Mercor's three founders were barely old enough to rent a car when the company crossed $2 billion in annualized revenue. Now the real question isn't whether the AI data business is real. It's how big it can get before the market runs out of room.
Brendan Foody said it plainly on Monday: Mercor had crossed $2 billion in gross annualized revenue, according to reporting from The Information. That's double where the company stood just four months earlier, when it hit $1 billion. Eight months before that, in October 2025, Mercor priced itself at $10 billion, a fivefold jump from its $2 billion Series B valuation eight months prior. The math isn't subtle anymore. Mercor's growth curve has gotten steeper, not flatter, at a stage when most startups start to slow down.
The business itself is unglamorous, almost clerical. Mercor pays doctors, lawyers, and engineers by the hour to answer questions, grade responses, and generate the kind of feedback large language models need to get smarter. It's reinforcement learning from human feedback, run at industrial scale. As of its Series C close, Mercor said it was paying out more than $1.5 million a day to a roster of over 30,000 contractors, who earn an average of $85 an hour. OpenAI and Google DeepMind are among the labs paying for that labor, and Mercor picked up much of that business after Meta invested $14 billion in rival Scale AI and hired away its CEO, a deal that reportedly made OpenAI and Google wary of routing training data through a competitor's platform.
Founders Foody, Adarsh Hiremath, and Surya Midha started Mercor in January 2023, while all three were still undergraduates: Foody and Midha at Georgetown, Hiremath at Harvard. In March 2024 they became one of the only full founding teams ever to win the Thiel Fellowship together, taking $100,000 from Peter Thiel's program on the condition that they drop out. They did. Eighteen months later, at 22, all three became the youngest self-made billionaires on record when Mercor's Series C priced the company at $10 billion.
That round, $350 million led by Felicis Ventures with Benchmark, General Catalyst, and Robinhood Ventures joining in, closed in October. It's already starting to look conservative. A company doesn't double its revenue every four months and expect last year's valuation to hold, and the venture market around Mercor has been repricing accordingly. Surge AI, a bootstrapped rival with roughly 110 employees that has never taken outside capital, is reportedly raising at a $15 billion valuation on more than $1 billion in trailing revenue. Scale AI, the company Mercor effectively displaced, is now valued above $29 billion following Meta's investment. Investors are pricing this category like infrastructure, not like a services vendor.
Not everyone is comfortable with how those numbers get set. Foody picked a fight with Sequoia Capital on X on June 8, accusing the firm of what he called "dual-pricing," structuring rounds in two tranches so a small slice of capital goes in at a much higher valuation than the bulk of the check, then letting the company announce only the higher number. "In the last 6 ive seen a half dozen rounds where sequoia invests in 2 tranches," Foody wrote. "everyone pretends they only did the higher valuation." Sequoia partner Shaun Maguire didn't deny it, saying the practice had happened "approximately five times" during his seven years at the firm, and pointed to Serval, an AI help desk startup that announced a $75 million Series B at a $1 billion valuation just days after a Series A extension had priced it under $400 million. Frankly, if a 22-year-old CEO is publicly correcting Sequoia's math, that tells you how much scrutiny now sits on every headline valuation in this market.
That's the tension underneath Mercor's growth story. The revenue is real and verifiable, paid out daily to tens of thousands of contractors doing actual work. The valuations wrapped around that revenue are set in a market where even sophisticated investors are arguing in public about whether the numbers mean what they say. Mercor doesn't need a splashy new funding round to prove its business works. The daily payroll already does that. What happens next depends less on Wall Street's appetite than on whether AI labs keep needing this much human judgment, or whether frontier models eventually need less of it than they do today.
Also read: Ollama Raises $65 Million as It Grows to Nearly 9 Million Developers • What Is MCP, the Model Context Protocol Powering AI Agents Now • Salesforce Keeps Promoting Anthropic's Claude Tag Inside Its Own Slack