Artificial intelligence company Recursive has announced a significant multi-year, $410 million agreement with Amazon Web Services to serve as its primary cloud provider. This strategic partnership is set to fuel Recursive's development of a self-improving superintelligence designed to automate knowledge discovery. The collaboration will provide the immense computational power required to scale the company's unique automated AI research system on a global level.
A Strategic Partnership for AI Advancement
The substantial financial commitment underscores the scale of Recursive's computational needs, which are driven by an AI that constantly generates new experiments. Richard Socher, CEO and co-founder of Recursive, stated that partnering with AWS provides the infrastructure to run this research at a capacity very few teams in the world can access. He further explained that as the research loops run autonomously, the system's performance advantage will compound over time.
Jason Bennett, a vice president at AWS, highlighted the unique demand for compute created by self-improving AI, where each research cycle initiates the next experiment. He noted that AWS offers the elasticity to run these experimental loops in parallel at a massive scale, dramatically shortening research timelines. This infrastructure turns what would be months of sequential research into just days of parallel discovery, accelerating the pace of innovation.
The Vision for Self-Improving Superintelligence
Recursive's core mission is to build AI that automates the scientific method by proposing ideas, running experiments, and using the results to inform subsequent tests. The company emerged from stealth in May with a formidable $4.65 billion valuation, backed by prominent investors including GV, Greycroft, AMD Ventures, and NVIDIA. Its founding team comprises distinguished former research leaders from industry giants like OpenAI, Google DeepMind, and Meta AI, ensuring a deep well of expertise.
The team's collective experience spans critical AI domains such as natural language processing, robotics, reinforcement learning, and systems research. They have contributed to major advances in areas like AI-generating algorithms, self-improving coding agents, and automated red-teaming. This diverse background is central to Recursive's ambitious goal of creating a truly open-ended process of automated scientific discovery.
Demonstrating Early Success and Future Collaboration
In its first public results, Recursive's automated system has already achieved state-of-the-art performance, breaking a two-year human leaderboard record. The system outperformed a live collaborative human-AI effort across three distinct benchmarks, showcasing its powerful optimization capabilities. These early successes provide tangible proof of the potential held by the company's self-improving AI framework.
The collaboration extends beyond a simple cloud service agreement, as Recursive and AWS plan to co-develop infrastructure purpose-built for this new paradigm of research. This joint effort aims to create a highly optimized environment for large-scale automated AI experimentation. The goal is to build the world's fastest and most efficient platform for invention generation, leveraging both Recursive's algorithms and AWS's hardware.
The Broader Implications for Scientific Discovery
According to Socher, the clearest path to superintelligence is through AI that recursively improves itself, a concept that forms the foundation of his company. He posits that by enabling AI to write and refine its own code in a continuous loop, we can automate the very process that has driven human progress for centuries. This represents a fundamental shift from using AI as a tool to establishing AI as an autonomous engine of discovery.
The $410 million partnership between Recursive and AWS is more than a cloud computing deal; it is a major endorsement of a new frontier in artificial intelligence. By providing the scalable infrastructure for a self-improving system, this collaboration is set to accelerate the automation of scientific research itself. The venture could ultimately unlock unprecedented advancements across countless fields by fundamentally changing how new knowledge is discovered and applied.