Mercor raises $750 million in Series D – the week's second-largest round in the US

The round is confirmed independently of the company itself, and places the firm behind only one larger US funding round that week. But most of the growth figures underpinning the valuation — 5 million experts, a run-rate in the billions…

Abstract illustration in dark tones, generated by AIMag.
Illustration

Mercor raises $750 million in Series D – the week's second-largest round in the US

The round is confirmed independently of the company itself, and places the firm behind only one larger US funding round that week. But most of the growth figures underpinning the valuation — 5 million experts, a run-rate in the billions, its own benchmarks — come from Mercor, and a data breach in March casts a shadow over the story.

The news itself

Mercor, a company that builds AI training and evaluation data using human expertise, announced on 10 November 2026 a $750 million Series D at a $20 billion valuation. That is according to the company's press release via Business Wire (0b138dc9).

The investor list is long: Benchmark, Felicis, Fundomo, funds managed by subsidiaries of Apollo, G Squared, General Catalyst, Glade Brook Capital, NVIDIA, Salesforce, Sapphire Ventures, StepStone Group, XTX Ventures and Zoom (0b138dc9).

The round is independently confirmed: Crunchbase News' weekly roundup of funding rounds places Mercor as the week's second-largest US round, beaten only by Mirendil's $1 billion Series A (58377551).

The money is anchored in a new market: enterprise customers, not just frontier labs. According to the company, the funds will accelerate Mercor's evaluation ("evals") and post-training business, so that companies in any industry can measure and set standards for how reliable AI agents are at task execution, and at what cost (0b138dc9).

What Mercor actually does

Mercor's core is a network of more than 5 million domain experts — lawyers, accountants, engineers and other professionals — who train and evaluate AI models within their own fields. The figure is the company's own, but it is repeated in the press release and the coverage (6aa42535; 0b138dc9).

The model has scaled quickly, at least on the company's own numbers: Mercor paid out more than $2 million per day to experts in March 2026, and had over 700 full-time employees as of November 2026 (6aa42535). The annual run-rate, measured as gross billings, reportedly passed $1 billion in February 2026 and $2 billion in June — up from around $1 million two years earlier. Contracts typically retain 60 to 70 percent of gross billings (6aa42535).

The only independent revenue figure in the coverage comes from The Information: $614 million in gross revenue in the first half of 2026, up 70 percent from the year before (6aa42535).

The enterprise push: benchmarks first, then post-training

Mercor's next chapter — and what the money will fund — is about moving the same model from frontier labs to enterprise customers. Co-founder and co-CEO Adarsh Hiremath explains the mechanism in the press release (0b138dc9):

"A model can score well on a public benchmark and still fail at the systems, policies and edge cases that define how a specific business is run. We build custom benchmarks together with each enterprise to measure models against its own standards, and that benchmark becomes the starting point for post-training."

The company has already built tools for this: the APEX benchmarks measure AI on economically valuable work. In October 2026, AI reportedly outperformed 12 licensed, state-certified accountants (CPAs) on APEX-Accounting, according to Mercor. The company has also post-trained an open model of 397 billion parameters using the SkyRL framework and achieved a 70 percent relative improvement on APEX-Agents — figures that originate from Mercor itself (6aa42535).

In July 2026, Mercor agreed to acquire Deeptune, a team that has spent two years recreating hundreds of enterprise applications as training environments — simulated office systems where AI agents can practice (0b138dc9; 6aa42535).

The founders and the numbers they themselves cite

The three founders Brendan Foody, Adarsh Hiremath and Surya Midha are all 23 years old. Forbes estimates that each holds around 21 percent of the company and is therefore worth roughly $4.2 billion — up from around $2.2 billion in October 2025, when the $350 million Series C valued Mercor at $10 billion (9517d3e7; 6aa42535). The wealth estimates are just that — estimates based on assumed ownership stakes, not verified figures.

The same day the round became known, Foody posted on X that Mercor scaled from $1 million to $3 billion in run-rate in 27 months — "fastest in history to $3 billion," according to the post cited by NDTV (9517d3e7). He also cited $250 million in gross sales for the latest month and a customer base that has more than doubled to over 100 — all the company's own claims.

It is worth noting that $3 billion (November, Foody) and $2 billion (June, the company's own figure) are not explicitly reconciled in the available sources. That could reflect four months of growth, but the definitions of "gross billings" and "gross annual revenue" are not specified.

The competitive picture

Mercor is still not the largest on paper. Scale AI was valued at around $29 billion when Meta bought 49 percent of the company in June 2025, and Forbes estimates Surge AI at around $24 billion (9517d3e7).

Staffing Industry Analysts (SIA) points out that before this round Mercor was the most highly capitalized B2B platform, with $450 million in funding (0ef92cfc). SIA vice president of research Brian Wallins frames the condition this way:

"There may be no larger market for AI to disrupt than the global market for human talent. Mercor's explosive growth demonstrates the scope of this opportunity, as demand for human expertise expands from training frontier AI models to enabling AI agents to perform real enterprise workflows."

The investors point to the same shift. Jai Das, president, partner and co-founder at Sapphire Ventures, says frontier labs have been the primary buyers of expert-driven training data, "and that is changing" (33ec277a).

Setbacks and open questions

The story has a shadow side. In March 2026, Mercor was hit by a data breach — described in the coverage as a supply chain attack late in March — and the company later announced additional cybersecurity measures (6aa42535; 0ef92cfc). According to AIMag, citing WIRED, Meta paused all work with Mercor while the company investigated the breach. The details of this sequence are, however, partly unclear in the sources, and WIRED's original reporting is not available here — the pause should therefore be read as reported, not fully verified.

Other uncertainties: the majority of the growth figures — 5 million experts, payouts of more than $2 million per day, the APEX results, Deeptune's "hundreds of recreated apps," customer counts — come from Mercor itself. Several outlets have reproduced the press release without independent verification. Hiremath's role also varies between sources: the press release and AIMag call him co-CEO, while NDTV/Forbes describe him as CTO.

What does stand on solid ground: the $750 million round, the $20 billion valuation, the 10 November 2026 date, the investor list and the ranking as the week's second-largest round (0b138dc9; 58377551).

AIMag.no
AIMag.no
The AIMag.no editorial team covers artificial intelligence, tools, research, and regulation.

Get the best of AI MAG in your inbox

News, analysis, and ideas at the intersection of AI and society.