ChainThink report: According to funding sources, Recursive Superintelligence has completed a financing round of at least $500 million, with a pre-money valuation of $4 billion. GV (Google Ventures, formerly Google Ventures) led the round, with Nvidia participating as a follow-on investor.

The round was oversubscribed, with the final size potentially reaching $1 billion. The company was incorporated in London late last year, currently employs around 20 people, and has not yet made a public debut.

The founding team primarily comes from core research roles at OpenAI, Google DeepMind, and Salesforce, including former Salesforce Chief Scientist Richard Socher, former DeepMind Lead Researcher Tim Rocktäschel, and multiple members from OpenAI, Google, and Meta.

The company's mission is to develop AI systems capable of autonomous, continuous self-improvement without human intervention. Unlike current mainstream large models, whose parameters are fixed post-training and require engineers to re-collect and re-train data for upgrades, self-improving AI can autonomously generate training data and update model parameters, thereby breaking free from human-driven training loops.

This approach represents a long-term objective in AI research, though no publicly verifiable results currently demonstrate sustained, stable operation over extended periods.

According to Crunchbase data, global AI startup funding reached a record high of $300 billion in the first quarter of this year, with OpenAI, Anthropic, xAI, and Waymo capturing the lion’s share.

Recursive Superintelligence is one of several new AI labs recently spun out from major tech firms, alongside peers such as Thinking Machines Lab, Safe Superintelligence, Ineffable Intelligence, and Advanced Machine Intelligence Labs.