xAI co-founder's River AI raises $1.1B to build AI you own

Igor Babuschkin — ex-DeepMind, ex-OpenAI, xAI co-founder — just raised $1.1 billion for River AI, led by General Catalyst and Amp PBC, with strategic money from Nvidia and AMD Ventures, plus Y Combinator and Temasek. That's a frontier-lab-sized round for a company whose entire pitch is that the frontier labs shouldn't own your model.

The product today is an API for fine-tuning open-weight models — LoRA and full reinforcement learning runs on models from 35B to 1T parameters, with River managing the GPUs so you don't need an infra team. The numbers it's leading with: complex RL training runs finishing in 15–20 minutes, at 2–4x lower cost than closed-source alternatives.

The longer bet is a personal AI — a model that keeps training on your preferences, goals and working style, and belongs to you rather than to whoever trained the base weights. Babuschkin frames it as moving ownership of AI from the labs to the people using it. Which, notably, is also a bet that open weights stay close enough to closed frontier models to be worth customizing.

Why it matters: if cheap, fast RL on open weights actually works, the moat shifts from who has the biggest model to who has the best data.