SAN JOSE, Calif. — From time to time, often at the end of the year, we get an opportunity to talk about issues of the day with Henry Samueli, Broadcom's co-founder and chief technologist. Earlier this month we asked his views on the outlook for Moore's Law, the platform wars in the Internet of Things, and more. Here's a digest of the interview.
EE Times: Has Broadcom found a path to designing with the emerging FinFET processes?
Henry Samueli: FinFET technology is becoming mature at multiple foundries. We're working on designs in 16nm and expect to be taping out several chips next year. The leading foundries are neck-and-neck aggressively working on it.
We've had a long history with TSMC — they were our fist foundry, and we've been working with them 23 years — so they are typically our partner of choice, but we are working with other foundries as well. As soon as their advanced nodes in 16/14nm become available, we'll take a look. We are open to anyone.
Moore's Law is marching on. We'll probably do test chips in 10nm FinFETs next year, and there's talk about how to get to 7nm, so its highly likely we'll get there — maybe well even get to 5nm but that's still not obvious. So there are two to three more generations left. Beyond 5nm is where a lot of skepticism comes into play. The barrers beyond 5nm look pretty tough.
The main thing is the cost curve has flipped upside down. We still get denser, higher performance and lower power chips but unfortunately they are more expensive. This is the first time in the history of the semiconductor business that we make something better and — amazingly enough — it's actually more expensive. That's the only downside.
It puts the onus on us to be more critical of which chips we move to advanced nodes and which will remain in legacy processes for cost reasons. Networking switches and multicore processors and high- density memories have to go to 16 and 10nm as fast as they can because they can support higher costs, but consumer-oriented products will stay in 40 and 28nm processes for quite a while. We can live with that.
Next page: The outlook in engineering
EET: What impact do you see the emerging hardware accelerators having on the engineering profession?
Samueli: I think it's fantastic. It's stimulating the entire community, lowering the hurdles to become an entrepreneur. It's crowd-sourced engineering basically.
We are happy to provide platforms, especially in IoT that a two or three person startup can run with. The onus is on us to provide the platforms, semiconductors and middleware and make it easy for them to write their apps and create IoT devices, whatever they might be.
It's a more challenging problem at the chip level. The entry for chips is very high, especially for advanced technology it's prohibitively high because semiconductors is a mature industry.
EET: So what's the startup pipeline like in semiconductors these days?
Samueli: It has slowed. There's no question it's slowed dramatically over the last few years. The cost of making a new chip is tens to hundreds of millions of dollars.
There are still pockets of chip startups. Israel is a great example. There's an amazing startup culture there these days.
On average the number of chip startups today is lower, so our acquisition appetite doesn't get filled as easily. We have acquired more than companies, but the pace has definitely slowed down. We love to find good talent and technology.
To read the rest of this article, visit EBN sister site EETimes.