Nvidia weighs $20 billion Mercor investment as open-source AI push grows

Nvidia is in talks to invest in Mercor, its AI data supplier, at a $20 billion valuation, deepening the chipmaker's push into open-source model development as it races to challenge DeepSeek and other open-source leaders.

The Information's Phoebe Liu and Julia Hornstein reported that Nvidia is considering funding Mercor at a $20 billion valuation, with venture capital firm General Catalyst in discussions to lead the round.

Mercor's revenue historically came primarily from closed-source AI model makers such as OpenAI, Google and Anthropic. But Nvidia-related revenue is growing as the chipmaker focuses on its Nemotron open-source large language models, built to compete with the world's top open-source systems. Mercor crossed $2 billion in annualized revenue in June, after hitting $1 billion in February, according to Forbes.

The deal would mark Nvidia's expansion beyond chipmaking into the AI model development value chain, where training data has become a strategic bottleneck. It also highlights the rising value of AI data providers as model makers race to secure high-quality training data.

Nemotron and the open-source race

The investment comes as Nvidia, the dominant supplier of AI accelerators, builds out its own open-source model lineup. Nemotron is Nvidia's answer to open-source rivals such as DeepSeek, whose R1 model showed that competitive AI can be built at a fraction of the cost of closed-source frontier systems. The Trump administration has also pushed for an American open-source alternative, backing startups like Reflection, a Brooklyn-based venture valued at $8 billion that counts Nvidia, Sequoia and Lightspeed among its backers.

For Mercor, the funding would cement its position as a critical supplier to the AI industry's biggest names. The company's customer roster already includes OpenAI, Google and Anthropic, and its revenue has more than doubled in the first half of the year. The startup also counts Neuralink among its customers, according to Forbes. Mercor's growth reflects the voracious appetite of frontier labs for labeled, high-quality data as they train ever-larger models.

Data as strategic infrastructure

The deal highlights the growing strategic importance of AI training data providers. As frontier models consume ever-larger datasets, the companies that source, label and curate that data have become essential infrastructure. A $20 billion valuation would place Mercor among the most valuable AI data companies, reflecting investor confidence in the sector's growth.

The funding also fits a broader pattern of chipmakers and cloud providers investing up and down the AI stack. Nvidia has backed a range of AI startups, from model developers to infrastructure providers, as it seeks to lock in demand for its GPUs. Mercor, which supplies the data that trains those models, sits at the center of that strategy.

For investors, the deal points to Nvidia's intent to control more of the AI value chain, from chips to models to the data that trains them. Nvidia shares, trading at roughly 35 times forward earnings, have been volatile as investors weigh the company's data center growth against rising competition from custom silicon developed by Google, Amazon and Meta. The Mercor investment, if completed, would give Nvidia a direct stake in the data layer that underpins the entire AI industry.

This article is for informational purposes only and does not constitute investment advice.