Recursive Superintelligence, an AI start-up founded by Google and OpenAI engineers, has raised over $500 million in an oversubscribed funding round according to reporting by the Financial Times.

The company hopes to design an AI system that can improve itself without human input, people familiar with its plans told the Financial Times, though this has not yet been proven to work for extended periods of time.

Despite not being officially launched, it has been valued at $4 billion, excluding the new capital. The funding round was led by GV, Alphabet’s venture capital arm, with contributions from Nvidia, according to people familiar with the matter. The round was oversubscribed, meaning the company could ultimately raise as much as $1 billion in total, they added.

The company currently has around 20 staff including co-founders Richard Socher, an AI researcher and former chief scientist at Salesforce, and Tim Rocktäschel, a professor of AI at University College London and former director of Google DeepMind. A person close to the startup told the Financial Times it also employs several former OpenAI researchers including Josh Tobin, Jeff Clune and Tim Shi.

The company is not yet fully operational, but was incorporated on 31 December, according to Companies House filings, where Rocktäschel is listed as a director. Recursive is based in London King’s Cross, close to Google’s London headquarters.

In February, AI-designed chip developer Ricursive Intelligence raised $335 million to fund the development of its own self-improving artificial intelligence. The company was founded by former Google and Anthropic engineers and is based in Palo Alto, California. The two companies have no known affiliation.

---