River AI Raises $1.1B Two Months After Leaving Stealth

River AI has raised $1.1 billion in its first round, roughly two months after coming out of stealth. General Catalyst and AMP PBC led, with Nvidia, AMD Ventures, Y Combinator and Temasek participating. The company did not disclose a valuation. TechCrunch reported the round on Tuesday, the same day it hit the wire.

The founder is Igor Babuschkin, who worked on generative modelling and reinforcement learning at Google DeepMind, ran large-scale training efforts at OpenAI, and then co-founded xAI — the lab that has since been absorbed into SpaceX. Neither River nor its investors have said why he left, and the round documents do not address it.

What River sells is narrower than the funding total suggests. It is an API for fine-tuning open-weight models, using reinforcement learning and low-rank adaptation, priced on token usage. The pitch is that a company can "train open models into ones that are truly yours" without standing up an infrastructure team: River claims a complex reinforcement learning run finishes in 15 to 20 minutes, at two to four times lower cost than the closed-model alternatives. Babuschkin frames the endpoint as everyone having their own continually improving agents rather than a single vendor's model standing in for a worker.

That is a direct bet against the shape the market has settled into. The dominant enterprise pattern in 2026 is renting frontier capability through an API and steering it with prompts, context and tooling — the weights stay with the lab. River's argument is that the open-weight tier has closed enough of the quality gap that owning a tuned copy beats renting a better one, provided the tuning is cheap and fast enough to run repeatedly. The 15-to-20-minute figure is the whole thesis compressed into a number: reinforcement learning stops being a research project and becomes something you do on a Tuesday afternoon.

The investor list is worth reading twice. Nvidia and AMD Ventures both being in suggests neither wants to miss a company whose business is selling short, bursty training runs to thousands of customers rather than long pretraining jobs to a handful of labs — a different demand curve for accelerators than the one the current buildout is designed around.

Whether $1.1 billion for a nine-week-old company with no disclosed customers is a bet on the thesis or on the founder is not really in question. It is the founder. That has been the pattern all year: Crunchbase counted 16 billion-dollar rounds in the second quarter alone, together taking 53% of all venture dollars in the period, and the ones going to brand-new entities have consistently followed people out of the frontier labs. River now has to convert a résumé into a product line before the open-weight window it is betting on either widens or closes.