River AI raises $1.1 billion two months after founding to rebuild AI stack

River AI’s distinguishing claim is not the size of its $1.1 billion raise but the assertion that every component of model training must be rebuilt from the ground up. Igor Babushkin, a co-founder of xAI who previously worked at DeepMind and OpenAI, proposes turning agents into personal assistants that each user can train themselves, rather than following the path most AI labs take that leads to replacing human workers. According to Babushkin, the entire technology stack—training, models, product layer and new hardware—needs end-to-end reconstruction to enable a personal AI to run close to the user.

investors and context

The round, described as seed/Series A, was led by General Catalyst and AMP PBC, with participation from Nvidia, AMD Ventures, Y Combinator and Temasek. AMP PBC is an investment firm founded in 2026 by Angeni Meda, a former general partner at Andreessen Horowitz who backed companies such as Black Forest Labs, Mistral AI, LMArena and OpenRouter. The $1.1 billion amount is unusually large for a two-month-old company and reflects the heightened tension in the AI market. The business backdrop is that organizations are seeking more control over their models by mixing them with open-weight components, and River positions its “neo-cloud” platform as a solution to the post-training expertise gap.

product: API with RL and LoRA

River’s first product is an API priced per million tokens, with rates varying by the chosen open model. The API lets developers perform both reinforcement learning (RL) training and low-rank adaptation (LoRA) fine-tuning on the models. The stated goal is to replace prompt engineering. River’s materials claim that prompting targets a model you do not own and cannot improve, whereas River enables training of open models that you truly own and running them like any other endpoint. The company says any organization can complete a complex RL run in 15 to 20 minutes, without an infrastructure team, and achieve cost savings of two to four times compared with closed-source alternatives.

vision and what remains unclear

The broader vision is that every person will have their own agents, trained by them and acting on their behalf. The source notes that the idea is already materializing in the rise of locally running personal agents such as OpenClaw and its derivatives, and in Nvidia’s collaborations with computer manufacturers like Dell, Microsoft and HP on AI-capable hardware. How River’s technology will differ in practice is still unknown. The company has not released performance metrics, benchmark results or technical details about the models, and the claims about runtime and cost savings are the company’s statements without independent verification.