Mercor's $20 billion Series D funds a move from AI labs to enterprise customers

Mercor raised $750 million at a $20 billion valuation on November 10. The company's own posts show a proven first market, the labs, and a second one, enterprises, it has yet to evidence.

Add FinanceInTime as a preferred source on Google
Illustration of an empty flight simulator cabin on hydraulic legs in a training hall at early morning
Illustration. Not a photo of any real training facility.Illustration: FinanceInTime

Mercor, a San Francisco company founded in 2023 that pays experts to train AI models, said on November 10 that it raised $750 million at a $20 billion valuation, and described the money as a way to take its expert network from AI labs to enterprises, according to its announcement. Read against the company's earlier posts, the round marks a change in who it expects to sell to.

From an implied $2 billion to $20 billion in three rounds

Mercor was founded in 2023, according to the company's mission page, by Brendan Foody and Adarsh Hiremath. In its Series C announcement on October 27, 2025, it said it raised $350 million at a $10 billion valuation, 5x its Series B valuation. That implies a Series B valuation of $2 billion (computed from $10 billion divided by 5).

The Series D valued the company at $20 billion, twice the Series C figure (computed). The company said the round "included an employee tender," so it did not say how much of the $750 million is new money for the business. Benchmark, Felicis, General Catalyst, NVIDIA, Salesforce and Zoom were among the investors it named.

The first customers were the labs

The company's own description of its business, in the Series C post, was that it connects "human expertise with leading AI labs and enterprises." In the Series D post, Foody wrote that "We’ve created a new category of work with more than five million experts in our network getting paid to train frontier AI models based on their unique knowledge." The mission page says experts on the platform are paid over $5 million per day.

Growth put pressure on the company's own systems. An engineer wrote that before August 2025 the service handling contractor contracts managed around 3,000 active contracts a month, and that between July and October contract volume grew roughly another 10x. At one point, the post said, more than 60% of engineering time went to manual triage and maintenance.

A second product: tests and practice grounds for agents

The Series D post describes the second product. Foody wrote that "We spent the last two years building the infrastructure to train and evaluate frontier models on real-world work with human intelligence," and that the company is "now bringing those same capabilities to enterprises across the economy."

Two steps in 2026 fit that description. On July 9, Mercor said it would acquire Deeptune, which it described as a company building environments for reinforcement learning, where AI models practice tasks. Deeptune's team "has recreated hundreds of enterprise applications," the post said, and Andreessen Horowitz had led its $43 million Series A earlier in the year. Foody wrote that Mercor was already a Deeptune customer and that he had been an early investor.

On October 8, Mercor said it was building a research team. It reported that a post-training guide raised the score of Qwen3.5-397B on its APEX-Agents benchmark from 16% to 27% on the Pass@1 measure, a gain of 11 percentage points (computed). The APEX benchmarks, the post said, cover management consulting, investment banking, corporate law, accounting and software engineering.

What the company says the money is for

The Series D post says the funding "accelerates our investment in the infrastructure and research for evals and post-training and in building more environments where agents can learn complex workflows." The company said it opened offices in New York and London this year, and noted that the United States has more than 800 distinct occupations, "each containing workflows that models have yet to be evaluated against."

Foody also made a claim about the company's own costs: "We already spend 3x as much on inference as on employee salaries, and it pays for itself by making our team better at their jobs." He added that, if the ratio holds as AI spreads, it "points to something closer to double-digit GDP growth over the next five years." That is the company's projection, not a finding.

What the sources leave open

None of the company's posts gives revenue, profit or the number of enterprise customers. The Series D post says Mercor's capabilities "are now being adopted by big tech, developers of open models, application layer companies, and the Fortune 500," without naming a count.

The company also disclosed a security incident. In a post dated June 25, it said it was affected in late March 2026 by a supply chain attack on LiteLLM, an open-source tool. Of its "nearly five million experts," it said, only a "very limited subset had sensitive information affected," and it reported that "all frontier labs have increased their work with us over the last few months." That last statement is the company's own and no customer is quoted.

Enterprise demand is the claim to test

We think the sources describe a company with a proven first market and an unproven second one. The first is clear in the numbers it published: a network of five million experts, contract volume that grew about 10x in a few months, and valuations that moved from an implied $2 billion to $20 billion. The second rests on a thesis stated in the company's own words and on two 2026 steps, the Deeptune purchase and the research team, that make the thesis plausible without proving it.

Our reading is that the opportunity is real if enterprises want what the labs already buy: custom tests and practice environments for their own workflows. The threat is concentration. A business that depends on a few labs for its revenue is exposed if their needs shift, and the posts do not say how large that share is. We expect the evidence to come in the form of named enterprise customers or a revenue figure, and we would change this view if the company publishes neither over the next one or two funding milestones. A valuation of $20 billion is a price set by the company's investors; it is not a forecast of what the business will earn.

The view in this final section is FinanceInTime's editorial opinion, for information only. It is not investment, tax, legal or other advice, and it does not take your circumstances into account. Forecasts are uncertain and past results do not guarantee future ones.

Sources

  1. Organizing human intelligence for the entire economy — MercorPrimary source
  2. Unlocking human potential in the AI economy — MercorPrimary source
  3. Mercor Mission — MercorPrimary source
  4. Mercor to acquire Deeptune — MercorPrimary source
  5. Update on Mercor security incident — MercorPrimary source
  6. Why Mercor is building a research team — MercorPrimary source
  7. Mercor Blog — MercorPrimary source
  8. When volume grew 10x in a month and we had one week to fix it — MercorPrimary source
Add FinanceInTime as a preferred source on Google

More in Startup Radar

View All Startup Radar

Get Finance News in Your Inbox

Daily headlines, market insights and analysis from our editors.