Nvidia Discusses Funding Its AI Data Supplier Mercor at a $20 Billion Valuation

Nvidia has discussed an investment in Mercor, a data labeling provider that helps the chip designer develop its open-source AI models, according to a person with knowledge of the process. The investment would be part of a $20 billion-valuation round. Existing investor General Catalyst has been in talks to lead the round.

Historically, most of Mercor’s revenue has come from the makers of closed-source AI models such as OpenAI, Google and Anthropic, The Information previously reported. But Mercor’s revenue from Nvidia is growing as the chip firm prioritizes the development of its Nemotron open-source models, which aim to compete with the world’s best open models.

Last quarter, Nvidia paid Mercor tens of millions of dollars, according to the person and another with knowledge of the contracts. Nvidia also uses data from other providers including Turing and Scale, in addition to its in-house data team, according to those people. (Nvidia invested in Scale’s 2024 funding round that valued the startup at $14 billion. Meta Platforms last year took a 49% stake in Scale and hired its CEO.)

Mercor and the other data labeling firms hire human experts in different fields to use and evaluate AI models in specialized tasks. The process generates data that the model makers use to refine the models to improve their capabilities in such tasks, which range from scientific research to legal and finance projects and, according to a Mercor job posting, even programming Nvidia’s graphics processing units.

Nvidia used Mercor data in its two latest Nemotron models, according to the company’s disclosures. Reflecting the tightening relationship between the firms, a handful of Mercor employees are focused almost entirely on their company’s work with Nvidia, said the two people.

Despite rising spending on such human-generated data, an executive said that Nvidia, like many other AI model developers, relies heavily on synthetic, or AI-generated, data to train its Nemotron models. Nvidia also receives critically useful data when developers access Nemotron models through OpenRouter, a service that helps developers use models from multiple providers, according to another person who worked on Nemotron.

The size of Nvidia’s proposed investment in Mercor and the total funding round couldn’t be learned.

Mercor generated $614 million in gross revenue in the first half of the year, up 70% from all of last year, The Information previously reported. It paid about 60% to 70% of its gross revenue to contractors as of last year, implying its net revenue in the first half of this year could be roughly between $180 million and $250 million. It expects gross revenue to total $2 billion this year.

It swung to a small operating profit last year and has told investors it expects the measure to rise to $226 million this year and $1.7 billion next year.

In recent months, Nvidia has been ramping up the pace and scale of its investing, though it primarily pumps money into its direct customers, such as upstart cloud providers, or the biggest users of its chips, such as OpenAI. In the quarter ending in April, it poured $18.6 billion into private companies and infrastructure funds, more than it invested in all of 2025.

This year alone, Nvidia has written separate $2 billion checks into cloud providers CoreWeave and Nebius, which buy Nvidia chips, as well as Lumentum and Coherent, which make lasers for Nvidia’s networking hardware. It has also funded a smattering of smaller startups that rent or buy Nvidia chips.

The three-year-old Mercor has raised about $520 million from investors including Benchmark and Felicis, according to PitchBook, and was valued at $10 billion in October.