On May 13, 2026, Recursive Superintelligence officially emerged from stealth mode, announcing a staggering $650 million funding round that values the company at $4.65 billion. The round was led by GV, Alphabet’s venture capital arm, and Greycroft, with major participation from chipmakers Nvidia and AMD. According to The Next Web and South China Morning Post, the funding round was “heavily oversubscribed,” indicating strong investor confidence in the startup’s ambitious vision.
Despite being only four months old at the time of the announcement, Recursive Superintelligence has already expanded its team to fewer than 30 employees, primarily researchers and engineers. The company operates out of offices in San Francisco and London but has yet to release any commercial product.
Leadership Combines AI Visionaries from Meta, Salesforce, DeepMind, and OpenAI
The startup is led by CEO Richard Socher, former chief scientist at Salesforce and founder of AI search engine You.com. Socher is joined by eight co-founders, including Tian Yuandong, whose involvement is particularly noteworthy from a China-AI perspective.
Tian Yuandong was a research scientist director at Meta’s FAIR (Fundamental AI Research) team, where he led projects on reinforcement learning, large language model (LLM) reasoning, and AI-guided optimization. He spearheaded Meta’s DeepMind-style Go projects, including DarkForest and ELF OpenGo. Tian holds a PhD in robotics from Carnegie Mellon University and graduated from Shanghai Jiao Tong University, one of China’s premier engineering schools.
Other co-founders bring equally impressive credentials: Tim Rocktaschel, professor of AI at University College London and former principal scientist at Google DeepMind; Alexey Dosovitskiy, a co-author of the influential Vision Transformer (ViT) paper; Josh Tobin, formerly of OpenAI; Caiming Xiong; Tim Shi; and Jeff Clune. The company also counts AI luminary Peter Norvig, co-author of the seminal textbook Artificial Intelligence: A Modern Approach, as an adviser.
Recursive Self-Improvement: The Next Frontier in AI Research and Autonomy
Recursive Superintelligence’s core mission is to build AI systems capable of recursive self-improvement, AI that can autonomously discover new knowledge, optimize its own algorithms, and evolve in an open-ended loop without human intervention. The company aims to pioneer a “Level 1” autonomous training system, with a public launch targeted for mid-2026.
This system envisions training AI with the capabilities equivalent to “50,000 doctors” to automate scientific research in AI itself, an audacious step toward accelerating AI innovation exponentially. A significant portion of the new funding will be allocated to securing large-scale computational infrastructure necessary to support such heavy compute demands.
Positioning in the Global AI Race
The launch of Recursive Superintelligence highlights an intensifying race to develop self-improving AI systems, which many experts predict could be transformative for the field. According to The Next Web, competitors like Anthropic have already automated much of their code writing using their AI Claude, while OpenAI’s GPT-5.5 boasts a new parallelization method that accelerates token generation speeds by over 20%. Meanwhile, Google DeepMind has developed AlphaEvolve, an AI agent focused on scientific and algorithmic discovery.
Jack Clark, co-founder of Anthropic, estimates roughly a 60% probability that a system capable of autonomously training a more powerful AI successor will exist by the end of 2028, with about a 30% chance as early as 2027. Recursive Superintelligence’s focus on recursive self-improvement places it squarely in this cutting-edge domain.
Tian Yuandong’s Role Reflects China’s Growing Influence in Global AI Innovation
Tian Yuandong’s involvement underlines a broader narrative of China’s increasing presence in the global AI ecosystem. Although Recursive Superintelligence is a U.S.-headquartered startup, Tian’s background, from a degree at Shanghai Jiao Tong University, a PhD in robotics from Carnegie Mellon University, and then onwards to Meta FAIR, exemplifies the cross-border flow of top AI talent.
This development complements recent EastFrontier coverage arguing that “China’s AI Race No Longer Looks Like Second Place.” Tian’s co-founding of arguably one of the most ambitious self-improving AI startup to date signals that Chinese-born AI researchers are not only contributing but leading on the global stage.
(Related: China’s AI Race No Longer Looks Like Second Place)
Funding Dynamics: Silicon Valley’s Big Players Place Big Bets
GV and Greycroft leading the $650 million round signals a high level of trust in Recursive Superintelligence’s vision and leadership. GV, Alphabet’s investment arm, has a history of backing breakthrough AI startups, and their involvement signals Alphabet’s strategic interest in recursive AI technologies.
Nvidia and AMD’s “major participation,” as described by SCMP, highlights the crucial role of hardware providers in enabling computationally intensive AI research. Both companies have been aggressively pursuing AI-focused chips to meet demand from startups and tech giants alike.
The oversubscribed nature of the round suggests that many investors see recursive self-improvement as a potentially game-changing breakthrough in AI, justifying the high valuation of $4.65 billion despite the absence of a released product.
Recursive Superintelligence’s emergence with heavy backing, world-class research talent, and a clear focus on self-improving AI systems marks a new phase in AI innovation. The startup embodies the convergence of deep AI expertise, cutting-edge research, and massive computational resources.
Tian Yuandong’s leadership role highlights the growing influence of Chinese AI talent in shaping global AI trajectories. As Recursive Superintelligence moves toward its public launch, all eyes will be on whether its vision of recursive self-improvement can become a reality, and how it might redefine the future of artificial intelligence.