Nvidia in talks to back Mercor at $20B valuation
What's the deal? Nvidia has discussed investing in Mercor, the AI data supplier that helps model developers create and evaluate specialised training data. The potential investment would form part of a funding round at a $20 billion valuation, although the size of Nvidia's proposed cheque and the total round have not been disclosed. Existing investor General Catalyst has been in talks to lead the round, according to a person with knowledge of the process.
Why now? Mercor's relationship with Nvidia has been deepening as the chip company prioritises Nemotron, its family of open-source AI models. Nvidia used Mercor data in its two latest Nemotron models, according to Nvidia's disclosures, and several Mercor employees are now focused almost entirely on work for Nvidia. The companies' commercial relationship has also expanded: Nvidia paid Mercor tens of millions of dollars in the latest quarter, according to people familiar with the contracts.
What Mercor provides: Mercor recruits human experts in fields including science, law, finance and engineering to complete and evaluate specialised tasks. Their work produces training and evaluation data that helps AI developers improve model performance in areas where generic internet-scale data is insufficient. A Mercor job posting even described work involving the programming of Nvidia graphics processing units.
The customer mix: Historically, most of Mercor's revenue came from developers of closed-source models such as OpenAI, Google and Anthropic. Nvidia's spending is growing as it builds Nemotron and competes in the open-model market. Nvidia also uses other data suppliers, including Turing and Scale, alongside an in-house data team. The chipmaker invested in Scale's 2024 funding round, which valued that company at $14 billion.
The numbers: Mercor generated $614 million in gross revenue in the first half of 2026, up 70% from all of 2025, and its annualised gross revenue reached $2 billion in June. The company expects gross revenue to total $2 billion for the full year. Because it paid roughly 60% to 70% of gross revenue to contractors as of last year, its net revenue for the first half of 2026 could have been approximately $180 million to $250 million. Mercor swung to a small operating profit last year and has told investors it expects operating profit to reach $226 million in 2026 and $1.7 billion in 2027.
Beyond human-generated data: Despite the increasing spend on expert-generated datasets, Nvidia relies heavily on synthetic, or AI-generated, data to train Nemotron. Nvidia also receives useful feedback when developers access Nemotron models through OpenRouter, a service that provides access to models from multiple providers, according to a person who worked on Nemotron.
Nvidia's investment push: The proposed Mercor investment would fit a broader acceleration in Nvidia's private-market activity. In the quarter ending in April 2026, Nvidia invested $18.6 billion in private companies and infrastructure funds, more than it invested throughout 2025. This year it has written separate $2 billion cheques to cloud providers CoreWeave and Nebius, as well as to Lumentum and Coherent, which make lasers used in its networking hardware. Nvidia has also backed smaller startups that buy or rent its chips.
The money behind Mercor: The three-year-old company has raised approximately $520 million from investors including Benchmark and Felicis and was valued at $10 billion in October 2025. A completed round at $20 billion would therefore represent a substantial increase in valuation, but the discussions may not result in a transaction.
Read more: The Information
Image credit: Mercor
