Mercor Doubles to $20 Billion With a $750 Million Series D as Nvidia and Apollo Chase Agent Training Data

Mercor, the San Francisco startup that has become one of the key suppliers of expert training data to frontier AI labs, said on Tuesday it raised a $750 million Series D at a $20 billion valuation. The price is exactly double the $10 billion the company fetched in October 2025, when it closed a $350 million Series C — a doubling that took roughly thirteen months.

The round drew participation from Benchmark, Nvidia, Felicis, Fundomo and funds managed by affiliates of Apollo, according to the company and Reuters. General Catalyst was also named among the investors in early reports of the deal. Mercor's earlier backers include Jack Dorsey and Peter Thiel, an unusual pairing that reflects the company's origin story: it began as an AI-powered hiring platform before pivoting into data supply.

That pivot is now the core of the business. Mercor connects technology companies with subject-matter experts — doctors, lawyers, engineers, financiers — whose knowledge is used to train, test and improve AI models and agents. As labs race from chatbots toward agents that can complete real professional work, demand has shifted from generic question-answer data toward expert evaluation of agent behavior on realistic tasks.

"Every job in the economy will need its own AI agents and every company will need a way to train and evaluate them on its own work," co-CEO Brendan Foody said in the announcement. The company said the new funding will go toward building out exactly that: infrastructure for training and testing AI agents on professional tasks.

The raise lands in a private market that remains strikingly concentrated on AI. According to Crunchbase data cited alongside the announcement, AI-driven startups raised $102 billion in the third quarter of 2026 — 64 percent of all global venture capital. Reports at the time of the Series C put Mercor's annualized run rate around $450 million; the company has not published an updated figure with this round.

The company's own record is not spotless. On March 31, Mercor acknowledged it had been the target of a data breach — disclosed six months after the Series C closed. The incident did not appear to slow fundraising, but it underlines the sensitivity of a business whose inventory is, in effect, expert knowledge and client model information.

All valuation and investor figures above come from the company's announcement and wire reporting and have not been independently audited. What the round does signal clearly is where the money thinks the next bottleneck sits: not in model architecture, but in the human expert layer that grades whether an agent can actually do the job it was built for. At $20 billion, investors are betting that grading layer becomes as strategic as the models themselves.