River AI is a full-stack AI company founded in 2026 by Igor Babuschkin, an xAI co-founder who previously worked at OpenAI and Google DeepMind, and a team of former xAI colleagues. It came out of stealth on 2026-06-10 with the thesis that the AI stack should be rebuilt around personal agents that users own and train, rather than intelligence rented from a handful of closed labs. Its first product is the River API, which offers LoRA fine-tuning and reinforcement learning on open-weight models reported to range from 35B to 1T parameters (including Qwen3.6, Kimi K2.6 and GLM 5.2), metered per million tokens for both training and inference instead of by GPU hour; the company claims a complex RL run can complete in 15-20 minutes with no infrastructure team and at 2-4x cost savings versus closed-source alternatives. River has also said it intends to build hardware that keeps personal AI on users' own machines so private data never reaches River or any other provider, and to release open-source technology, though that work was described as months away at the time of the funding announcement. On 2026-08-11, roughly two months after launch, it announced $1.1B raised across Series Seed and Series A co-led by General Catalyst and AMP PBC, with Nvidia, AMD Ventures, Y Combinator and Temasek participating, at a reported ~$5B valuation — among the largest early-stage rounds of the year. Headcount was reported at about 20.

Company Profile

Founders & Key People

Investors

Products

• River API (per-token billed training and inference on open-weight models)
• LoRA fine-tuning
• Reinforcement-learning training runs
• Personal AI hardware (announced, not yet shipping)

Competitors

AIAI InfrastructureOpen WeightsFine-TuningPersonal AI

Private-company numbers are not real-time. Reflects publicly disclosed valuations from press releases, news reports, and tender offers as of 2026-08. Refreshed quarterly.