River AI raises $1.1 billion to build personal AI stack

River AI, a startup founded by former xAI co-founder Igor Babuschkin, announced on August 11, 2026, that it has secured $1.1 billion across its Series Seed and Series A funding rounds. The massive capital injection, led by General Catalyst and AMP PBC, also features strategic backing from rival chipmakers Nvidia and AMD Ventures, alongside Y Combinator and Temasek. This funding arrives just two months after the startup emerged from stealth on June 10, 2026. According to Babuschkin, the company wants to "allow people and companies to own their intelligence" rather than renting it from centralized labs.

At the core of River AI's current offering is the River API, a post-training cloud platform that allows developers to fine-tune and run reinforcement learning on open-weight models ranging from 35 billion to 1 trillion parameters. Supported models include Qwen3.6, Kimi K2.6, and GLM 5.2. Instead of billing by the GPU hour, River charges by the token. For example, training costs $1.00 per million tokens on the Qwen3.6 35B model and $12.84 per million tokens on Kimi K2.6 at a 262k context length. Checkpoint storage is priced separately at $0.10 per gigabyte per month.

For enterprise practitioners, this setup eliminates the need for dedicated infrastructure teams. River claims companies can execute complex reinforcement learning runs in 15 to 20 minutes, achieving two to four times the cost savings of closed-source alternatives. A full production-scale reinforcement learning run on a mathematics dataset can cost under $1,000. Crucially, the resulting checkpoints belong entirely to the customer and deploy directly to an OpenAI-compatible endpoint. Looking ahead, River plans to use its capital to develop a personalization layer and proprietary hardware, aiming to transition AI from rented cloud services to locally owned, continuous-learning agents.