Quote · All-In with Chamath, Jason, Sacks & Friedberg
Thomas Laffont: The $4T AI IPO Wave, 2026's Unicorn Economy, and the 10X Paradox
Where this was said
Segmenting AI markets and future impact
At 18:32 · chapter starts 15:33
Laffont introduces Cerebras — where he served on the board and led the Series B — as a cautionary and ultimately triumphant tale. The company spent years in what the slide labels 'construction': no new capital, grinding through hard technology development, with no guarantee of survival. Then came a massive OpenAI contract that tripled the company's value, and eventually an IPO. [1] — Thomas Laffont "Memory per user could quintuple: Coatue projects that the amount of memory required per user could quintuple as AI systems increasingly nee…" 22:20 The lesson isn't just about Cerebras — it's that the semiconductor sector broadly is on a generational run since the 2024 All-In Summit. Laffont then pivots to the memory thesis: as AI systems require increasingly personalized, persistent memory to deliver useful services — remembering your restaurant preferences, your schedule, your habits — the memory demand per user could quintuple. This, he argues, explains the explosive moves in memory company valuations and implies they still have room to run.
