Artificial intelligence startup Recursive Superintelligence has signed a multiyear agreement worth $410 million with Amazon Web Services for computing capacity, as the company seeks to scale its work on AI systems designed to improve themselves.
The deal, announced Tuesday, will give the startup access to AWS computing resources as it develops what it describes as open-ended self-improving AI systems. The approach could require substantial computing power as the company works to build systems capable of identifying weaknesses in its performance and improving their capabilities.
Recursive Superintelligence emerged from stealth in May with $650 million in funding. The company was founded by a group of prominent AI researchers and is led by CEO Richard Socher, a former chief scientist at Salesforce. Its backers include GV, the venture capital arm of Alphabet, as well as Nvidia and AMD.
READ: Amazon to lay off employees in Artificial General Intelligence unit (July 22, 2026)
The company’s central research focus is based on the idea that AI systems could eventually contribute to their development by analysing their performance and helping design improved versions of themselves.
That approach could create a potentially accelerating cycle of AI development. However, the extent to which current AI systems can autonomously improve their own underlying capabilities remains an active area of research.
The agreement with AWS underscores the growing importance of computing infrastructure in the race to develop increasingly advanced AI systems. Training and operating large AI models require substantial amounts of specialised computing capacity, making access to chips and cloud infrastructure a critical factor for AI companies.
Recursive Superintelligence’s research strategy places particular emphasis on this relationship between AI development and computing resources. In an earlier interview with TechCrunch, Socher discussed the possibility that compute could become an increasingly important resource in determining which problems AI systems can tackle and how quickly they can advance.
The company has attracted significant funding despite being relatively new. Its emergence from stealth with $650 million in financing highlighted investor interest in research focused on AI systems that could potentially improve with less direct human intervention.
READ: Amazon to raise at least $25 billion for AI infrastructure through bond sale (July 7, 2026)
The AWS agreement also reflects the growing role of cloud providers in the AI industry’s infrastructure race. Amazon has been expanding AWS’s role in providing computing resources for AI developers, while companies across the sector are seeking access to increasingly powerful hardware to train and run their models.
For Recursive Superintelligence, the multiyear deal provides greater flexibility as it works to scale its research without having to build and operate all of its computing infrastructure.
The agreement comes as major technology companies and emerging AI startups increasingly compete for access to specialised processors and data centre capacity. The surge in AI development has made computing availability one of the industry’s key constraints, alongside funding, talent and energy.
Recursive Superintelligence has not yet disclosed a consumer product based on its research, but the company has indicated that it expects to develop products as its technology matures.
The $410 million AWS agreement represents a significant commitment to the company’s research strategy and highlights the enormous computing requirements associated with efforts to develop AI systems capable of improving themselves.
As the AI industry moves toward increasingly autonomous systems, Recursive Superintelligence’s approach will be closely watched for evidence of whether recursive self-improvement can deliver meaningful gains in AI capabilities and whether the additional computing resources required can translate into practical advances.