- Recursive Superintelligence has signed a multi-year, $410 million compute agreement with AWS.
- This deal accounts for most of Recursive’s recent $650 million funding round. Socher describes it as his smallest deal so far.
- AWS is not taking an equity stake. The two companies will work together to build AI research infrastructure.
Recursive Superintelligence, an AI lab focused on self-improving systems, has signed a multi-year, $410 million agreement with Amazon Web Services to run its automated research platform on AWS.
“Partnering with AWS gives us the infrastructure to run self-improving AI research at a scale very few teams in the world are capable of. We’ve already shown our automated AI research system can outperform years of human optimisation. As our research loops run autonomously, that advantage will compound. We’re excited to partner with AWS and make our system and its infrastructure the fastest at invention generation,” says Richard Socher, chief executive and co-founder of Recursive.
Socher describes this as his smallest deal.
The AWS agreement accounts for most of the $650 million that Recursive raised at a $4.65 billion valuation in May 2026. The round was led by GV and Greycroft, with AMD Ventures and Nvidia also participating.
Socher told TechCrunch this is likely one of the smallest compute deals they will sign in the coming years. He explained that the company is investing more in compute than in hiring. Unlike many AI lab financings, AWS is not taking an equity stake. The companies will work together to build infrastructure for large-scale automated AI research.
Jason Bennett, vice president and global head of startups and venture capital at AWS, welcomed the partnership: “Self-improving AI creates a compounding demand for compute. Every research loop generates the next experiment. In just a few months, Recursive has produced results that outpaced years of community-driven optimisation. AWS gives them the flexibility to run those loops in parallel at massive scale, and the reliability, security, and purpose-built compute to turn months of sequential research into days of parallel discovery.”
Key team members and their significance
Socher, who is also chief executive of You.com and founder of AIX Ventures, leads a team of seven co-founders. The group includes Yuandong Tian, former research scientist director at Meta’s Fundamental AI Research lab; Tim Rocktäschel, professor of AI at University College London and former principal scientist at Google DeepMind; Alexey Dosovitskiy, co-author of the Vision Transformer paper; and Jeff Clune, creator of the Darwin Gödel Machine at Sakana AI.
The company is based in San Francisco and London, with more than 25 researchers and engineers.
Recursive’s system automates the scientific method by proposing experiments, running them, checking results, and deciding what to do next. In its first public results, the system beat a two-year human leaderboard record and outperformed a live human-AI collaboration, achieving top results on NanoChat, NanoGPT Speedrun, and SOL-ExecBench benchmarks. Socher expects the company to release its first usable products by October 2026.
Recursive is not the only company working on self-improving AI. Ilya Sutskever’s Safe Superintelligence focuses on safety, Yann LeCun’s AMI Labs works on world models instead of scaling language models, and David Silver’s Ineffable Intelligence prioritises reinforcement learning. Recursive stands out by putting most of its resources into compute rather than hiring, and by working closely with a single major cloud provider.
For AWS, bringing Recursive on board helps attract top AI labs seeking options beyond Microsoft Azure and Google Cloud. AWS uses custom Trainium chips and Nvidia GPU clusters to compete. The global AI infrastructure market was valued at $142.8 billion in 2026 and is expected to reach about $947.46 billion by 2035, growing at a 23.4% annual rate. Big compute deals like this are fueling much of that growth.
All self-improving AI projects face the same question: does the system really build on its progress, or just spend money faster? Recursive has put more money into this effort than any other area in its four months so far. What happens in October, whether it’s a product launch or more debate, will be important.
