Summary
This a16z podcast episode features an investor discussion on why AI demand is consistently outrunning compute supply, and what that imbalance means for the market. The conversation centers on the acceleration observed across AI deployments in July and August 2026, with the guests noting that despite public stock volatility, every major player — OpenAI, Anthropic, open-source models, and inference clouds — appears to be growing simultaneously.
A key thread is the counterintuitive scenario where token prices could rise rather than fall. If data center construction is constrained by permitting and regulatory pushback, the resulting compute scarcity could produce pricing power for frontier labs and create a form of compute inequality where only well-capitalized companies and individuals can access the best AI. The discussion references Kimi’s license structure, which stipulates a 30% revenue share from any downstream use, as an underappreciated constraint in the open-source ecosystem.
The guests also analyze Anthropic’s quiet period ahead of its anticipated IPO, speculating that its next disclosure will show a revenue reacceleration, and debate the “everyone wins” thesis popularized by investor Eric Vishria — the idea that Anthropic, OpenAI, Meta, Google, NeoCloud inference providers, and open-source all have room to grow in parallel. The episode provides a useful framework for thinking about the infrastructure economics and competitive dynamics shaping the current AI investment landscape.
📺 Source: a16z · Published August 31, 2026
🏷️ Format: Podcast
