The capital injection includes strategic backing from chip giants Nvidia and AMD Ventures, alongside contributions from Y Combinator and Temasek. River AI intends to challenge the dominance of large-lab general-purpose models by providing an API capable of executing complex reinforcement-learning training runs in under 20 minutes. According to the company, this infrastructure-free approach offers a cost reduction of two to four times compared to existing closed-source alternatives.

Babuschkin argues that the current trajectory of artificial intelligence favors the labs that train the models rather than the users deploying them. By focusing on open-weight models, River AI hopes to make high-performance computing affordable and accessible. The company did not disclose its post-money valuation. Babuschkin previously held key roles at Google DeepMind and OpenAI, where he focused on generative modeling and large-scale training architectures.