If accurate, a $2 billion gross annualized revenue figure would put Mercor in rarefied territory among data-labor vendors, a category that barely existed as a venture asset class three years ago. The company built its business matching contractors — often domain experts — to AI labs hungry for reinforcement-learning data, evaluation work, and bespoke datasets that off-the-shelf scraping can’t produce.
Human-in-the-loop data has quietly become one of the fattest line items in any frontier lab’s budget, and Mercor’s numbers are the receipt.
The Information’s report is thin on specifics — no breakdown of margins, customer concentration, or how ‘gross’ revenue nets out contractor pay, which for marketplace models like this can be the difference between a real business and a pass-through with a fee attached. While topline revenue seems impressive, the devil is in the details – and may not be revealed until Mercor or its competitors go public. For example, what is retention among the big-lab customers? Is pricing holding as labs build in-house data ops teams to cut out the middleman.
Still, the headline number is a signal flare for the sector: as AI labs compete on post-training quality rather than raw pretraining scale, the money is following the humans who supply judgment, not just text. Watch for competitors like Scale AI, Handshake, Micro1 and Surge to point to their own run-rate figures in response, and for more scrutiny of how ‘gross annualized revenue’ gets defined across this crowded field.
Exclusive: Mercor Hit Over $2 Billion in Gross Annualized Revenue
