A company most people have never heard of just posted revenue numbers that would make some publicly traded firms jealous. Mercor, a San Francisco-based AI recruiting and data platform founded just three years ago, pulled in $614 million in gross revenue during the first half of 2026, with more than 90% of that coming from AI foundation model companies.
That’s not a typo. A three-year-old startup is generating revenue at a pace that implies a $2 billion annualized run rate, according to internal documents referenced by The Information. Mercor crossed the $1 billion annualized threshold just four months before hitting $2 billion.
The AI training supply chain, explained
Here’s what Mercor actually does. Think of it as a massive matchmaking service, but instead of dates, it connects domain experts and contractors with AI labs that need human brainpower to train their models.
The work includes data labeling, model evaluation, and reinforcement learning from human feedback (RLHF). Mercor’s client list reads like a who’s who of the AI arms race. OpenAI, Anthropic, and Google are among the foundation model developers fueling that 90%-plus revenue concentration. The company compensates roughly 30,000 contractors at an average rate of $105 per hour, with daily payouts exceeding $2 million.
