Oct 20, 2026 · 20 min
AI infrastructure race shifts from chips to inference
Eric Schmidt's $80 Billion Anthropic Bet
The episode connects faster models, open-source inference software, and major AI financing into one contest over cost and scale.
- 1DeepSeek’s DeepInfer open-sources an inference stack designed to improve chip efficiency and reduce compute costs.
- 2Inception positions Mercury 3’s diffusion architecture as a faster alternative to conventional autoregressive language models.
- 3Eric Schmidt’s Anthropic stake could turn AI investment exposure into a major IPO windfall.
Don't miss
The episode’s most consequential moment is the suggestion that Eric Schmidt’s Anthropic stake could generate a major IPO windfall.
The brief
Brian McCullough previews a technology cycle increasingly defined by efficiency: DeepSeek opens its inference stack, Inception touts faster models, and AI investors eye enormous returns.
DeepSeek’s DeepInfer targets the machinery beneath AI products, optimizing chip performance while reducing energy and compute costs; the move puts software efficiency alongside hardware in the spotlight.
Inception’s Mercury 3 represents a different route to speed, using diffusion-based language modeling to challenge the slower inference assumptions of conventional autoregressive systems.
The episode also widens the lens to Spotify’s paid podcast memberships and Cloudflare’s Clairic extraction model, showing AI and media businesses competing over monetization.
The standout financial angle is Eric Schmidt’s Anthropic stake, which could produce a major IPO windfall and underline how quickly AI progress is reshaping investment fortunes.
Taken together, the preview frames AI’s next contest around throughput, cost, distribution, and ownership—not simply which company builds the largest model.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.