On Tuesday, Recursive Superintelligence revealed a $400 million compute agreement with Amazon Web Services. The firm, which left stealth in May after raising $650 million, focuses on open-ended self-improving AI systems that require heavy computational resources. This multi-year pact is designed to give flexibility while the company scales those systems.

Recursive's $410 million spend accounts for the majority of its total funds raised to date. Yet during a discussion with TechCrunch, Socher noted he expects this to be the first of several similar deals. The current announcement is "likely going to be one of the smallest compute deals we are going to sign in the next few years," Socher said.

Recursive's focus on self-improving AI means most of the budget that would normally cover headcount and operations goes directly to compute, as the firm works to automate its own product development.

"For us, it is less about headcount and more about agent count," Socher said.

There is no investment element in Amazon's role, unlike the hybrid funding models common at other major labs. Still, the size of the commitment lets AWS dedicate significant resources to meet Recursive's specific needs, which could help attract other foundation-model AI companies in the future.

"Part of the agreement is that we are going to co-develop infrastructure purpose-built for these types of company," said Jason Bennett, VP for startups and venture capital at AWS.

Recursive self-improvement has long been seen as a pivotal moment for AI, with some forecasting rapid progress once systems improve without human input. As more groups pursue the idea, requirements remain unclear, and views differ on whether a breakthrough is near or whether self-improvement is a gradual process.

In Recursive's case, the goal is to apply RSI to build real products, and Socher expects to release the first examples before year end.

"We are excited to build like really amazing products that people can use, and you will see those within a few months, not within a few quarters or years," Socher says. "In October or so, you will see some actually tangible, useful things that you will be able to play around with."