A rising startup, Mercor, founded in 2023 by three college dropouts, is reportedly in the final stages of securing a funding round that values it at about $10 billion. Based in the Bay Area, Mercor has rapidly become a linchpin in the generative‑AI ecosystem by recruiting thousands of domain experts—doctors, lawyers, journalists—to train and evaluate chatbot outputs, including for major clients like OpenAI and Anthropic.
From its inception as a recruiting startup, Mercor pivoted quietly but decisively into supplying human feedback loops for AI models—a behind‑the‑scenes function that is increasingly recognized as mission‑critical. Its surge reflects a broader trend: as generative‑AI capabilities escalate, so too does the demand for high‑quality human‑in‑the‑loop systems that can refine, evaluate and “teach” the machines.
Founded by CEO Brendan Foody, CTO Adarsh Hiremath and Chairman‑turning‑COO Surya Midha—all of whom dropped out of college and became Thiel Fellows—the company leveraged its recruiting origins into a roster of more than 30,000 contractors worldwide who label data, review model‑answers and help train chatbots to “think and speak like humans.”
Mercor’s business model charges the AI firms for expert human feedback—engineers, doctors, lawyers—and in turn pays the contractors hourly rates (for example, up to ~$170 /hour for medical professionals) while retaining a service spread of roughly 30–35 %. The company’s revenue run‑rate is estimated at around $450 million annually, putting the valuation target at more than 20× revenue—underscoring the faith investors are placing in the “human feedback” layer of AI.
But the rapid rise comes amid competition and controversy. Rival firm Scale AI—itself valued at nearly $29 billion after a major investment by Meta Platforms—recently sued Mercor for alleged trade‑secret theft, highlighting how hot and contested the data‑labelling market has become. In parallel, broader questions are surfacing about whether human‑in‑the‑loop systems can scale ethically, how labor conditions will evolve, and whether such firms will be able to maintain quality and neutrality as stakes grow.
Looking ahead, Mercor’s trajectory invites reflection on both the promise and paradox of the AI revolution. On one hand, it illustrates how the “boring” infrastructure of expert‑feedback is becoming a gold mine in an AI landscape obsessed with shiny model announcements. On the other hand, it underscores a tension: as machines learn faster, the human “teachers” may themselves be commoditized, and business models will need to reconcile explosive growth with fairness, sustainability and regulatory scrutiny. The rise of Mercor signals that the AI arms race is no longer just about algorithms—it’s about the human networks underpinning them.
Cybericonic 