Mercor buys Deeptune for $10B valuation, after $43M Series A angel check from Brendan Foody

Brendan Foody, 23 and still the founder at the center of Mercor’s rocket ship, just used his $10 billion business to buy Deeptune, the simulation-environment startup he was listed as funding three months earlier. Fortune reports that Mercor acquired Deeptune on Thursday, and that Foody quietly helped fund Deeptune months before buying it, with financial terms not disclosed.

This is the part that matters for decision-makers: Deeptune’s tech is not “AI tooling” in the abstract. It builds the simulation environments where AI agents practice real-world enterprise tasks before touching production systems. Think of it as a flight simulator for agents learning how to use tools like Excel, Salesforce, and Slack. If you are trying to ship agentic workflows safely, you need exactly that layer, and Mercor now owns it.

Deeptune is Andreessen Horowitz-backed, and Foody’s name appears as an angel investor in its $43 million Series A just three months before the acquisition closed. Fortune also reports that Foody told them the angel check was written with acquisition already in mind, saying, "It was in a lot of ways the main motivation, actually." That quote is a reminder that in fast-moving AI markets, “strategic investing” is often less about passive upside and more about positioning for an eventual roll-up.

So why is the market paying attention now? Because AI training, especially for enterprise agents, is shifting toward full digital replicas of business software environments. Fortune points to Frontier Labs customers and peers like Anthropic and OpenAI needing these environments so their agents can practice, fail, and learn without breaking real systems. Mercor’s angle is that it already runs the network and scoring layer: it has a network of more than five million domain experts who build the tasks and scoring rubrics that tell a model whether it did the job right. Deeptune covers the apps those tasks run inside. Together, that is a “full stack” story, from tasks and evaluation to the environment where the agent actually operates.

Foody also claims Mercor is already the market leader in building these so-called reinforcement learning environments, and says that all of the Mag Seven except Tesla are customers. He frames the latest growth as aggressive even by AI standards. Mercor says it reached $2 billion in annual recurring revenue in June, up from $1 billion last year. Foody described the jump as "the fastest growth trajectory ever" from $1 million to $2 billion ARR in 24 months.

This acquisition lands in the middle of a credibility test Mercor did not ask for: a high-profile cybersecurity breach in March. Fortune reports that hackers exploited a supply chain vulnerability in LiteLLM, an open-source Python library widely used to build AI tools, and exfiltrated an estimated four terabytes of Mercor data. That included contractor Social Security numbers, passport scans, interview recordings, facial biometrics, and internal records showing how Anthropic labels safety data and how Meta builds reinforcement learning pipelines. The hacking group Lapsus$ claimed responsibility and offered the data for sale on Telegram.

The legal fallout is also concrete. A class action was filed in April against Mercor in California. In the telling, though, Foody spoke about the breach like it is already in the rearview mirror. He claimed, "Every frontier lab has expanded their relationship with us since the data breach," and continued, "The business grew from a $1 billion revenue run rate to $2 billion in ARR in the last four months." For executives, that statement is doing two jobs at once: it reassures customers about continuity and it argues that the breach did not permanently damage trust.

That said, there is a strategic shadow behind the confidence. Fortune notes that this could reflect customer loyalty, or it could be that Frontier Labs have nowhere else to go at this scale. In other words, buying Deeptune is not just about product depth. It is also about lock-in, reducing the number of reasons customers might shop elsewhere, especially when they are operating in environments where switching costs can be existential.

For peers building enterprise agent platforms, this deal signals where budgets and power are concentrating. If agent training requires simulation environments that map to real business tools, then owning that layer becomes a competitive advantage, not a nice-to-have. And if you are a board member watching revenue accelerate while security risk simmers in the background, the message is blunt: growth may continue, but the operational maturity required to protect sensitive data becomes part of the valuation conversation. Mercor is trying to buy its way into the full stack right when the market is deciding what it trusts.