Mercor Raises $750M at $20B Valuation for AI Agent Evals

By Nathan Brooks

Mercor raised $750 million in a Series D at a $20 billion valuation, the San Francisco startup said Tuesday. The financing doubled the roughly $10 billion valuation reached after its $350 million round in October 2025, according to Reuters.

Mercor supplies expert knowledge for training, testing and improving AI models and agents. It began in 2023 as an AI-powered hiring platform and now connects AI labs and technology companies with subject-matter experts.

What will $750 million buy?

Mercor says the proceeds will accelerate its evaluation and post-training business, including simulated environments where agents learn professional workflows. The company builds custom benchmarks with each enterprise, measures how reliably agents complete tasks and at what cost, then uses those results as the starting point for post-training.

Those simulations build on Deeptune, a team Mercor acquired earlier this year. The acquired team spent two years recreating hundreds of enterprise applications as training environments for agents.

"Every job in the economy will need its own AI agents," co-CEO Brendan Foody said. His forecast frames Mercor’s opportunity broadly, although the financing announcement provides no market-size estimate or rollout schedule.

Mercor says its expert network contains more than 5 million experts. Customers include frontier AI labs, large technology companies, AI startups and Fortune 500 businesses, but Mercor named none of them.

Who backed the round?

Mercor did not name a lead investor. Investors in the round included:

• Benchmark, Felicis, Fundomo, G Squared, General Catalyst, Glade Brook Capital and Nvidia.
• Salesforce, Sapphire Ventures, StepStone Group, XTX Ventures, Zoom and funds managed by affiliates of Apollo.

Does the $20 billion valuation match the reported revenue?

Mercor’s release did not disclose revenue. In August, The Information reported that Mercor’s annualized gross revenue had reached $2 billion in June, while also saying General Catalyst was discussing a lead role in the round. The Information’s revenue figure and account of the lead-investor talks were not confirmed in Mercor’s financing announcement.

The same August reporting said Nvidia had paid Mercor tens of millions of dollars for expert-curated data used in Nvidia’s Nemotron models. Nvidia’s role as both an investor and a reported customer gives its participation a commercial context, though Mercor did not quantify the data purchase.

The round sits within a larger funding surge. Reuters cited Crunchbase data showing that AI startups took $102 billion, or 64% of global venture funding, in the third quarter.

Why does evaluation now carry commercial value?

Measuring models on paid professional work has become a standard test for AI labs, with Anthropic’s Claude Fable 5.2 on the Remote Labor Index serving as an example. Mercor also publishes the APEX benchmark series, giving it assets in both published benchmarking and enterprise-specific testing.

The commercial proposition is straightforward: companies pay for assessments that reveal whether an agent can finish real work reliably and economically. Mercor’s expansion into post-training turns that assessment into a development loop, using test results to improve a model or agent.

Can Mercor scale beyond the reported revenue figure?

At $20 billion, investors are underwriting more than expert matching: they are also funding the prospect that assessment becomes a development loop. More than 5 million experts and hundreds of recreated enterprise applications give Mercor substantial potential capacity to supply evaluations and training environments. Commercial execution now depends on whether enterprises commission repeat post-training projects and whether the task costs measured by Mercor support broader deployment.

The $2 billion annualized gross revenue figure came from The Information rather than a Mercor disclosure, so the announcement does not establish its basis or durability. Pricing will depend in part on the cost of completed agent work, because reliability and cost are the two variables Mercor says its benchmarks measure.

Going forward, the financing funds more evaluation capacity; customers will decide whether the rubric earns recurring revenue.