Ex-xAI Co-founder's River AI Raises $1.1B at Two Months Old

Igor Babuschkin's newest venture has been operating in public for barely two months, and it has already secured $1.1 billion in funding. River AI, the startup he founded after leaving xAI, closed the round with General Catalyst and Anjney Midha's AMP PBC leading the check.

The investor list reads like a who's-who of AI infrastructure bettors. Nvidia, AMD Ventures, Y Combinator, and Temasek all joined the round, a signal that chipmakers and sovereign capital alike see River AI's approach to model customization as strategically important rather than merely promising.

Babuschkin's resume explains some of that appetite. Before co-founding xAI, he held research roles at DeepMind and OpenAI, giving him a rare vantage point across three of the field's most influential labs.

What River AI actually sells

The product is a token-based API that lets developers fine-tune open-source models using reinforcement learning combined with low-rank adaptation (LoRA), a technique that updates a small subset of a model's parameters rather than retraining it wholesale. That keeps compute and storage costs down while still allowing meaningful behavioral changes.

River AI claims enterprises can run complex reinforcement learning jobs in just 15 to 20 minutes, without needing a dedicated infrastructure team to manage the underlying clusters. The company positions this as a shortcut past the operational overhead that typically makes custom RL training a specialist's job.

On pricing, River AI says its service delivers two to four times the cost savings compared with closed-source alternatives, though that figure comes from the company's own framing rather than an independent benchmark.

A bigger bet on the AI stack

River AI came out of stealth in June 2026, and its stated ambition goes well beyond an API product. The company says it wants to rebuild the AI stack from training methods through models to hardware, aiming to produce assistants that can be trained around an individual user's needs rather than shipped as one-size-fits-all systems.

River AI's vision centers on personally trainable assistants built on a fully reworked stack, not just a fine-tuning layer bolted onto existing models.

That framing helps explain why hardware players like Nvidia and AMD Ventures joined a round this early. If River AI's thesis holds, cheaper and faster RL-based customization could shift where value accrues in the AI supply chain, away from a handful of frontier labs and toward tooling that lets any enterprise adapt open models on its own terms.