Mercor Captures High-Skill AI Labor

Mercor’s growth shows that AI labs have moved from buying cheap generic labeling to buying expensive expert judgment as a core training input. When models are being tuned on medicine, law, finance, coding, and reasoning, labs need people who can catch subtle errors, rank outputs, and explain why one answer is better. That pushes spend per worker far higher, and a platform that can quickly vet and deploy those experts can scale revenue extremely fast.

• The shift in AI training demand is from low wage commodity tasks to high skill RLHF and post training work. Mercor built around doctors, lawyers, PhDs, and engineers paid at much higher hourly rates, while earlier data labeling markets were built on much cheaper generalist labor.
  2 sacra 3 sacra
• Mercor sits between broad freelance marketplaces and full service data vendors. Like Upwork, it uses software to source, vet, match, contract, and pay remote workers at global scale. But like Turing or a staffing firm, it wins by tighter curation for higher value hires where buyers care more about quality than raw marketplace breadth.
  4 techcrunch 5 sacra
• The comparable set explains the revenue acceleration. Mercor reached roughly $50M run rate by the end of 2024 and $500M annualized revenue by August 2025, while Invisible and Scale also grew quickly as labs spent more on reasoning data and human feedback. This is a category wave, not just one company taking share.
  1 sacra 2 sacra 3 sacra

The next phase is a race to own the human layer of advanced model training. As labs build more reinforcement learning environments and need auditable expert feedback, Mercor can move from matching workers into becoming workflow infrastructure for training, evaluation, and benchmarking. That would make it less like a recruiter and more like a core supplier to frontier model development.