A huge seed-stage bet on personal agents

River AI, founded by xAI co-founder Igor Babuschkin, has raised $1.1 billion in a seed/Series A round led by General Catalyst and AMP PBC. Nvidia, AMD Ventures, Y Combinator, and Temasek also participated. For a company that only emerged from stealth in June, the size of the round is striking and signals continuing investor appetite for foundational AI infrastructure.

River’s thesis is not simply to build agents that replace human workers. Babuschkin, whose background includes AI roles at DeepMind and OpenAI, argues that the stack should be rebuilt end to end — including training, models, the product layer, and hardware — so that agents can become personally trainable assistants that remain aligned with their users.

Investors and early product direction

AMP PBC is an AI-focused investment firm founded in 2026 by former Andreessen Horowitz general partner Anjney Midha, who previously backed companies such as Black Forest Labs, Mistral AI, LMArena, and OpenRouter while at a16z. The participation of Nvidia and AMD Ventures also places River near the broader AI compute and hardware ecosystem, although no specific strategic partnership details were disclosed.

Key facts:

• Funding size: $1.1 billion;
• Round: seed/Series A;
• Lead investors: General Catalyst and AMP PBC;
• Participants: Nvidia, AMD Ventures, Y Combinator, and Temasek;
• Founder: Igor Babuschkin, formerly associated with DeepMind, OpenAI, and xAI.

River already offers an API priced per 1 million tokens, with rates depending on the open model used. A token is the basic unit of text processed by a model. The API supports reinforcement learning, a method that improves model behavior through feedback, and LoRA fine-tuning, a lightweight technique for adapting models to specific needs.

From prompt engineering to post-training

River frames its first product as an alternative to prompt engineering. Prompt engineering tries to steer a model through carefully written instructions, but users generally do not own or improve the underlying model. River’s pitch is that developers can train open models into versions that are more genuinely theirs, then serve them like regular endpoints.

That places the company in the growing market for post-training tools. As enterprises adopt a mix of AI systems, including open-weight models, they increasingly want control over model selection, adaptation, and deployment. River says its neocloud offering can help any enterprise complete a complex reinforcement learning run in 15 to 20 minutes without an infrastructure team, while delivering two to four times the cost savings versus closed-source alternatives.

Neocloud, in this context, refers to cloud infrastructure designed specifically around AI training, inference, and optimization workflows rather than general-purpose computing alone.

The bigger vision and the test ahead

River’s long-term vision is that every person will have agents trained by themselves and working on their behalf. The article points to early signs of this direction in locally running personal agents such as OpenClaw and its derivatives, as well as Nvidia’s partnerships with PC makers including Dell, Microsoft, and HP on AI-capable hardware.

The ambition is compelling: agents that are not merely task-based assistants, but persistent software companions that understand a user and act in that user’s interest. Still, the technical differentiation remains to be proven. River will need to show that its training tools are simple, reliable, cost-effective, and meaningfully better than existing approaches.

The round gives River an unusually large war chest for a very young company. It also raises expectations. In the near term, the clearest opportunity may be enterprise and developer post-training workflows for open models. Over the longer term, if local agents, AI PCs, and open model ecosystems continue to mature, River’s personal-agent vision could become a major application layer. For now, the market will be watching whether the company can turn a bold thesis into repeatable products and real customer adoption.