Cerebras CEO Andrew Feldman: Inside the Largest Chip Ever Built | TomorrowX Summit
The episode goes into how a chipmaker describes the value of fast tokens, the economics of compute and the shift from an OpenAI contract to new capacity, which is the kind of operating detail behind AI infrastructure exposure. It is a management interview rather than a market call, so it is useful as background on how one supplier frames demand and product cadence.
Summary
By the alternatives in wealth deskCerebras co-founder and CEO Andrew Feldman speaks with Gavin Baker of Atreides on the Arena Stage at the TomorrowX Summit in Austin about the company's wafer-scale chip, its OpenAI partnership and disaggregated inference, where prefill runs on GPUs and decode runs on Cerebras hardware. He explains the first-principles case for wafer-scale computing with SRAM, the CS-4, CS-5 and CS-6 roadmap, and how the company sells both hardware and cloud capacity. The conversation closes with reflections on Jensen Huang, AI tutoring and data centers as neighbors.
- Feldman sets out disaggregated inference with prefill on GPUs and decode on Cerebras.
- He argues that fast tokens on frontier models are the most valuable tokens to sell.
- The company runs two go-to-market motions: hardware and cloud.
- The roadmap runs from CS-4 to CS-5 and CS-6.
Chapters
From the publisher’s description| Time | Chapter | Length |
|---|---|---|
| Cold open | 1 min | |
| Welcome | 1 min | |
| The OpenAI relationship | 2 min | |
| Why fast tokens are the most valuable | 1 min | |
| What compute is worth today | 2 min | |
| Disaggregated inference | 4 min | |
| Why wafer-scale computing | 6 min | |
| SRAM and memory bandwidth | 2 min | |
| Pairing with every GPU | 1 min | |
| The roadmap: CS-4, CS-5, CS-6 | 2 min | |
| From the OpenAI contract to new capacity | 2 min | |
| Hardware and cloud: two ways to buy | 4 min | |
| On Jensen Huang | 2 min | |
| Winning with the product | 1 min | |
| How AI is making lives better today | 2 min | |
| Being a good neighbor | 2 min | |
| AI in three years | 1 min |
Similar items
Same topicGet the weekly 3 to watch
Free, in the Morning BriefEvery week the Morning Brief picks three videos and podcasts worth your time, with why each one matters.
Free. Every morning. Unsubscribe from any edition in one click.
Videos and podcasts are embedded from their publishers. Selections and summaries are editorial and are not investment advice.




