Mercor Raises $750 Million at $20 Billion Valuation to Sell AI Agent Testing to Enterprises

The San Francisco company has done most of its work for frontier AI labs so far, drawing on a network of more than 5 million domain experts. It said the new money is meant to take that business to companies in every industry. Its argument is that each business will need its own way to check whether an AI agent can be trusted with a particular job.

The pitch matters because it addresses a question many corporate buyers of AI still cannot answer: how a model that does well on public benchmarks will perform inside their own systems. Mercor co-chief executive Adarsh Hiremath said a model can score well on a public benchmark and still fail on "the systems, policies and edge cases that define how a specific business runs." For each enterprise customer, Mercor says it builds a custom benchmark that measures models against that company's own standards. Post-training then starts from that benchmark.

Brendan Foody, the other co-chief executive, said "every job in the economy" will need its own AI agents, and that companies will need a way to train and evaluate those agents on their own work. He said Mercor spent the past two years building that infrastructure alongside the frontier labs.

That business with the labs has grown quickly. Foody said in a post on X in July that Mercor's annualised revenue had passed $2 billion in June.

Part of the enterprise offering relies on Deeptune, a New York startup that Mercor bought in July. Before the acquisition, Deeptune's team had spent two years recreating hundreds of enterprise software applications as practice environments for AI agents. Mercor said the new funding will pay for research and evaluation infrastructure, including more of those environments.

The company did not name a lead investor. Nvidia, Salesforce and Zoom Communications invested, as did returning backers Benchmark, Felicis and General Catalyst. Other participants were Fundomo, G Squared, Glade Brook Capital Partners, Sapphire Ventures, StepStone Group, XTX Ventures and funds managed by affiliates of Apollo Global Management.

Not all of the $750 million is new capital for the company. Foody wrote in a blog post on the round that it included a tender offer for employee shares, and Mercor has not said how the total is split between that sale and fresh money.

Mercor's previous round came in October 2025, when Felicis led a $350 million Series C that valued the company at $10 billion.

Jai Das, president, partner and co-founder of Sapphire Ventures, said frontier labs have been the main buyers of expert-driven training data so far. He expects that to change as open models become good enough for enterprises to build on, and said Mercor has built "something hard to replicate" from several years of work with the labs.

Mercor has not named any enterprise customers or given prices for the custom benchmarks. Those are the details that will show whether companies outside the AI labs will pay for this kind of evaluation work.