River AI, an artificial intelligence startup founded by xAI co-founder Igor Babuschkin, has raised $1.1 billion in funding as it looks to expand tools that allow businesses and developers to customise AI models using their own data.
The funding round was led by General Catalyst and AMP PBC, with strategic investments from NVIDIA and AMD Ventures. Y Combinator and Singapore-based investment company Temasek also participated.
The financing gives the young company significant backing as it enters a competitive market for enterprise AI infrastructure. River AI is positioning itself around an open AI stack that enables customers to train and customise models rather than relying entirely on general-purpose systems developed by major AI laboratories.
The company expects enterprise AI adoption to increasingly shift towards businesses owning and adapting models for their specific requirements, particularly as open-weight models give developers greater control over how AI systems are trained and deployed.
River AI provides an application programming interface, or API, through which developers can customise models using reinforcement learning. According to the company, customers can complete complex reinforcement-learning training runs in approximately 15 to 20 minutes without requiring a dedicated infrastructure team.
River AI also claims its approach can be two to four times more cost-effective than closed-source alternatives. These performance and cost claims have been made by the company and have not been independently verified.
The startup's proposition addresses a growing question in enterprise AI: whether companies should depend primarily on large general-purpose models or build more specialised systems around their own data, workflows and requirements.
Open-weight models have expanded the options available to businesses by allowing organisations to modify and deploy AI systems with greater control over their underlying models. For enterprises, customisation can also provide more flexibility around performance, costs and the way proprietary information is incorporated into AI applications.
Babuschkin said River AI's approach is based on making AI open, accessible and affordable, with models designed to work more directly for the people and organisations using them rather than remaining controlled by the laboratories that originally trained them.
His background spans several of the companies that have shaped the current generative AI market.
Before founding River AI, Babuschkin co-founded Elon Musk's xAI. He previously worked on generative modelling and reinforcement learning at Google DeepMind and led large-scale model training at OpenAI.
The funding also brings together investors from both the venture capital and semiconductor industries. NVIDIA and AMD are competing aggressively in the market for processors and infrastructure used to train and run AI models, while General Catalyst has backed companies across enterprise software and artificial intelligence.
River AI's focus on reinforcement learning comes as AI developers increasingly look beyond initial model training towards techniques that can improve how models perform on specialised tasks.
The company's emergence also reflects growing investor interest in the infrastructure surrounding AI models rather than only the developers of foundation models themselves.
River AI will use the new capital to expand its AI model tools and support its broader effort to build an open AI technology stack.
As businesses seek greater control over how AI systems interact with proprietary data and workflows, River AI is betting that the next stage of enterprise adoption will involve more companies customising and operating models around their own requirements instead of relying exclusively on standardised AI services.