Nvidia discusses backing Mercor in a General Catalyst-led round that would value the three-year-old data provider at $20 billion, reflecting surging demand for expert-curated AI training data.
Nvidia is in discussions to back Mercor, a three-year-old data-labeling startup, at a $20 billion valuation. The deal would double the company's worth from its October round.
Venture firm General Catalyst, an existing Mercor investor, is in talks to lead the financing, according to a report from The Information cited by Yahoo Finance. The precise size of Nvidia's potential contribution remains undetermined.
The chipmaker paid Mercor tens of millions of dollars last quarter alone for high-quality data curated by domain experts across legal, financial, and scientific fields. That data fed directly into Nvidia's two most recent Nemotron open-model releases, Yahoo Finance reported.
A dedicated group of Mercor employees now focuses almost entirely on Nvidia operations, reflecting how quickly the relationship has scaled. Mercor historically derived most of its revenue from closed-source developers such as OpenAI, Google, and Anthropic.
The potential investment aligns with Nvidia's aggressive capital deployment, which reached $18.6 billion in the quarter ending in April. The company has been building a moat around its open-source Nemotron family, designed to rival top open-weight systems globally, according to Yahoo Finance.
The data supply chain
Expert-curated data has become a critical differentiator as model builders exhaust public web corpora. Mercor's specialization in recruiting professionals — lawyers, physicists, financial analysts — to generate supervised fine-tuning examples addresses a bottleneck that synthetic data alone cannot solve.
Nvidia's willingness to pay premium rates for this labor-intensive pipeline signals that the next performance gains will come from data quality, not just compute scale. The company's open-model strategy also creates a flywheel: better Nemotron models attract more developers to Nvidia hardware, Yahoo Finance noted.
The valuation jump from $10 billion to $20 billion in less than a year reflects investor conviction that data-labeling platforms will capture outsized value in the AI stack. But the concentration of Mercor's revenue around a single hyperscaler introduces risk if Nvidia shifts strategy or builds internal labeling capacity, Yahoo Finance indicated.
Nvidia shares eased 1 percent at Wednesday's close. The question now is whether this partnership deepens into an acquisition or remains a strategic minority stake — and whether Mercor can diversify before its biggest customer decides it no longer needs a middleman.
What is Mercor and why does Nvidia want it? Mercor is a data-labeling platform that recruits domain experts to create supervised training examples for AI models. Nvidia needs this high-quality human data to improve its open-source Nemotron models.
How much has Nvidia paid Mercor recently? Tens of millions of dollars in the last quarter alone, according to company disclosures cited by Yahoo Finance.
What is the Nemotron model family? Nemotron is Nvidia's series of open-weight large language models designed to compete with the best open-source systems while driving demand for Nvidia GPUs.
Who else invests in Mercor? General Catalyst is an existing backer and is in talks to lead the new round. Other prior investors have not been disclosed in recent reports.
