Baseten's $13B bet on 100x inference demand — and open source as the antidote to centralised AI

Key points

Key takeaways from The Sachin and Adam Show interview with BaseTen co-founder and CEO Tuhin Srivastava (October 2026):

Inference is the industrial revolution. Every AI action — thinking, reading, doing — requires inference, and BaseTen now processes over 1 bn of these requests every day across 20 cloud providers, giving active-active resilience so traffic reroutes if any provider or region fails.

A 100x demand curve. Tuhin argues there is no way inference demand is not 100x or even 1,000x higher in five years: most of the world has not yet come online, making this a deliberately secular bet.

Multi-model, not either-or. BaseTen is betting on open source but also works with closed models — a multi-model future is the core thesis, because customers want whichever model gives the best cost, quality and latency per task.

Best open weights come from China. The team says the strongest open-source models today come from China, и open-weight models are now good enough that more companies start owning their own intelligence.

Base Labs, research in the open. Training co-heads Charlie O'Neal and Moody now run Base Labs, an independent research lab making open models as useful as possible — open-sourcing post-training environments so any customer can post-train on their own data, workflows and user signals.

Closed labs spend billions. Billions are being spent by frontier closed labs, and that gap risks becoming too big to close — open-sourcing the environments, not just weights, is how the field keeps an alternative alive.

Democracy of values. If one or two labs train every model, values, censorship and morality are decided by a few thousand people without an election — open source ensures values are partly voted in by everyone.

Closed-source lead is a canary. The six-month capability lead of closed models acts as a canary in the coal mine, giving society time to harden itself before the next-generation capabilities arrive.

Read more: YouTube