Why this matters in practice
Infrastructure choices affect latency, economics, and which AI workflows are practical; teams should connect compute decisions to the operating experience they need.
Chapters
- 00:00Intro
- 01:06What Cerebras Actually Builds
- 03:41Before LLMs Existed
- 06:09The GPT Wake-Up Call
- 10:45First Principles of AI Compute
- 15:40One Giant Chip
- 18:43So Why Do We Still Use GPUs?
- 20:59Why Inference Took Over
- 22:27Where Fast Tokens Win
- 25:07Is AI Infra a Bubble?
- 29:56The Memory Shortage, Explained
- 32:35The Energy Problem
- 35:34Rolling Their Own Data Centers
- 38:20What a Chip Actually Costs
- 40:18AI Designing AI Chips
- 41:40The Supply Chain Reality
- 43:45Cheap Tokens vs Fast Tokens
- 46:11Why Every Millisecond Matters
- 48:56Inside the OpenAI Deal
- 51:56The Gigawatt Future
- 54:00Selling to the Government
- 58:36Exporting the US AI Stack
- 01:03:17Should We Sell Chips to China?
- 01:06:00Why Europe Is Falling Behind
- 01:07:59Why Enterprise AI Moves Slow
- 01:13:05"I Haven't Read Code in Months"
- 01:14:32The Next Cerebras Chip
- 01:16:50Closing Thoughts
Related topics
More episodes
- Spencer Whitman - Gray Swan AI's $200M Plan to Secure AI SystemsSpencer Whitman — Gray Swan AI
- Tony Gentilcore - Glean, the $7.2B Startup Sam Altman Warned Investors AboutTony Gentilcore — Glean
- Russ Salakhutdinov - Kimi K3 CEO’s PhD Advisor Predicts the Future of AI AgentsRuss Salakhutdinov — Sooth Labs
