The operative question isn’t why Nvidia wants access to Mercor’s data — it already pays for that, tens of millions of dollars a quarter, according to The Information’s reporting. It’s why Nvidia wants to own a piece of the company doing the labeling rather than just keep writing checks. The answer is supply-chain control: Nemotron is Nvidia’s bet on open-source relevance against Meta and Alibaba, and expert-labeled data for legal, financial and scientific tasks is the scarce input that differentiates a frontier open model from a commodity one. An equity stake buys priority access and alignment at a moment when Mercor’s other clients — OpenAI, Google, Anthropic — are drawing on the same finite pool of domain experts.
Nvidia already ran this playbook with Scale AI, and it didn’t end cleanly — Meta’s later stake in Scale spooked rival labs into diversifying away, which is exactly the dynamic a Mercor investment risks recreating.
Nvidia’s participation in Scale AI’s 2024 round, which valued that company at $14 billion, is the direct precedent, and it’s a cautionary one: once Meta took a larger, controlling-adjacent position in Scale in 2025, several major labs reportedly pulled volume rather than hand a competitor’s shareholder visibility into their training pipelines. A Nvidia stake in Mercor — even a minority one, terms undisclosed — puts OpenAI, Google and Anthropic on notice that their labeling vendor now has a board-adjacent relationship with a chipmaker they all depend on for compute. Whether that erodes Mercor’s non-Nvidia revenue is the first thing to watch.
The valuation math also deserves scrutiny. Mercor’s annualized gross revenue reportedly hit $2 billion in June 2026, per figures cited by finance.biggo.com from The Information, up from roughly $614 million in gross revenue over the first half of the year. But Mercor reportedly pays contractors 60% to 70% of that gross figure, which puts estimated net revenue for H1 closer to $180 million to $250 million — a very different number to underwrite a $20 billion valuation than the headline $2 billion run rate suggests. A $20 billion price tag on a three-year-old labeling shop, doubling in ten months, is a bet that the take-rate business scales faster than the margin math implies.
Watch three things next: whether Nvidia’s stake size and board rights become public, whether OpenAI or Anthropic diversify their labeling spend in response, and how Mercor’s pending data-breach class actions — which allege exposure of contractor biometric data and interview recordings, per PYMNTS — factor into diligence on a company now handling sensitive work for the world’s largest chipmaker.
In the previous quarter alone, Nvidia paid Mercor tens of millions of dollars for high-quality data curated by domain experts across legal, financial, and scientific fields. According to company disclosures, Mercor's data was integrated into Nvidia's two most recent Nemotron releases.
