Mercor raises $750M at $20B valuation to expand AI agent evaluation to enterprises

AI training data company Mercor said it has raised $750 million in a Series D round at a $20 billion valuation. The company supplies human experts, drawn from a network of more than 5 million, to train and evaluate AI models, and most of that work so far has been for frontier AI labs.

Mercor plans to use the funding to move into enterprises, which it says need their own way to test whether an AI agent can be trusted with a job. For each customer it builds a custom benchmark based on that company's standards, and post-training starts from there. Part of that work runs on software from Deeptune, a New York startup Mercor bought in July.

The new money will go toward research and evaluation infrastructure, including more enterprise-style environments. Mercor did not name a lead investor, and the round included a tender offer for employee shares, though the company has not said how the $750 million is divided between that and new capital.

Teams building agents for enterprise use may see more vendors offering custom benchmarks built around a company's own workflows and policies, not only public tests. Developers should note that Mercor has not said how much of the $750 million is fresh capital versus employee share purchases, so the amount available for new evaluation tooling is unclear.