In this NYSE Floor Talk interview, host Judy Shaw speaks with Andrew Feldman, co‑founder and CEO of Cerebras Systems, about the company’s wafer‑scale processors and AI supercomputers designed specifically for generative AI workloads. Feldman explains how Cerebras identified a market need in 2016, built a chip 50‑100 times faster than traditional GPUs, and optimized it for AI training. He contrasts Cerebras’ dedicated AI engine with NVIDIA’s GPU‑based approach, emphasizing speed and ease of use. The conversation then moves to Cerebras’ recent strategic partnership with the UAE‑based G42, which resulted in the deployment of Condor Galaxy One – the world’s largest AI supercomputer – and plans to build nine such systems. Feldman also discusses the company’s open‑source initiatives, notably an Arabic large language model aimed at democratizing AI access for 400 million speakers. Finally, he highlights real‑world applications across sectors such as life sciences (GlaxoSmithKline), energy (Total), and academic research (KAUST), underscoring Cerebras’ belief that AI can solve societal challenges in healthcare, climate, and beyond.
Chapters
I'm Judy Shaw for NYC Floor Talk. Joining me today is Andrew Feldman. He is co-founder and CEO at Cerebus Systems, and he's joining me today at NYSE Pacific. Andrew, it is wonderful to have you here. Thanks for joining me today.
Judy, thank you so much for having us.
My pleasure. So now, Andrew, the generative AI industry exploded in 2023 with the introduction of ChatGPT, and there's been a lot of innovation and focus on AI compute. Tell us about your wafer-scale processors and supercomputers, which I understand are custom-built for generative AI.
Yeah, thank you, Judy. We, we saw an opportunity in 2016 to build a processor, a new type of computer engine for generative AI. And we founded a company. All our effort and focus and drive is for this one application. And the market unfolded very much as we thought it would. We built a chip that was 50 times bigger, that had trillions of more transistors. And all of it was optimized for this single purpose of training AI workloads faster and easier.
Now, there's been a lot of buzz in the investor community around NVIDIA. Can you tell me or explain Sariya's current role in the AI compute market and how your solutions compare to NVIDIA GPUs?
Sure. Clearly they're Goliath and we're David. And we're just lining up stones in this battle with, with Goliath. NVIDIA makes graphics processing units and they were originally aimed at the graphics market and they've tried to move them to the AI market. We built a dedicated part optimized for AI. We're 50, between 50 and 100 times faster. We're vastly easier to use. And that means that practitioners can test more ideas in less time.
All right. Now, last month you announced an international strategic partnership with G42, including the deployment of Condor Galaxy One, which is the world's largest AI supercomputer. So tell me about that partnership and what it means for Cerebras and the AI industry on a global scale.
Sure. Well, we were really proud to, to build a partnership with, with G42, and they shared our vision that we could together build some of the largest AI supercomputers in the world. It's called Condor Galaxy One because we intend to build 9 such supercomputers. We thought that together we could build enough compute to change the worldwide inventory of AI compute. We've already done an enormous amount of interesting work on Condor Galaxy One. We've trained models that are in the open source community. We've trained models that are being used in the private sector by Fortune 50 companies, by little tiny startups, and all of it has moved the industry forward.
All right. Now, you recently released several open source models and datasets. So tell me, how does this fit into your broader mission to democratize AI?
Well, democratizing AI is sort of a buzzword that you hear a lot in our industry. I think by making models more available, by making them cheaper, by putting them in people's natural language. We just announced in combination with our partners, a large language model in Arabic. Now, there are 400 million Arabic speakers across 25 nations worldwide. Now they have a model that captures Arabic. We are constantly thinking about how to bring this extraordinary technology to more people around the world, to students, to academics, to people in different nations and with different experiences. All right.
And finally, Andrew, tell me, how are you and your customers advancing AI in various industries like healthcare, energy, climate, and more? And what societal problems are you trying to solve?
We, we have customers in life sciences. We've announced that GlaxoSmithKline is doing research using our equipment. We've announced customers in the oil and gas industry, now called energy, with Total. Basic research is being done on our systems around around the world, including by universities like KAUST in Saudi Arabia. All of them are in pursuit of sort of fundamental advances in science. Our belief, and we hold this very close to our heart, is that AI can change the way we live and work and play. And each of these industries, and the healthcare industry has vast amounts of data that haven't been put to use. To improve our healthcare outcomes. And we're collaborating with companies around the world to use AI to improve healthcare outcomes.
All right. Well, Andrew, it's been wonderful to talk with you. Thanks for joining me on Floor Talk today at NYSE Pacific.
It's a pleasure to be here. Thank you for having me.
