“The scaling laws were going to drive everything — and we understood that back in 2020, 2021, before the ChatGPT moment. Computing dominates at scale.”
Four CEOs building the AI infrastructure stack — CoreWeave's Michael Intrator, Perplexity's Aravind Srinivas, and two others — make the case at Nvidia's GTC conference that the AI buildout is nowhere near its ceiling. Intrator reveals CoreWeave's "box" financing model, which pays back principal and interest within 2.5 years of a 5-year contract and has already cut their cost of capital by 600 basis points, while dismissing GPU obsolescence fears with a blunt fact: clients are signing 5- and 6-year deals. Srinivas reframes Perplexity not as a search tool but as an AI orchestration layer — a conductor directing hundreds of specialized models — and announces a local/cloud hybrid architecture built around the Mac Mini that keeps private data off third-party servers. If you want a ground-level read on where AI infrastructure money is actually flowing and what the next wave of AI products looks like from the people building them, this episode delivers specifics that most coverage misses.
Jason Calacanis welcomes listeners to Nvidia's GTC conference and introduces the upcoming interviews with four AI CEOs.
No quotes extracted for this segment.