Recursive Superintelligence Raises $650M to Build Self-Improving AI

Recursive Superintelligence Inc., a startup focused on creating self-improving artificial intelligence, launched today with $650 million in funding.

Led by Alphabet's GV fund and Greycroft, with participation from Nvidia and AMD, the round values the company at $4.65 billion.

Founded earlier this year by Richard Socher, former chief scientist at Salesforce and founder of You.com, Recursive now employs over 25 people in San Francisco and London. The goal: build a recursive self-improving superintelligence-an AI capable of discovering new knowledge like a human scientist.

Its first priority is an AI that can improve its own code through automated simulations. The system will develop experiment ideas, test them, validate results, and refine its training and inference infrastructure, all within guardrails to prevent risky output.

OpenAI already uses GPT-5.5 for similar tasks, boosting token speeds over 20% by optimizing request parallelization. Alphabet designs its TPU chips using neural networks trained on blueprints.

Socher says the technology will eventually expand into physics, chemistry, and pre-clinical biology, calling AI a new language for understanding complex systems.