River AI has secured $1.1 billion to build a personalized AI stack, aiming to make customized AI models faster, cheaper, and more accessible to developers and companies. The funding round was led by General Catalyst and AMP PBC, with participation from NVIDIA and AMD Ventures, Y Combinator, and Temasek. Founded by former DeepMind, OpenAI, and xAI researcher Igor Babuschkin, the startup is building infrastructure that lets users customize and continually train models for their own needs. The company said general-purpose models are rarely tailored to specific organizational data and workflows, while custom models previously required specialized hardware, an infrastructure team, and months of development. River’s API aims to reduce that complexity, with enterprises able to complete complex reinforcement-learning training runs in as little as 15 to 20 minutes without maintaining an infrastructure team, achieving two to four times the cost savings of closed-source alternatives. The platform supports LoRA fine-tuning and reinforcement learning for frontier open-weight models, managing underlying infrastructure automatically, including fast movement of model weights, consistency between sampling and training, and elastic compute allocation. Models can be deployed immediately after training, with token-based billing charging for actual training and inference usage rather than unused GPU capacity. Babuschkin said River’s goal is to shift AI ownership from training companies and laboratories toward users. The company is also developing hardware and consumer-facing products, with a long-term strategy to build a vertically integrated stack where personal AI operates close to users, continually learns from them