River AI has raised $1.1 billion in seed/Series A funding, just around two months after its founding. The round was led by General Catalyst and AMP PBC, with participation from Nvidia, AMD Ventures, Y Combinator, and Temasek.
The company was founded by Igor Babushkin, xAI’s co-founder, who previously held artificial intelligence roles at DeepMind and OpenAI. River emerged from stealth in June, announcing a vision focused on rebuilding AI from the ground up, starting with how models are trained.
Personally Trainable Agents
Babushkin wants to turn AI agents into assistants that users can train themselves, rather than directing the technology toward replacing human workers, according to his company launch blog post.
River believes achieving this vision requires rebuilding the entire stack, including training, models, and the product layer, as well as new hardware that allows personal AI to remain close to the user. Babushkin describes future agents as quietly present in everyday life, knowing the user well and acting in their best interest.
An API for Customizing Models
The company currently offers an API priced per million tokens, with prices varying according to the open model used. The API allows developers to use reinforcement learning (RL) and low-rank adaptation (LoRA) to fine-tune models.
River presents its first product as an alternative to relying solely on prompt engineering. According to the company’s description, writing prompts directs a model that the user does not own and cannot improve, while River allows open models to be trained to become customized for the user and then run like any other endpoint.
A Bet on Enterprise Needs
The large round comes at a time when enterprises are showing growing interest in controlling the future of their models by using a mix of models, including open-weight models. River is betting on providing post-training expertise as part of its cloud offering, which it calls neocloud.
The company says any enterprise can carry out a complex reinforcement learning process within 15 to 20 minutes, without the need for an infrastructure team, and with savings of between twofold and fourfold compared with closed-source alternatives. These figures appear in the company’s funding announcement.
A Broad Vision and Massive Funding
River’s broader vision is for every person to have their own agents, train them personally, and have them act on their behalf. The market context points to the emergence of personal agents that run locally, such as OpenClaw and its forks, alongside Nvidia’s efforts to partner with computer manufacturers such as Dell, Microsoft, and HP to provide hardware capable of running AI applications.
It remains unclear how River’s technology will differ in practice from other solutions, but the company is beginning its journey with massive funding that gives it substantial resources to develop its vision.