London-based startup Recursive Superintelligence is betting heavily on a new frontier in artificial intelligence. Co-founder Tim Rocktaschel says the company will deliver recursively self-improving AI within two years. Unlike current models that rely on human engineers for upgrades, Recursive’s systems are designed to autonomously enhance their own capabilities, drawing on principles of evolution and automated scientific discovery.
The venture has already secured $650 million in funding at a $4.65 billion valuation, with backing from GV, Nvidia, and AMD. That round could expand to one billion dollars. Despite having fewer than thirty employees and no public product, the startup’s valuation implies roughly $155 million per worker.
The involvement of major chipmakers signals strong expectations for surging compute demand as these systems iterate. Over the next two years, success will not be measured by a consumer launch, but by whether Recursive can prove measurable, autonomous improvement in a controlled environment. With Rocktaschel’s Google DeepMind pedigree backing the timeline, the market is now waiting to see if the technology can match the valuation.