IMPLICATOR
Recursive Superintelligence, the four-month-old AI startup founded by former Salesforce chief scientist Richard Socher, has raised at least $500 million in a round led by Google's venture arm GV with backing from chipmaker Nvidia, the Financial Times reported on Thursday. The financing values the London-incorporated company at $4 billion before new capital and was so oversubscribed it could eventually reach $1 billion, people familiar with the deal told the FT. The round formalizes figures Socher told Implicator.ai last week were "directionally right," and it places Recursive among the most richly capitalized of a new wave of labs spinning out of OpenAI, Google, and Meta.
Key Takeaways
- Recursive Superintelligence has raised at least $500 million at a $4 billion valuation, with Google's GV leading and Nvidia contributing, according to the Financial Times.
- The four-month-old London-incorporated startup names co-founders from Google DeepMind and OpenAI, including Berlin-born UCL professor Tim Rocktäschel and former OpenAI researchers Josh Tobin, Jeff Clune, and Tim Shi.
- Richard Socher told Implicator.ai the company has already hit undisclosed technical milestones and aims to automate the entire AI development pipeline, from evaluation through research direction.
- Socher dismissed the EU regulatory environment as turning Europe into 'the most beautiful open-air museum,' a pointed framing given Recursive's London legal base sits outside the EU AI Act.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
A team built from DeepMind and OpenAI
The FT named Recursive's core founders for the first time. Alongside Socher: Tim Rocktäschel, the second German on the founding team. Berlin-born, Humboldt-educated. Rocktäschel holds an AI professorship at University College London and until recently ran Google DeepMind's Open-Endedness group as director and principal scientist. His team's Genie interactive world model won the ICML 2024 Best Paper Award. Former OpenAI researchers Josh Tobin, Jeff Clune, and Tim Shi round out the publicly identified lineup, with others from Google and Meta also on staff. About 20 people work there. "It's a ridiculously strong team," one person close to the start-up told the paper.
Socher told Implicator.ai last week that Recursive had eight co-founders and had "grown considerably" since that initial group. He declined to name them then, saying the full lineup would come at public launch. That lineup just came early, via the FT.
Option C, all of the above
Asked which slice of frontier AI development Recursive wants to automate, whether evaluation, data selection, training, post-training, or research direction itself, Socher answered: "Option C, all of the above."
That scope goes further than earlier reporting suggested. Internal milestones exist, Socher said, though naming them is off-limits. "We already have them, but I unfortunately can't share them yet." He frames self-improving AI not as science fiction but as an engineering target for the next few years.
Unproven over extended periods, the concept. That was the FT's careful framing, citing people familiar with Recursive's plans. Research stage, not product stage.
The bottleneck isn't compute
Most AI scaling arguments start with chips, clusters, and power bills. Socher disagrees. "The biggest bottleneck is in people's heads," he told Implicator.ai. "In the ideas and the speed at which you have to manually implement and validate them."
That framing positions Recursive not as another lab burning capital on bigger clusters but as a company trying to remove the human researcher from the loop entirely. If AI can generate hypotheses, run experiments, and evaluate results faster than a $20-million-a-year researcher, the economics of frontier development shift. Socher traced the arc in three stages. First, neural networks learned features on their own, eliminating feature engineers. Then unified models killed task-specific architectures. Now: AI that trains itself. "The third and perhaps final stage of neural networks," as he put it.
He mentioned losing a hire recently because the candidate bet AI researchers themselves would be automated within a few years.
A crowded field of new labs
Recursive joins a growing cohort of labs spinning out of frontier shops. Thinking Machines Lab, Safe Superintelligence, Ineffable Intelligence, and Advanced Machine Intelligence Labs have all landed at eye-watering valuations in recent months. The bet among their backers is simple enough. These new shops might out-flank OpenAI and Anthropic, which are largely pinned to scaling ChatGPT and Claude.
Q1 2026 alone pulled $300 billion into startups, per Crunchbase, most of it chasing the OpenAI, Anthropic, xAI, and Waymo mega-deals. No prior quarter has come close.
London base, skeptical of Europe
Recursive's legal home is London. Companies House records the incorporation at the end of last year, and Rocktäschel works from there. Two Germans building a frontier AI lab abroad, one in San Francisco, one in London. Neither picked Berlin. Socher is sharp about why. The EU AI Act, he told Implicator.ai, "has slowed the whole region down even more than it already was," stacking bureaucracy on bureaucracy. "It's really sad how often Europe shoots itself in its own feet," he said, nearly actively declining to help shape the future. The line Socher keeps hearing from founders: "Europe is just the most beautiful open-air museum."
A pointed remark, given that his own company's legal home sits inside that museum. London sits outside the EU AI Act, of course. That is part of the point.
Three roles, one founder
Socher still runs You.com and his venture fund AIX Ventures alongside Recursive. He addressed potential conflict-of-interest concerns by noting that only AIX handles deal flow, and the two operating companies serve different layers of the stack. You.com builds search infrastructure for language models. Recursive builds the models themselves. "Both can be customers of each other," he said.
You.com now generates revenue in the "high double-digit millions," Socher disclosed to Implicator.ai, without specifying whether that figure is annual or run-rate. Thousands of customers, some with billions of end users.
Whether the separation between the three roles holds as Recursive scales is an open question. GV declined to comment to the FT. Nvidia and Socher did not respond to the paper's requests. The public launch is expected in May, Socher told Implicator.ai, and that is when the technical claims about self-improving AI meet the first real test.
Frequently Asked Questions
Who funded Recursive Superintelligence's round and how much did it raise?
Google's venture arm GV led the round with backing from Nvidia, securing at least $500 million at a $4 billion pre-money valuation, the Financial Times reported. The round was oversubscribed and could eventually reach $1 billion. GV declined to comment and Nvidia did not respond to the FT.
Who are Recursive Superintelligence's co-founders?
The FT named Richard Socher, Berlin-born UCL professor Tim Rocktäschel (until recently director and principal scientist at Google DeepMind, where his team won the ICML 2024 Best Paper Award for the Genie interactive world model), and former OpenAI researchers Josh Tobin, Jeff Clune, and Tim Shi. Other researchers come from Google and Meta. Socher said the startup has about 20 staff.
What is Recursive Superintelligence building?
The company aims to build AI systems that continuously improve themselves without human intervention. Socher told Implicator.ai the target is to automate the entire frontier AI development pipeline: evaluation, data selection, training, post-training, and research direction itself. He calls it the 'third and perhaps final stage of neural networks.'
When will Recursive Superintelligence launch publicly?
Socher told Implicator.ai the public launch will come roughly a month after the April 9 interview, placing it around mid-May 2026. The FT reporting on April 17 formalized the funding round ahead of the planned announcement, but the company has not yet officially announced its existence.
Why did Recursive incorporate in London rather than the US or EU?
UK Companies House filings show Recursive was incorporated in London at the end of 2025. The UK sits outside the EU AI Act, which Socher told Implicator.ai has 'slowed the whole region down.' London offers a European talent base, notably Tim Rocktäschel and Google DeepMind alumni, without the EU regulatory overhead.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: [email protected]
