Legendary software engineer Jim Keller on the significance of an alliance with Rapidus

Q.

You are known in Japan as a legendary engineer for your development of various innovative chips. How did you become interested in semiconductors, and why have you been so successful?

A.

I started working on semiconductors in college because my advisor at Penn State ran a small semiconductor lab, and I took all of the courses on semiconductor physics. The cool thing in the lab was we got to literally make transistors. To be able to not only study electrical engineering but to actually build electronics is great. And then when I joined a semiconductor group at Digital Equipment I was there as a computer architect, but we also co-designed the fab and the process with the architecture. And that was really interesting because we made some really interesting trade-offs. Then over the years, I've been involved with a number of projects where we're doing computer design but we're also working with process development. So it just keeps coming back that even when you build computers, we always have to keep in mind what we're building it on and that's been really interesting. And it's basically been going on for me for over 40 years.

Q.

Semiconductors have changed society in many ways. You are developing semiconductors for AI that have capabilities beyond GPUs. What kind of changes will your new AI chips bring to society and the market?

A.

AI is on a very steep technology curve. You can read in the news that the current high-end AI is very expensive; people are talking about it, and companies are raising hundreds of millions and billions of dollars to build it. That means that AI will be for big things and that users can only use it if they're running on somebody else's computer. Tenstorrent would like to greatly reduce the cost of AI and make AI available to many more companies to build into their own products. So we've been focused on superefficient low-cost technology choices and licensing and partnering with quite a number of companies so they can build AI into their own products, and I think that will let AI be more part of normal product development and use.

Q.

Japan was a leader in the global semiconductor industry until the 1980s but is no longer in that position. What are your views on Japan's current and future status in the semiconductor industry?

A.

I don't know that Japan stopped being one of the leaders of technology development. They provide equipment in all the factories. Japan is one of the world's leaders on materials used in semiconductors. Obviously, PlayStation, Nintendo, lots of the top gaming companies, gaming computers, and product development is done in Japan, and Japan has been one of the leaders in HPC computing continuously through that whole period.[*] Some of the top computer designers in the world are from Japan and went to Japanese universities, including some who work at Tenstorrent. So we think the possibilities are quite open and interesting, and one reason we've been working in Japan is because the government and the many businesses are very interested in the next big wave of innovation, and we think the future is bright.

Q.

What was the reason for your decision to form an alliance with Rapidus, which has no track record in this field? What specific actions do you plan to take to develop semiconductors for AI based on the 2 nm generation agreed on with Rapidus?

A.

Rapidus is interesting for a bunch of reasons. First of all, if you want to do something new, you need to work with some new players. That's really important. Second, the CEO and the team are very experienced, and they've done a lot of technology development. And third is one thing they said is that they wanted to really focus on speed-the speed of building the fab, the speed of the time to make a chip and the speed to get to 2 nm. I think that's really, really important, and when we got to meet them it was really good conversations, and we thought we could work together. One of the things I want to do is be able to build computers faster, and having a foundry partner that also believes we need to do this faster is really important.

Q.

In Japan, the challenge is to attract talented people to the semiconductor industry. What are the important aspects of human resource development related to semiconductors? Could you give your message to the youth of Japan?

A.

A career in computer design, semiconductors and chip design-it's a long process. You need to attract talented students out of high school, to have colleges that really teach people the basics like math, physics, chemistry. There are some basics like programming and computer architecture that people need to really learn well. Then you need a good path for students to be interns at companies that are doing new and innovative things and then a place for people to go and have valuable careers. And you can't leave any piece of that out. We hire dozens of interns every year, they often return multiple times and join us when they start their careers. Being engaged at the university level through internships and partnerships is what makes that happen. To the youth of Japan, I would say I am a technology optimist. We are building computers and tools that are going to make life essentially more interesting, and there're many, many opportunities to do design going forward. Find companies and places that are doing new things with new groups of people and new arrangements. When companies become very successful, they become very set in their ways. Every generation needs to stake out some new territory and do something new, so make some new partnerships and make some new friends.