Mercor acquires a16z-backed Deeptune to build AI-agent training environments

Mercor, the AI data-labeling company last valued at $10 billion, said on July 9, 2026, that it is acquiring Deeptune. The startup builds simulation environments where AI agents rehearse real-world software tasks before touching production systems.

The deal gives Mercor a way to generate the practice environments — for applications like Excel, Salesforce and Slack — that increasingly feed reinforcement learning for AI agents, a market where demand for training data has outrun supply.

Deeptune raised a $43 million Series A earlier in 2026 led by Andreessen Horowitz. Mercor co-founder and chief executive Brendan Foody had personally invested in that round as an angel, three months before agreeing to buy the company. Terms of the acquisition were not disclosed, and Deeptune’s team is joining Mercor’s New York office.

Mercor said its annualized revenue run rate crossed $2 billion in June 2026, roughly double the figure four months earlier. The company has positioned itself as a supplier of human expertise and training data to AI labs, and buying Deeptune moves it further into the tooling that turns that data into agent behavior.

The purchase lands as Mercor is reportedly weighing a new funding round that would double its valuation, a signal of how quickly capital is flowing to companies that supply the raw material for training AI agents. Mercor did not detail how Deeptune’s environments will be packaged or sold.