River AI, founded two months ago by xAI co-founder Igor Babuschkin, closed a combined $1.1 billion seed and Series A round led by General Catalyst and AMP PBC on August 11, with strategic participation from Nvidia and AMD Ventures, plus Y Combinator and Temasek. The round values the company at approximately $5 billion and marks one of the largest founding-stage raises in AI history.
The Palo Alto company is building personalized AI infrastructure, starting with River API—a cloud service enabling enterprises to customize open-source LLMs in 15–20 minutes using LoRA fine-tuning, at 2–4× the cost savings of closed-source alternatives. Babuschkin's thesis: most enterprises rely on general-purpose models trained on the entire internet; River enables companies to train, tune, and own models tailored to their own data and workflows. The platform will support models up to 1 trillion parameters.
The longer ambition is 'personal AI'—models that live close to the user (locally or decentralized), learn from your data, and serve your interests rather than a corporation's. River plans to build the full stack: training infrastructure for any developer to fine-tune, products around personalization and continual learning, and new hardware to run personal AI at the edge.
For architects, this signals venture capital's confidence in open-weight models and edge/on-device AI as viable alternatives to hyperscaler-controlled inference. Nvidia and AMD's participation hints that chipmakers see modular fine-tuning infrastructure as a high-volume TAM complementary to their core GPU sales. The speed to $1.1B from zero also shows investor appetite for reproducible personalization as a category within agentic AI.