The interview, recorded at Web Summit Qatar for CNBC TV18, features Andrew Feldman, CEO of Cerebras, discussing the company’s flagship AI hardware and broader strategy. Feldman explains that Cerebras builds extremely large AI chips—about 56 times larger than typical GPUs—and a full system that accelerates both training and inference while using a fraction of the power, claiming roughly 70× faster inference performance. The company hosts a variety of open‑source (Llama, Deepseek) and proprietary (Mistral, Perplexity) models on its cloud platform, attributing speed gains to reduced data movement on the massive chip. He emphasizes the need for users to apply common sense when interacting with AI, suggesting cross‑checking answers across models. On governance, Feldman notes that both industry and government share responsibility for AI guardrails, citing the need to restrict dangerous content (e.g., nuclear weapon design) and protect customer data. The conversation shifts to Cerebras’ India operations: currently ~70 staff in Bangalore, plans to double the Bangalore office, expand to Hyderabad, and partner with G42 to deploy a large AI supercomputer cluster, targeting 350‑400 employees and 10‑12 exaflops of compute within two years. Finally, Feldman confirms that Cerebras has filed an S‑1 for an IPO, hinting at upcoming announcements.
Chapters
Web Summit Qatar, in association with CNBC TV18.
We're here at the Web Summit in Qatar and it's all about AI and innovation. And today we have with us Andrew Feldman of Cerebra Systems. Andrew, thank you so much for taking time out and being here with us at the CNBC TV18 studio in Doha.
Thank you so much for having me. What a fun event.
Absolutely. And, you know, Before we get into the nitty-gritties of everything, if you could sort of explain to us what it is you do at Cerebras.
Sure. We build big, fast computers whose job is to accelerate AI. AI has two parts. It has training, which is as we make AI, and inference, which is how we use AI. And we built a chip and then a system and the entire computer that makes both training and inference faster and easier and use less power.
Right. And, you know, sort of to explain that to us a little bit further, you know, when we talk about chips, the first thing that comes to mind is NVIDIA, right? When it's AI and chips. But if you could tell us, how are you doing this differently?
Sure. There are lots of ways to make chips. We chose a way that, that focused on the exact problem posed by AI. And the hard part about AI for a computer is the amount of data that needs to be moved. And so we built a very big chip. In fact, we built the biggest chip in the history of the compute industry. And that allows us to move data less often, to keep it on chip, to work on it faster, and to use less power.
And you know, that, that was what I was going to come to because when I was reading about Cerebras, the one thing I saw is about the size of the chip that you make, right? So, you know, how much faster is it? How much more efficient is it? Like what makes it different?
Sure. So the chip is about 56 times larger than any other chip made. So compared to an NVIDIA chip, it's 56 times larger. It's order 70 times faster in inference, uses a fraction of the power. So this one chip is faster, it uses less power, and it makes your AI experience more enjoyable.
So it's faster, it uses less power, and it's more enjoyable. And as we know, you now host Deepseek on your platform, right? And it's, it's, it's making waves. It's what people are talking about. And the, and the thing about Deepseek is that it's so much cheaper than, you know, what a ChatGPT would do or how you would do that. So why is that the case?
So we host a whole range of models. We host open source models like Llama from Meta, Deepseek from Deepseek, and we also host closed source models for particular customers. Like Mistral and Perplexity and many others. And all of those are served to customers via our cloud. What makes it faster? It's a good question. By being a bigger chip, we have to move information less far. Right. And that's what takes time in compute. And so these big chips move information more quickly, using less power. And as a result, all these different models, whether it's Deepseek or Llama or Perplexity or Le Chat from Mistral, all of them provide a better user experience.
Right. And, you know, one of the things that comes up is also about accuracy, right? Like, for instance, on Deepseek, there are certain questions that you can't sort of ask or you won't get an answer to. So how does someone circumvent that, you know, when so much information is being passed through?
Sure. It's a good question. When you use AI, I encourage you not to give up your common sense. And so I think these are amazingly powerful things, but they don't replace our common sense. So go ahead and think a little bit. Use these models, make progress. You might even compare the answers from two models.
Right.
Right. From Llama, which comes from Meta, and Deepseek, and see what is the similarity. And if they're, if they're both giving very similar answers, you can have more confidence.
Yeah. So common sense is what we need to use when we're using all of these AI models, right? I think that's right. But another question that comes up is then about privacy and security, right? When it comes to AI, when it comes to how you're using it and the information that's being passed through. So what about guardrails when it comes to AI? Who needs to put that up? Does government need to work with corporates and figure this out? Or what sort of guardrails do you think need to come into play to sort of make AI more safe for everybody to use?
Well, I think that's a bigger question than we have time for today, but I think the following: I think government has a role, but I think government wants to regulate, right? And as entrepreneurs, we want to innovate, and we'd like the government out of innovation. I think that we are, as a community, careful and thoughtful. I think we are aware that powerful technologies can be used for good and for evil.
Right.
I think the, uh, there is a responsibility on the consumer. Don't put your Social Security number in. Don't put your tax information in. You use your, you use your common sense at every stage, just as you would any other technology. Right? Just like when someone, when a Nigerian prince asks you to send them money, don't do that. Right? It's the same. It's the same for, for AI as it is for any other part of the internet.
So you're saying that everybody has to be responsible for how they use AI because then it helps regulate it much better and much—
They do. And we're not abdicating responsibility as the industry. We need to put in guardrails. We need to make it safer. And our users need to use some common sense.
And what sort of guardrails would these be, though?
Well, I think there's some questions we'd like not answered. Yes, right. Weapons making, correct. For example, we should make it harder to gain information about the chemistry of nuclear weapons, about all sorts of, uh, uh, war-making techniques that I, I think are probably better left outside of your chat. But I, I think open information, many of that information can be found in the library. I think it's okay if you make people do a little bit of work. You got to get up off your chair. You got to get a library card. You got to go to the library. I think that we should be very careful with our customers' data. Obviously, we should destroy it quickly when we don't need it. If we have to save it, we should put up a whole set of thoughtful security around our customers' data. It should be treated with respect.
So everything needs to be treated with respect at this point. But, you know, I also want to come to the fact that Cerebras has an office in India, right?
We do.
And our audience would love to know, what are your plans for India and where does India stand right now in this AI race? Because I know the government's been pushing for chips and, you know, manufacturing of chips in India. So where does India stand in your plans at this point?
Well, today we have about 70 people in, in Bangalore and we're looking to double our, our, our office this year. We just took more more office space. And so we're looking to continue our very rapid growth. Look, India has some of the best universities in the world, both their IITs and their NITs. These are extraordinary universities. They have people who are, who are driven and some of the great engineers. The, the environment is ripe for, for India entrepreneurship and for India to play a leading role in AI.
Absolutely. And India is ripe for that opportunity. Are you opening more offices in Bangalore? Are you looking at other places?
We are. Right now we're growing our office in Bangalore. Right. We do have a handful of guys in Hyderabad. Not just because I like the biryani, but because— I mean, the biryani is phenomenal. And also we are looking for a site to deploy together with our partner G42, a giant supercomputer cluster.
Right.
In India, one of the largest supercomputer clusters for AI in the entire subcontinent. And so we'd like to make a commitment of size, of people, of compute. I think there's tremendous opportunity there.
So your India plan for the next 2 years?
Oh, I think in 2 years we should have 350 or 400 people in our India office. I think we'd like to have 10, 12 exaflops of Cerebras compute in a single site, providing one of the largest AI supercomputers in India. I think I'd like to get there sometime in much less than 2 years.
We hope to see you in India.
It would be enjoyable. I would like that very much.
And, you know, before I let you go, I have to ask you, I know that Cerebras has been in talks for an IPO. What's going on on that front?
Look, we filed our S-1, and beyond that, as you know, I'm not permitted to say. I thank you for asking nonetheless. But good stuff to come.
Absolutely. All right, Andrew, thank you so much. It's been a pleasure having you on our show today.
Thank you so much for having me. Be well now.
And with that, we'll be back with more news and updates.
