Mignano believes the AI infrastructure buildout is largely complete, with future value creation moving to the application layer, similar to the internet's evolution from broadband to apps. He suggests that frontier labs will vertically integrate down to inference, not up to applications, due to market heterogeneity and the need for specialized startups. USV's investment thesis focuses on reinventing workflows, maximizing token spend for coding, and betting on energy infrastructure like small nuclear reactors and micro data centers.
AI's next value wave is in applications, not core infrastructure, but labs will control inference; energy remains a critical, early-stage investment for long-term compute.