W.081Mercor doubled its valuation to $20B in about a year while its revenue run-rate quadrupled. Investors now pay about 10 times its gross run-rate, half the multiple of its last round
Mercor, which pays doctors, lawyers, coders and other experts to produce training and evaluation data for AI labs, raised $750M at a $20B valuation on November 10, 2.0× the price of its round a little over a year ago. Its gross run-rate rose about 4.0× in the nine months to June, to more than $2B in June, so the valuation is about 10.0× that run-rate, from 20.0×.
The round
Nvidia, Salesforce and Zoom joined Mercor’s existing backers. Benchmark, Felicis and General Catalyst returned, with funds managed by Apollo affiliates and others; no lead was named. The round included a tender offer for employees’ shares, and none of the reports we found said how much of the $750M goes to the company. Nor did they say whether $20B is the valuation before or after the money; The Information had described the round Mercor was planning in July as $500M at $20B after the money.
Mercor says it will spend on research and evaluation infrastructure. It plans to sell more of that work to companies building their own AI agents, according to SiliconANGLE, beyond the AI labs that are now most of its revenue. Across its four priced rounds since 2024 it has raised about $1.2B, counting the full Series D, and its valuation has risen 80-fold since the Series A.
Mercor’s gross run-rate rose from about $75M to more than $2B in 16 months
Gross annualized revenue, $M, by the month each figure refers to: about $75M in February 2025 (TechCrunch), $100M in March and $500M in September 2025 (Forbes), and more than $2B in June 2026 (co-founder Brendan Foody). Dashed: the $2.8B run-rate for the end of 2026 Mercor forecast to investors, as The Information reported in July. Gross revenue includes what Mercor pays its experts.
What the revenue is
Most of each dollar Mercor bills goes to the experts it hires. It pays them 60–70% of gross revenue, The Information reported, and its gross margin was 27% in 2025 and 33% in the second quarter of 2026, according to BigGo’s account of documents The Information reported on. On a 30–40% share kept, the June run-rate is about $600M–$800M a year after paying experts, and the valuation 25–33 times that.
AI labs paid most of the bills. First-half gross revenue was $614M, The Information reported, about 91% of it from companies that build foundation models; reports differ on whether that means OpenAI and Anthropic alone. That share would be about $559M of lab spending in six months through one supplier. The first half annualized is $1.23B, below the June run-rate, consistent with revenue rising through the half.
Valuation has doubled since the Series C; the step-ups before it were 8 and 5 times
$M. Mercor’s valuation at each priced round since 2024. The Series A date is given only as 2024. Whether $20B is before or after the new money was not disclosed.
At about 10.0× its run-rate, Mercor is priced below Scale AI’s multiple at Meta’s investment
Valuation as a multiple of revenue. Mercor: each round’s valuation over the gross run-rate nearest before it ($75M in February 2025, $500M in September 2025, $2B in June 2026), and the Series D over the $2.8B forecast for the end of 2026 (dashed). Peers, in grey, on mixed bases and dates: Scale AI at about $29B (Meta, June 2025) over 2025 revenue of just under $1B; Invisible at $2B (September 2025) over 2024 revenue of $134M; Turing at $2.2B (March 2025) over $167M of ARR.
What this changes on inthesingularity.ai
- Who funds the AI labs?: spending by labs on human data is a cost line the pack does not break out. Mercor’s first half, about $559M from labs on The Information’s figures, is one supplier’s floor for it.
- Worth tracking: Nvidia now invests in a data supplier whose data, The Information reported citing Nvidia’s disclosures, went into Nvidia’s two latest Nemotron models; and whether Mercor’s next run-rate figure keeps pace with the $2.8B it forecast for December.
Sources and method
| Item | Figure | Source |
|---|---|---|
| Series D | $750M at a $20B valuation, including an employee tender offer; investors including Nvidia, Salesforce, Zoom, Benchmark, Felicis, General Catalyst and Apollo-managed funds; Series C was $350M at $10B in Oct 2025; run-rate passed $2B in June | SiliconANGLE, Nov 10 |
| Series D | Round size and valuation; more than 5 million experts in its network | The Next Web, Nov 10 |
| Series D | Valued at $20B | Reuters, via Investing.com, Nov 10 |
| Series C | $350M at $10B, led by Felicis, five times the Series B; contractors paid more than $1.5M a day | TechCrunch, Oct 27, 2025 |
| Series A, B | Series B $100M at $2B led by Felicis; Series A $32M at $250M; about $75M of annualized revenue | TechCrunch, Feb 20, 2025 |
| Run-rate | $100M in March 2025 and $500M in September 2025 | Forbes, Oct 30, 2025 |
| Run-rate | Foody: crossed $2B of gross annualized revenue as of June | TechCrunch, Jul 8, 2026 |
| Financials | H1 2026 gross revenue $614M; about 91% from AI labs; end-2026 run-rate forecast $2.8B; gross margin 27% in 2025 and 33% in Q2 2026 | The Information, via AI Weekly, Jul 2026 |
| Financials | End-2026 run-rate forecast $2.8B; gross margin 27% in 2025 and 33% in Q2 2026; planned round of $500M at $20B post-money (The Information, Jul 9) | BigGo Finance, Jul 22, 2026 |
| Nvidia | Nvidia used Mercor data in its two latest Nemotron models, per Nvidia’s disclosures (The Information) | Dealroom, Aug 20, 2026 |
| Financials | 60–70% of gross revenue paid to contractors | The Information, via Dealroom, Aug 2026 |
| Peers | Meta paid $14.3B for 49% of Scale AI, valuing it at about $29B | Reuters, via The Star, Jun 13, 2025 |
| Peers | Scale AI revenue just under $1B in 2025 | Forbes Australia, May 2026 |
| Peers | Invisible raised $100M at more than $2B; 2024 revenue $134M | SiliconANGLE, Sept 16, 2025 |
| Peers | Turing raised $111M at $2.2B; ARR $167M | TechCrunch, Mar 6, 2025 |
Step-up: 20000 ÷ 10000 = 2.0; 2000 ÷ 250 = 8; 10000 ÷ 2000 = 5; 20000 ÷ 250 = 80. Run-rate: 2000 ÷ 500 = 4.0 (September 2025 to June 2026; the rounds were about 12.5 months apart). Multiples: 20000 ÷ 2000 = 10.0; 10000 ÷ 500 = 20.0; 2000 ÷ 75 = 26.7; 20000 ÷ 2800 = 7.1. Because June’s figure is “more than” $2B, the current multiple is at most 10.0×. Net of experts: 2000 × (1 − 0.7) = 600; × (1 − 0.6) = 800; 20000 ÷ 800 = 25.0; ÷ 600 = 33.3. Lab revenue: 614 × 0.91 = 558.7; first half annualized 614 × 2 = 1228. Raised: 32 + 100 + 350 + 750 = 1232, excluding the seed round and counting the tender. Peers: Scale AI 29000 ÷ 1000 = 29.0; Invisible 2000 ÷ 134 = 14.9; Turing 2200 ÷ 167 = 13.2. Mercor’s Series D announcement was not available to us; terms are as reported by Reuters, SiliconANGLE and The Next Web.
Wire reports are written within hours of the news, on a six-hour schedule. Each lists its sources and shows the arithmetic behind its figures. Not investment advice.