Mercor at $20 Billion: AI's Most Expensive Input Is Still Human Judgment

On November 10, San Francisco-based Mercor announced a $750 million Series D at a $20 billion post-money valuation, including an employee secondary. Benchmark, General Catalyst, Felicis, StepStone and Apollo-affiliated funds invested alongside NVIDIA, Salesforce and Zoom; no lead was disclosed. A year ago the company raised $350 million at $10 billion. Forbes estimates the three 23-year-old founders - CEO Brendan Foody, CTO Adarsh Hiremath and chairman Surya Midha, teammates on a high-school debate team - each hold about 21%.

From a recruiting desk to a judgment supply chain

Founded in 2023 to place freelance Indian engineers with US companies, Mercor pivoted when it saw who was really paying: AI labs that need doctors, lawyers and senior engineers for post-training data and model evaluation. Today it claims a vetted network of more than five million experts, counts OpenAI among 100-plus customers - double last year - acquired the reinforcement-learning environment company Deeptune earlier this year, and runs APEX, a public eval benchmark. The new capital goes into evaluation and post-training infrastructure: testing what agents do before businesses let them loose.

Read the $3 billion carefully

Founder Foody says annualized total revenue grew from $1 million to $3 billion in two-plus years, with the latest month near $250 million. Those are unaudited company figures - and gross ones. In a marketplace, experts are paid out of that number; what accrues to the platform is the take rate. Mercor disclosed $614 million of gross revenue for the first half of 2026 and said inference spending already runs at three times payroll. At $20 billion, the round prices the business at roughly six to seven times annualized gross volume - a GMV multiple borrowed from e-commerce, not a net-revenue multiple from software.

The demand signal is still real: someone is writing very large cheques for human expertise, by the hour and by the task, because model capability now bottlenecks on exactly that. For founders across Asia Pacific the lesson is not to chase the model but to own the supply of scarce judgment - domain experts, evaluators, operators who can certify what an agent actually did. That is the asset this region is unusually rich in. Chainfir Capital holds no position in Mercor or any company named above.