A London born AI lab that thinks it can race past human intelligence has finally pulled back the curtain. Recursive Superintelligence, founded just months ago, walked out of stealth this week with a jaw dropping $650 million war chest and a $4.65 billion valuation. The tiny team, big names, and a wild promise of self improving AI have already shaken Silicon Valley and London’s tech scene.

Inside the $650 Million Mega Round

The final funding round totals $650 million at a $4.65 billion valuation, led by GV (Google Ventures) and Greycroft, with AMD Ventures and Nvidia also participating. Sources close to the deal say it was heavily chased by investors.

The funding round, which was described as “severely oversubscribed,” was led by GV (Google Ventures) and Greycroft, with participation from NVIDIA and AMD Ventures. That mix of cash and chip muscle is rare for a company this young.

Here is a quick snapshot of where the startup stands today:

Recursive Superintelligence, incorporated in London on December 31, 2025, has raised at least $500 million in a pre-Series A round led by GV with strategic backing from Nvidia. The deal values the company at $4 billion before new capital and the round was so oversubscribed it could eventually reach $1 billion. The new $650 million figure confirms how fast the buzz turned into hard cash.

recursive superintelligence self improving AI startup funding launch

The Star Studded Team Behind the Bet

The startup’s co-founders include Richard Socher, its CEO, who was previously chief scientist at Salesforce, and Tim Rocktäschel, a professor of AI at London’s UCL and a former Google DeepMind scientist. Others who work at the startup, which has a team of less than 30, previously worked at Meta and OpenAI.

Richard Socher has assembled an exceptional group of seven co-founders, bringing together Tim Rocktäschel, Alexey Dosovitskiy, Josh Tobin, Caiming Xiong, Yuandong Tian, Tim Shi, and Jeff Clune. Many of them are household names inside the AI research world.

Clune is a pioneer in evolutionary algorithms and open-ended AI systems; his Darwin Gödel Machine work at Sakana AI demonstrated that AI agents could autonomously rewrite their own code to improve benchmark performance. That résumé fits the company’s mission almost perfectly.

“AI is code, and now AI can code. When these two realities connect, the self improvement loop can be closed.” That short line, shared by GV after the deal, sums up the whole pitch.

The Bold Bet on AI That Improves Itself

Most labs today are chasing bigger models. Recursive is chasing a different idea altogether.

In a blog post on X, it said: “The fastest path to superintelligence will be realised by AI that recursively improves itself, and does so via open-ended algorithms that drive endless innovation. We will first focus on the science of AI itself (by creating AI that improves AI), but the playbook we create will soon allow us to revolutionise every scientific discipline.”

Following this logic, instead of relying on human engineers to hand-design optimizations, Recursive is building systems that conduct experiments to safely improve their own capabilities. The system learns to identify its own limitations, write its own benchmarks, and actively rewrite its own codebase to become more capable.

“We will start with AI research itself but eventually hope to expand its aperture to physics, chemistry and especially pre-clinical biology,” Socher wrote in a post on X. “AI will be to biology what calculus was to physics, a new language and way of thinking that deals with complex systems and helps us understand and engineer them better.”

Key pillars of the company’s plan include:

• Self writing code: Models that edit and patch their own training pipelines.
• Open ended algorithms: Inspired by biological evolution and cultural learning.
• Automated science: First applied to AI, then physics, chemistry and biology.
• Hardware harmony: Tight ties with Nvidia and AMD for raw compute power.

A Crowded Race for Superintelligence

Recursive is not the only big bet in town. Several superstar researchers have launched rival labs in the past year.

AMI Labs, founded by Yann LeCun, is pursuing world models. Ineffable Intelligence, founded by DeepMind’s David Silver, is focused on reinforcement learning. Safe Superintelligence, Ilya Sutskever’s company, is targeting safety first paths to superintelligence.

The London-based company announced a record $1.1 billion seed round in April, co-led by U.S. venture capitalists Sequoia and Lightspeed, with participation from Nvidia, DST Global, Index, Google and the U.K.’s Sovereign AI Fund. That was Ineffable, and it shows how loose the purse strings are right now.

Nvidia CEO Jensen Huang has even picked sides publicly. “The next frontier of AI is superlearners, systems that learn continuously from experience,” said Nvidia CEO Jensen Huang. He added: “We are thrilled to partner with Ineffable Intelligence to codesign the infrastructure for large-scale reinforcement learning as they push the frontier of AI and pioneer a new generation of intelligent systems.”

Yet Recursive stands apart in scope. What separates Recursive is scope. The others are building toward something; Recursive wants to automate the process of building itself. The company plans to first focus on the science of AI, creating AI that improves AI, and then turn that playbook toward every other scientific discipline.

What Comes Next for Recursive

The road ahead is short on talk and heavy on action. Looking ahead, Recursive Superintelligence plans to use the fresh funding to secure large-scale compute infrastructure and run its first “Level 1” autonomous training system. A public launch is targeted for mid-2026, with ambitions extending beyond AI research into broader scientific discovery over time.

So far, the company hasn’t published any concrete technical results. That silence has critics worried. Safety researchers warn that systems which rewrite their own code can drift into behaviour their makers never planned for.

Still, the talent stack is hard to ignore. Recursive’s founding team includes researchers and entrepreneurs who previously helped create AI research organizations at Salesforce and Uber and led teams at OpenAI, DeepMind, Google Brain, and Meta.

Recursive Superintelligence is now the boldest face of a much bigger question. Can machines really teach themselves to be smarter than the humans who built them, and do it safely? The team behind this London startup believes the answer is yes, and they have $650 million and some of the world’s sharpest minds betting on it. For readers watching this race unfold, the next 12 months could be the most defining chapter in modern AI history. What do you think, is self improving AI a leap forward for science or a risk we should slow down? Drop your thoughts in the comments and share your take using #RecursiveSuperintelligence across X and LinkedIn.