River AI Raises $1.1 Billion Just Months After Launch

River AI, the artificial intelligence startup founded by xAI co-founder Igor Babuschkin, has raised $1.1 billion just months after emerging from stealth, giving the young company substantial backing for its effort to rethink how AI models are trained and personalised.

The seed and Series A financing was led by General Catalyst and AMP PBC, with participation from Nvidia, AMD Ventures, Y Combinator and Temasek. River publicly launched in June to develop AI systems that can become personally trained assistants rather than general-purpose agents designed primarily to replace human work.

River AI wants users to train their own models.

Babuschkin previously held AI roles at DeepMind and OpenAI before helping found xAI. At River, he wants to rebuild multiple parts of the AI stack, including training, models, products and eventually hardware designed to keep personal AI systems close to their users.

Thecompany’ss broader vision is for individuals to have AI agents that understand their preferences and work specifically on their behalf.

“Capable agents will be a normal part of everyday life,” Babuschkin wrote when River launched. “They will know you well, and they will be yours, not someone else's.”

River has already launched an API that lets developers customise open models using reinforcement learning and low-rank adaptation, or LoRA, fine-tuning. Pricing is based on the number of tokens processed and varies depending on the open model being used.

River targets an alternative to prompt engineering

The startup positions model training as an alternative to relying heavily on prompt engineering. Instead of repeatedly instructing a model controlled by another provider, River wants developers and businesses to train open models for their own requirements and then deploy them through an API endpoint.

The approach arrives as more enterprises consider using combinations of proprietary and open-weight models rather than relying entirely on a single AI provider. River is targeting the post-training portion of that market with infrastructure intended to simplify reinforcement learning.

River claims enterprises can complete complex reinforcement learning runs in 15 to 20 minutes without maintaining a dedicated infrastructure team. The company also says its approach can offer two to four times the cost savings of closed-source alternatives.

River AI pursues personalised agents

River’s longer-term ambition extends beyond enterprise model customisation. The company expects personally controlled AI agents to become increasingly common and wants to develop technology that allows users to train systems around their own needs.

The idea is emerging alongside greater interest in locally operated AI agents and increasingly powerful personal computing hardware. Nvidia has also been working with PC manufacturers including Dell, Microsoft and HP on computers designed to handle more AI workloads locally.

River has not yet demonstrated how its broader personal AI technology will differentiate itself from competing approaches. With $1.1 billion in new funding, however, the startup has secured substantial resources to develop its training infrastructure and pursue Babuschkin’s vision of AI agents that users can customise and control themselves.