Mercor pays over $1.5 million a day to humans training AI, says its CEO (www.businessinsider.com)

Mercor — a startup that recruits teams of human experts to train AI models — says it is paying contractors more than $1.5 million per day and has grown to over 30,000 contractors after a funding round that valued the company at $10 billion. CEO Brendan Foody told media the firm works with major AI labs including OpenAI and Anthropic, hires specialists from software engineering, finance, law and medicine, and frames “humans training AI” as a nascent category of work. Foody also signaled a possible IPO on the horizon. The announcement underscores how large-scale human-in-the-loop workflows (labeling, fine-tuning, RLHF-style supervision and domain tutoring) remain central to building performant models — and are financially lucrative. Mercor’s spending and client list highlight sustained demand for curated human judgment, taste and nuance that current models still need. The market dynamics mirror a broader gold rush: startups connecting skilled contractors to AI labs are fetching lofty valuations and creating new labor markets where some trainers can earn high hourly rates (reports cite up to ~$100/hr), even as workers report monotony and emotional strain. For the AI/ML community this reinforces that scaling models still depends on massive, organized human annotation and domain expertise, shaping where investment and operational focus will go next.