Summary
Perplexity co-founder and CEO Aravind Srinivas joins Harry Stebbings on 20VC for a wide-ranging interview covering competitive strategy, infrastructure constraints, and bold predictions about where AI value will accrue. Srinivas argues that Perplexity reshaped Google’s product roadmap more than any internal Google PM ever did—pointing to Google AI Mode’s near-identical UI (inline citations, bolded text, follow-up suggestions) as evidence—and explains why he has shifted from aggressive public messaging to a quieter offense-first posture.
The conversation digs into macro infrastructure dynamics. Srinivas contends that power and high-bandwidth memory are the true bottlenecks to AI scaling, and makes a pointed prediction: Micron, as the dominant supplier of HBM, could surpass Meta’s market cap within 6–12 months given that whoever controls the bottleneck commands the price. He also offers a counterintuitive read on U.S. export controls, arguing they may have inadvertently accelerated China’s domestic AI ecosystem by forcing self-sufficiency.
Srinivas touches on the resurgence of CPU relevance as agentic workloads shift compute demand away from pure GPU inference, why he believes cloud compute providers like CoreWeave and Nebius can build durable businesses if they solve the power-supply equation, and why resistance to new data center construction will intensify. This is one of the sharper CEO-level perspectives on AI competitive dynamics and infrastructure economics available in podcast form.
📺 Source: 20VC with Harry Stebbings · Published June 15, 2026
🏷️ Format: Interview
