Nigel Toon, Graphcore CEO joins Yahoo Finance Live to discuss how the AI chipmaker is faring during the pandemic and break down whether or not the company plans to go public in 2021.
- UK-based chip maker Graphcore has flown a bit under the radar this year, but that unlikely will be the case in 2021, as the Nvidia rival this week raised $222 million in fresh capital. The company is now valued at $2.77 billion. Let's welcome in Graphcore co-founder and CEO Nigel Toon for more on the company's plans.
Nigel, good to have you here this morning. So for those viewers not too familiar with Graphcore, how is what you're doing different compared to some of the more names that they are familiar with, like an AMD and Nvidia?
NIGEL TOON: Yeah, we have developed a completely new type of processor built from the ground up especially for AI. We have many more processor cores inside the chip. We have a very different interconnect between those processor cores. And it really delivers a very different set of performance capabilities for customers, really accelerates for a wide range of AI applications, just built exactly for that application.
- So Nigel, this is a nice chunk of money that you did, in fact, raise this week. Does how you use that money change if an Nvidia does go out there and they're able to get that deal, that $40 billion deal, confirmed with ARM.
NIGEL TOON: Well, I think for us, we are absolutely focused on running as fast as we can. We're driving a strong road map. We've already built the world's most sophisticated microprocessor. We have 59 billion transistors on our chip. We're delivering much higher performance than the Nvidia GPUs.
We're obviously concerned about Nvidia buying ARM, not so much for ourselves, but more just because it's not good for competition. We don't think it's good for the industry. We think it cuts off access to this advanced technology, potentially, for other companies in the semiconductor space. And I think it will be bad for competition overall.
- And then, Nigel, going back to the technology itself, I'm curious, maybe for a lay audience, I mean, I don't really know anything about semiconductors and chips and all that, what is the difference between the computing power that's necessary to run some of the AI applications that we hear so much buzz about? And how much different does what you guys are making need to be? How much more power, I guess, does it need to have than what is out on the market relative to the goals, again, that I think the real evangelists for AI see as possible in the next couple of decades?
NIGEL TOON: Yeah. Well, I think the big thing to realize is that for 75 years, we have been telling computers what to do step by step in a program. And now the computers are learning from data. And that is a completely different type of compute.
It requires a completely different type of processing solution. And you can't just take what exists today and apply it. You can get some performance, but really, you need to sit down and create a completely new type of processor for this new market. And that's what Graphcore has done. And that's why we're able to get this much higher level of performance. This is a processor designed to learn from data, designed to support artificial intelligence.
JULIE HYMAN: So Nigel, it's Julie here. That sounds expensive, again, to a layperson, both the R&D to develop that kind of chip, and then perhaps the production of it. So talk to me about your cost structure. It's my understanding that you are not yet profitable. What would be the critical mass that you would need to reach in terms of your sales in order to get there?
NIGEL TOON: Well, we have about 500 people in the company today. We're growing still, very, very fast. We have raised so far since the company started just four years ago, we've raised over $700 million. But we actually have nearly $450 million of cash at the bank.
So we have all the resources we need to be able to grow our revenues and to really be able to build a very large business here. And obviously, these very large investors who are coming in and supporting the company, they can see the potential behind what we're doing and the huge market opportunity that is in front of us. So this is very much about running as fast as we can, building this incredible technology, and growing to become a major player in the marketplace.
- Nigel, you're no failure to doing deals. You sold a business, I believe, back in 2011 to Nvidia. What are you more inclined to do? Do you see Graphcore being part of another major chip maker, or is 2021 the year that you continue to build your business and then IPO later in the year?
NIGEL TOON: I think 2021 would be too early for us to IPO. We need to really knuckle down and build our business over the next 12, 18 months. And then we'll see where we are at that stage.
But this is a massive market opportunity. And I think there is the opportunity to build a business that is designing and supplying these new types of processors for what will be the most important computing workload on the planet. AI is going to dominate a lot of new application areas.
And there is the opportunity to create a specialist company that is focused in on that, just as we saw in mobile, just as we saw originally with computers. This is a new paradigm and new companies will emerge here. And I think Graphcore is at the leading wave of that.
- Well, we look forward to following your journey for sure in the new year. Graphcore co-founder and CEO Nigel Toon. Happy new year, we'll talk to you soon.
NIGEL TOON: Happy new year. Thank you very much.