Recursive Superintelligence has signed a $400 million multi-year deal with Amazon Web Services to buy computing power, the company announced on July 28, 2026. Richard Socher, the company's founder, told TechCrunch the figure is $410 million. He spoke alongside Jason Bennett, AWS's vice president for startups and venture capital. TechCrunch

What makes this deal unusual is what it does not include: an investment. In recent years, most big AI companies and their cloud providers have made deals that combine computing resources with ownership stakes — the cloud provider gets a piece of the company along with the contract. Amazon's involvement here is purely a business transaction. As part of the agreement, AWS and Recursive will work together to build computer systems specifically designed for self-improving AI, meaning AI that can get better at its tasks over time without humans stepping in at every step. TechCrunch

Recursive Superintelligence launched publicly in May 2026 with $650 million in funding led by GV (Google Ventures) and Greycroft, a venture capital firm. The company was valued at $4.65 billion and was only four months old at the time. Founded by former researchers from Meta and DeepMind, Recursive is focused on building AI systems that improve themselves and wants to use those systems to automate its own product development. The Next Web Crowdfund Insider

Socher said this computing agreement is a starting point, not a limit. He expects it to be the first of many and called it "likely going to be one of the smallest compute deals we're going to sign in the next few years." TechCrunch

The approach fits Recursive's overall philosophy. Socher said the company's focus is "less about headcount and more about agent count," meaning the company plans to grow by adding autonomous AI agents — software programs that can perform tasks and make decisions on their own — rather than by hiring more human engineers. Socher said Recursive expects to release products people can actually use around October 2026. TechCrunch

The TechCrunch report is original journalism based on direct interviews with Socher and Bennett, not a press release.

The broader context here is a pattern taking shape across well-funded AI companies. The biggest AI labs — including OpenAI, which works with Microsoft, and Anthropic, which works with Google Cloud and AWS — have all committed hundreds of millions of dollars to multi-year computing contracts. That has become the basic cost of entry for building advanced AI. Recursive's deal matches that pattern in dollar terms but is structured differently: no equity, and a specific plan to build infrastructure tailored to self-improving AI rather than general-purpose computer systems.

That difference matters for understanding the competitive landscape. Other AI labs have largely tied themselves to the big cloud providers through deals that mix computing credits with ownership stakes, creating financial ties that go both ways. Recursive's agreement keeps the relationship straightforward: AWS sells computing power, Recursive buys it. Whether that turns out to be an advantage or a limitation will depend on whether the custom-built systems deliver real performance gains for self-improving AI that standard systems do not.

It is worth flagging the speed implied by Socher's comments. Calling a $410 million deal likely the smallest Recursive will sign in the coming years suggests a spending trajectory that would place the company among the highest-spending AI labs within a short window. A company that was four months old when it raised $650 million at a $4.65 billion valuation is, within roughly two more months, committing to $410 million in computing power alone. The question of how fast the company is spending its money is straightforward, and the answer will arrive partly in October, when Recursive says it will release its first products. If those products succeed, the path from research to revenue gets validated on an unusually fast timeline. If they slip, the computing commitments still have to be paid.

The "agent count" idea also deserves attention. Most AI labs talk about getting bigger in terms of model size, amount of text processed, or computing power used for training. Recursive is talking about getting bigger by having autonomous AI agents do work that would otherwise require human engineers. That is either a genuine bet on a new way to scale or a repackaging of familiar automation talk for investors who reward ambitious stories. The October product release will be the first real signal.