Nvidia plans to invest in AI data supplier Mercor, which is valued at $20 billion

Nvidia is seeking to acquire a stake in AI data annotation company Mercor, a move that would double the valuation of this three-year-old startup to $20 billion. This reflects the chip giant's accelerated push into open-source AI models while extending its investment reach upstream in the data supply chain.

According to a recent report from technology outlet The Information, insiders reveal that Nvidia has entered discussions to participate in Mercor's new financing round, with existing investor General Catalyst negotiating to lead the round. Last quarter, Nvidia paid Mercor tens of millions of dollars in data services fees, and the business relationship between the two is deepening rapidly.

The potential investment comes against the backdrop of Nvidia's overall acceleration in investment activity. In the quarter ending April this year, Nvidia invested $18.6 billion in private companies and infrastructure funds, surpassing the total annual investment of 2025. If Mercor successfully completes this round, it would mark Nvidia’s first significant bet in the data services sector.

Nvidia and Mercor: From Customer to Potential Shareholder

The relationship between Nvidia and Mercor is evolving from a simple procurement contract to equity binding. According to reports, insiders stated that Nvidia has used Mercor's data to train its two newest Nemotron open-source models, with several Mercor employees dedicating nearly all their efforts to serving Nvidia.

Mercor’s core business is recruiting human experts in various fields to use and evaluate AI models on professional tasks, with the resulting data provided to model developers to enhance model capabilities. These tasks span scientific research, legal, financial projects, and even programming work for Nvidia GPUs.

The Nemotron series is central to Nvidia's open-source model strategy, aiming to compete with the world's leading open-source models. As Nvidia increases its investment in Nemotron, its payments to Mercor have also surged, making Mercor’s revenue from Nvidia increasingly significant.

It is worth noting that Nvidia is not Mercor’s only data supplier. The report notes, citing insiders, that Nvidia is also using services of other data providers such as Turing and Scale, and it maintains an internal data team.

Additionally, Nvidia gathers valuable data generated by developer access to Nemotron models through the OpenRouter platform. A person involved in the Nemotron project commented that like many AI model developers, Nvidia relies heavily on synthetic data (i.e., AI-generated data) to train its models.

Mercor: A High-Growth Data Annotation Unicorn

Mercor’s financial figures underscore the booming market for AI data annotation. According to The Information, Mercor achieved gross revenue of $614 million in the first half of this year, a 70% increase over the whole of last year, and is expected to reach $2 billion in gross revenue for the full year.

In terms of profitability, Mercor turned a small positive operating profit last year and has informed investors that operating profit will rise to $226 million this year and further leap to $1.7 billion next year. Since the company pays about 60% to 70% of its gross revenue to contractors, its estimated net income for the first half of this year is between $180 million and $250 million.

Mercor was founded just three years ago and has previously raised around $520 million from investors such as Benchmark and Felicis. As of last October, it was valued at $10 billion. Should this round close at a $20 billion valuation, its value will have doubled in less than a year.

Mercor’s main clients have traditionally been closed-source AI model developers, including OpenAI, Google, and Anthropic. The rapid growth of Nvidia’s business marks a shift in its client structure, expanding towards the open-source model ecosystem.

Nvidia’s Investment Landscape: From Chip Clients to Ecosystem Deployment

Nvidia’s recent investment strategy shows a clear trend toward scale and acceleration.

This year, Nvidia has each invested $2 billion in cloud service providers CoreWeave and Nebius, both major customers of Nvidia chips. Additionally, Nvidia has injected capital into its network hardware laser suppliers Lumentum and Coherent, and provided funding to several startups that rent or purchase Nvidia chips.

Nvidia’s investment logic typically revolves around its direct customers or the largest users of its chips. The potential investment in Mercor is somewhat different—Mercor is not a chip buyer, but rather an upstream data supplier to Nvidia’s AI model development capabilities. This layout reflects Nvidia’s strategic consideration in strengthening its competitive edge in open-source AI.

For reference, Nvidia participated in Scale AI’s financing round in 2024, when Scale was valued at $14 billion. Subsequently, Meta took a 49% stake in Scale and appointed its CEO. If Mercor completes this round, its $20 billion valuation will surpass Scale’s, making it the highest-valued company in the AI data annotation field.