Descriptions:
Recorded live at Nvidia’s annual GTC conference, this All-In Podcast session brings together four AI company CEOs for back-to-back interviews: Michael Intrator of CoreWeave, Aravind Srinivas of Perplexity, Arthur Mensch of Mistral, and the CEO of IREN. The result is a rare multi-perspective snapshot of where the AI infrastructure and application stack is heading, delivered by people actively building it.
Intrator traces CoreWeave’s unlikely origin story — starting as an algorithmic hedge fund that pivoted to Ethereum mining in 2017, weathering crypto winters, and eventually discovering that GPU compute could serve ML researchers. He credits donating A100s to open-source AI group EleutherAI as the “tuition” that taught CoreWeave how to run large-scale parallelized computing, and explains how those early academic relationships became the commercial foundation of the business.
The Perplexity segment covers Aravind Srinivas’s thesis that AI is becoming the operating system itself — users set high-level objectives rather than specific instructions, and the model orchestrates file systems, browsers, and connected hardware underneath. He argues that $10,000 local Linux workstations could soon make economic sense for heavy AI users, replacing hundreds of dollars per month in cloud subscription fees while offering privacy guarantees. The discussion connects to broader themes around on-device inference, the role of Linux in the AI era, and where consumer and enterprise AI spending is heading.
📺 Source: All-In Podcast · Published March 23, 2026
🏷️ Format: Interview
