River AI, a startup founded by xAI co-founder Igor Babuschkin, has raised $1.1 billion in a seed and Series A round led by General Catalyst and AMP PBC. Nvidia, AMD Ventures, Y Combinator, and Temasek also invested.
AMP PBC is an investment firm established in 2026 by Anjney Midha, a former general partner at Andreessen Horowitz. Midha backed companies such as Black Forest Labs, Mistral AI, LMArena, and OpenRouter.
The mission
River emerged from stealth in June with a specific goal. Babuschkin, who worked at DeepMind and OpenAI, wants to rebuild how AI models are trained. The company aims to turn agents into personally trainable assistants instead of tools designed to replace human workers.
Babuschkin wrote that the entire stack must be rebuilt. This covers training, models, the product layer, and new hardware that allows personal AI to run close to the user.
He describes capable agents as a normal part of everyday life. These tools would be quietly present and helpful with what matters to the user. They would know the person well and belong to them, rather than to another company.
What the product does
River already offers an API priced per million tokens. Rates depend on the open model used. Developers can use reinforcement learning and low-rank adaptation fine-tuning on the models.
This first product acts as an alternative to prompt engineering. Prompting steers a model you do not own and cannot improve. River lets you train open models into ones that are truly yours. You can then serve them like any other endpoint.
The company claims any enterprise can complete a complex reinforcement learning run in 15 to 20 minutes. This requires no infrastructure team. The cost is two to four times lower than closed-source alternatives.
Why the timing matters
Enterprises are increasingly interested in controlling their AI model destiny. They want to use a mix of models, including open weight. River promises to solve the post-training expertise part of that problem with its neocloud offering.
The bigger vision is that everyone will have their own agents trained by themselves and working on their behalf. We are already seeing this concept with the rise of personal, locally running agents like OpenClaw and its derivatives.
Nvidia is also partnering with PC makers such as Dell, Microsoft, and HP for AI-capable hardware.
How River’s technology will differ remains to be seen. But the company starts with a large amount of cash to try.
