Recursive raises $650M at $4.65B valuation

Recursive Superintelligence emerged from stealth on May 13 with more than $650 million in Series A funding at a $4.65 billion valuation, according to the company and law firm Wilson Sonsini, which advised on the deal. GV said it co-led the round with Greycroft, while Wilson Sonsini and multiple reports said AMD Ventures and Nvidia also participated. Richard Socher, the former Salesforce chief scientist who is now Recursive’s CEO and co-founder, is leading a startup that says it wants to build AI systems that improve themselves. The company surfaced publicly after earlier reports in April said it had raised at least $500 million at a $4 billion pre-money valuation, with the final disclosed round landing above that level. (wsgr.com) GV’s write-up and outside reports describe Recursive as a London-founded lab with offices in London and San Francisco and a team that includes former researchers and executives from OpenAI, Google DeepMind, Meta, Salesforce AI and Uber AI. The company has not publicly tied the launch to a commercial product or revenue line. ### Who is behind Recursive? (gv.com) GV named Socher’s seven co-founders as Tim Rocktäschel, Alexey Dosovitskiy, Josh Tobin, Caiming Xiong, Yuandong Tian, Tim Shi and Jeff Clune. GV said the group has worked on projects including the Vision Transformer and safety work such as “rainbow teaming.” The New York Times described Recursive as founded by former Google, Meta and OpenAI researchers. (tech.eu) Tech.eu reported that Rocktäschel is a University College London professor and former Google DeepMind scientist, while Socher previously served as Salesforce’s chief scientist. ### What exactly is the company saying it wants to build? (gv.com) Recursive said its core bet is “recursive self-improvement” — AI systems that can identify weaknesses, run experiments and improve their own code and capabilities. GV said the startup is building an “open-ended architecture” meant to teach AI to improve its own codebase rather than rely only on human engineers to hand-design upgrades. (nytimes.com) Tech.eu reported that the company said it would start with “AI that improves AI” and then try to apply the same playbook to other scientific disciplines. That framing also appeared in secondary coverage citing the company’s public materials. ### How does the final round compare with the earlier reports? April 20 reporting citing the Financial Times said Recursive had raised at least $500 million at a $4 billion pre-money valuation and that the round could grow because investor demand was strong. (gv.com) Wilson Sonsini’s May 13 note said the company ultimately announced more than $650 million at a $4.65 billion valuation. (tech.eu) That sequence suggests the company either expanded or finalized the financing between the earlier report and the public launch. GV separately characterized the financing as an “early $650M funding” at the same $4.65 billion valuation. ### Why are Nvidia and AMD Ventures part of the syndicate? Nvidia and AMD Ventures were listed as participants in the financing by Wilson Sonsini, GV coverage and multiple follow-on reports. (techfundingnews.com) Neither source disclosed the size of their individual checks in the material reviewed. (gv.com) Chip-company participation is notable because Recursive is pitching a research-heavy approach that would likely require significant computing infrastructure, though the company did not disclose a hardware procurement plan in the materials reviewed. That is an inference based on the startup’s stated research agenda and the presence of strategic chip investors. (wsgr.com) ### What should readers watch next? May 13 is the key public launch date, but Recursive has not yet published a product launch timetable in the sources reviewed. Earlier April reporting said the company was targeting a public debut around mid-May 2026, which has now occurred. Future disclosures are likely to come from Recursive’s own public channels and from named backers such as GV, which has already published its investment thesis and the list of co-founders. (gv.com) The next concrete milestone to watch is whether the company releases technical results, research papers or infrastructure details to support its self-improving AI claims. (techfundingnews.com)