Mercor doubles to $20bn, slower than last time
Mercor, which pays experts to train and test AI models, raised $750m at a $20bn valuation on Tuesday, twice its price of October 2025. The last step was fivefold in about eight months. This one took 379 days, and the multiple of revenue has halved.
Clay Gilliland, CC BY-SA 2.0, via Wikimedia CommonsMercor, a San Francisco company that recruits professionals from hundreds of occupations to train and grade AI models, has raised $750m in a Series D round at a $20bn valuation, it said on Tuesday 10 November. The round included a tender for employees' shares. No lead investor was named.
The valuation is double the $10bn Mercor was given in October 2025, when it raised $350m in a Series C led by Felicis, TechCrunch reported at the time. That round had itself been five times the $2bn price of a $100m Series B earlier that year. The new step is smaller and slower: 2x in 379 days, by our count, against 5x in about eight months.
Thirteen investors are named in the announcement by Brendan Foody, co-founder and co-chief executive: Benchmark, Felicis, Fundomo, funds managed by affiliates of Apollo, G Squared, General Catalyst, Glade Brook Capital, Nvidia, Salesforce, Sapphire Ventures, StepStone Group, XTX Ventures and Zoom. Three of them, Nvidia, Salesforce and Zoom, are also companies that build or sell AI products. The round was announced the same week Nvidia's own shares lost about $428bn of market value in two days.
| Round | Announced | Raised | Valuation | Step-up | Time since previous round | Company figure at the time |
|---|---|---|---|---|---|---|
| Series B | February 2025 | $100m | $2bn | n/a | n/a | n/a |
| Series C | Mon 27 Oct 2025 | $350m | $10bn | 5x | about 8 months | "on track" for $500m ARR; $1.5m a day paid to contractors |
| Series D | Tue 10 Nov 2026 | $750m | $20bn | 2x | 379 days | $2bn gross run rate in June 2026, per Dealroom citing the company |
The multiple has come down. In October 2025 the company had told investors it was on track for $500m of annual recurring revenue, TechCrunch reported, which made the $10bn price 20 times that figure, by our division. Dealroom reported on Tuesday that Mercor's annualised gross revenue run rate reached $2bn in June 2026, attributing the figure to the company. On that number the new price is 10 times run rate. Neither the release nor the blog post gives a revenue figure, so the comparison rests on figures Mercor has put out elsewhere, and the two are not measured the same way: one is a forecast of recurring revenue, the other a gross monthly rate annualised. For comparison, Modal's $15.75bn price last month was about 15 times its own run-rate figure.
"Gross" is the word to watch. Mercor's business is paying people. In October 2025 it was paying more than $1.5m a day to its contractors, TechCrunch reported, to more than 30,000 experts earning over $85 an hour on average. At that rate, payouts come to about $548m a year, by our multiplication, slightly more than the $500m of recurring revenue it was aiming for at the time. The public numbers do not let us say how much of a gross run rate the company keeps after paying its experts, and the figures come from different points in the year.
The headcount figures have also changed meaning. The October 2025 figure was more than 30,000 experts doing paid work. The new announcement speaks of "more than five million experts in our network" across more than 800 US occupations. A network is people who have signed up, not people being paid in a given month, and the release does not say how many are active.
The money is meant to move Mercor beyond the frontier labs that are its core customers. The company plans to spend on evaluations, post-training infrastructure and research, and on environments in which AI agents learn professional workflows, helped by Deeptune, which it bought earlier this year. "Every company deploying AI needs to know whether it works, and the answer is different for each one," Adarsh Hiremath, the other co-founder and co-chief executive, said in the release. Foody's post adds a ratio of the company's own: "We already spend 3x as much on inference as on employee salaries."
Concentration is the risk investors are paying around. Dealroom cited an unnamed report that 91 per cent of Mercor's first-half 2026 gross revenue came from foundation-model companies. Komposite has not been able to check that figure. Those customers are spending heavily, but their own financing has tightened at the margin: OpenAI cut the size and price of its latest round last month, and labs have been changing how they train, as when OpenAI took frontier training off the live internet in October.
Mercor has opened offices in New York and London alongside San Francisco, according to the post. It has not said whether it will report revenue on a net basis, or how many of its five million experts were paid in the latest quarter. Those are the two numbers that would let an outsider check the multiple.