Nvidia Eyes Mercor as AI Data Becomes the New Gold

Nvidia has reportedly discussed investing in AI data supplier Mercor at a valuation of around $20 billion, highlighting how the AI race is expanding beyond GPUs and data centres into the specialized human expertise and training data required to build increasingly capable models. The discussions remain just that—an investment has not yet been confirmed. 

Mercor has become an important supplier to the AI industry by recruiting specialists—including scientists, programmers, lawyers, bankers and PhDs—to create and evaluate sophisticated training data. This expertise is increasingly important as frontier models move beyond basic pre-training toward reasoning, reinforcement learning, agentic AI and domain-specific intelligence. 

Mercor’s Extraordinary Rise

The potential valuation represents remarkable growth. Mercor raised $350 million at a $10 billion valuation in 2025 and has since been discussing approximately $500 million in new funding at twice that valuation. The company said its annualized gross revenue run rate exceeded $2 billion in June 2026, doubling in roughly four months. 

Mercor is also moving beyond traditional data labeling. Its acquisition of Deeptune, which develops simulation environments where AI agents can learn to perform real-world tasks, demonstrates an expansion toward the infrastructure required to train and evaluate autonomous agents. 

Why Nvidia Could Want Mercor

For Nvidia, the strategic logic goes beyond financial investment. The company already dominates the compute layer of the AI economy through GPUs. Supporting companies such as Mercor potentially gives Nvidia greater exposure to another increasingly valuable layer: the data and post-training ecosystem that makes models more capable.

Sources says,Nvidia already supplies Mercor with business: The Information reports that Mercor helps Nvidia develop its open-source AI models. An investment could therefore deepen an existing supplier relationship. 

Analysist Predicts AI’s Bottleneck Is Moving

The significance of the deal is that AI's scarcity is changing. Initially, GPUs were the major constraint. As compute infrastructure expands and foundational models mature, high-quality domain expertise, post-training data, evaluation and environments for teaching AI agents how to perform complex tasks are becoming increasingly valuable.

There is also risk behind Mercor's extraordinary valuation. Its business remains heavily dependent on major foundation-model developers; reporting indicates roughly 91% of first-half 2026 gross revenue came from foundation-model makers. Customer concentration could become a vulnerability if large AI labs internalize more training work or shift suppliers. 

For Nvidia, however, the broader strategy is compelling: owning the GPU opportunity is no longer enough—the company increasingly benefits when the entire AI ecosystem expands.

The next AI gold rush may not be about who owns the most data, but who can turn expert human knowledge into intelligence that machines can learn from.