Mercor has doubled its valuation in about thirteen months. The San Francisco company announced a $750 million Series D on November 10 at a $20 billion valuation, up from $10 billion when it raised a $350 million Series C in October 2025. Investors named include Benchmark, Felicis, General Catalyst, Menlo Ventures, Nvidia, Salesforce and Zoom. No lead was named.
Mercor’s business is human expertise sold to AI builders. Its network of more than five million experts (lawyers, doctors, bankers, engineers) grades model output and produces the training data that frontier labs use for post-training. Co-founder Brendan Foody says run-rate revenue went from $1 million to $3 billion in 27 months. The release itself doesn’t disclose revenue.
The pivot to enterprises
The interesting part is where the money goes. Mercor spent two years selling to a handful of frontier labs. The new round is aimed at selling the same tooling to ordinary companies. Its argument is that every business putting AI agents to work needs to know whether they actually do the job, and the answer differs for a bank, a hospital and a law firm. Co-CEO Adarsh Hiremath framed it that way in the announcement.
So the funding goes into evaluation and post-training infrastructure, including simulated work environments where agents practise professional workflows before they touch production systems. That capability came largely from Deeptune, which Mercor bought earlier this year. Deeptune had spent two years recreating hundreds of enterprise applications as training environments. Mercor also runs APEX, a public benchmark series that scores AI on economically valuable professional tasks.
Why the labs-only model needed a second leg
Revenue concentrated in a few AI labs is fragile. Any of them can decide to build evaluation in-house, and Mercor lost Meta as a customer after a data breach earlier this year. Spreading the same product across thousands of companies is the obvious hedge.
It’s also a bet on where AI spending goes next. Training frontier models is a game for a few players. Testing whether a deployed agent handles your insurance claims or your contracts correctly is a problem every large company will have. If Mercor can turn its expert network into a standard way of measuring that, the $20 billion price looks less strange.