don’t work out. We design processes and technologies and ecosystems and supply chains. And then you take a couple of guys who have made up for the sins of the industry out of that equation and it looks even worse.

SE: At 65nm you could take a design and move it from one foundry to the next. Are we getting to the point where the foundry processes are so different that you’re committed to working with one foundry?

Kengeri: That’s already happened. 40nm was perhaps the last node where you could move. After that, high k was so complex that there was no way different foundries could have similar processes—even to the extent where one ecosystem was on gate first, another was on gate last. Things got complicated after that. Customers now have to make a bet.

SE: Does that matter?

Kengeri: For some very high-volume customers, that matters, because now they have to invest a lot more in the design.

Soheili: It’s not just cost. It’s capacity. If you need to ship by Christmas, you have to do that or you’ll lose the market.

SE: How real is the Internet of Everything at this point?

Semeria: We are more and more solicited by industry for the Industrial Internet of Things (IIoT). I see this application appearing right now.

Boudre: We see very opportunistic applications at this point. We are talking about connecting things together, so there is some need for standardization. What is missing is power. It is very critical. I was driving to Paris with my son and he was wearing his standard watch, not his smart watch. I asked him why and he said he was traveling, and he didn’t want to bring everything to plug in his watch. This is a problem. Kids will love it. But we need to get it to the point where it is useful.

Soheili: I walked into a store the other day to look at a smart watch and I asked the salesman, “What’s the killer app for this so that I have to have one?” He scratched his head but couldn’t answer that question. We can connect things to each other, but what are we going to get out of that? In the semiconductor industry we’d love to put a chip in everything. The refrigerator needs one, the toaster needs one, the industrial controller needs two. Cars have 150; they should have 300. And then there are the software guys who can stitch it all together and deliver it. So if I have complete data on all my health indicators from my watch and the pill that I just took that happens to be an IoT device itself, what do I do with all that data and how does it all get analyzed? We’re far away from that.

SE: So does the technology drive the market, or does the market drive the technology with the IoT/IoE.

Soheili: It’s both.

Semeria: If you use the example of implantable medical devices, if we can develop a low-energy platform that uses some energy harvesting solution, it will change life in many ways.

SE: In the industrial IoT, though, it’s the market that’s driving the technology. In the consumer market, it may be the other way around, and many of these devices such as smart watches measuring sleep patterns will need to be incredibly efficient. Will 22nm FD-SOI add enough energy efficiency and performance?

Kengeri: When we define technology we don’t do that in isolation. We talk to customers, get their feedback, understand their road maps, figure out how to reduce their pain points. At the same time, you have to look at what’s possible, what’s manufacturable, and what the risks are. So we’ve put all of that together. Technology development takes about two years, and in the meantime we do technology co-optimization with some of our key customers. It’s very hard to say whether the market is driving it or technology is driving it, but we do collaborate.

Semeria: Technology will enable markets, but markets will select technology.

SE: Where are we in terms with energy harvesting, and will it show up with this technology?

Semeria: It’s definitely a new area where we can catch low amounts of energy, but from multiple sources. It’s clearly a new space for efficiency.

SE: Looking out at the market for FD-SOI, are you optimistic this will all work as planned, or is this just one more possible technology in a long list.

Soheili: From our perspective, the more technologies with full ecosystems, the better. It provides more options for our customers to make a better selection that allows them to stay in business and move to the next platform and the next technology. This is fantastic. We’re fully supportive of this, either as part of an ecosystem or as a service provider. When we talk to customers, they don’t always know what they want. Sometimes we have to rely on intuition to put a complete solution in front of them and show them how this can be developed into an idea or concept. A lot of the burden also falls on us.

Kengeri: We are very optimistic from the customer feedback we’ve seen. This isn’t just going to the next node and here is the yield, here is the cost and the performance. There is some additional feedback that comes with it.

Boudre: We are very optimistic, as well. We have been talking to customers for the past two years and very focused on bringing this platform to life and to manufacturing. Talking to end customers we know there is a need. We need to execute as planned, but if we do it will be a big success.

Semeria: We have to use this technology, and the best way to extend this technology is to make it accessible for more and more designers. More and more companies are switching to FD-SOI. That’s the best sign of beginning success.