On 10 November 2026, Mercor announced a $750 million Series D round at a $20 billion valuation. The company runs a marketplace that pays domain experts to train and evaluate frontier AI models, and the round doubles the $10 billion valuation it set in October 2025.
What was announced
Mercor published the news on its own blog on 10 November 2026. The round was priced at $20 billion and included an employee tender alongside new investment. Named participants were 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. The announcement did not name a lead investor. Crunchbase News listed the round among the week’s biggest funding rounds in its 13 November 2026 roundup.
The previous round was a $350 million Series C at a $10 billion valuation in October 2025, reported by TechCrunch. TechCrunch also reported in July 2026 that Mercor was in talks at a $20 billion valuation.
What Mercor says it does
In the announcement, Mercor says it has more than five million experts in its network “getting paid to train frontier AI models based on their unique knowledge”. It also says it works with enterprises and experts in their fields to build custom evaluations based on the company’s actual workflows. The company says the new funding will go toward infrastructure and research for evaluations and post-training, and toward environments where AI agents can learn complex workflows. It also says it has opened offices in New York and London. The announcement does not break down the network by qualification, so it does not say how many of those experts hold doctorates or work as researchers.
Why it matters for researchers and research offices
The round is a data point on how much capital is flowing toward expert human judgment as an input to AI development. Three practical questions follow for universities and research offices. These are editorial observations, not claims made by Mercor.
- Researcher time. Expert annotation and evaluation work for AI companies is paid outside the normal grant and appointment structures. Research offices may need to check how it interacts with institutional outside-activity and conflict-of-commitment policies.
- Confidential and unpublished material. Evaluation tasks built from real workflows raise questions about what a researcher may share, including unpublished data, grant content or material under a confidentiality agreement.
- Evaluation methods. Expert-written tests are one way AI developers measure model ability. Researchers who read results from such tests can use frameworks such as model evaluation suites and AI evaluation cards to ask what was tested and how.
For background on the techniques this market supports, see RLHF and training data provenance. For an example of a research-level expert-written benchmark, see FrontierMath.
What is not yet known
The Series D announcement did not disclose revenue, expert pay rates or the share of the network made up of researchers. Anyone weighing the opportunity should rely on the terms they are offered directly rather than on funding headlines.







