A profile of Mercor, which pays ~$2M per day to ~30K experts training AI models at $95 per hour on average, with roles like radiologists earning up to $375/hour
Tens of thousands of professionals are joining tech start-up Mercor to hone technology in their skills — At first Lola thought the job ads were fake.
Context & Ripple Effects
Mercor has moved from an AI-vetted jobs marketplace into a specialist training-work intermediary: its earlier AI-based candidate-screening marketplace created the supply-side machinery, while later coverage described a business hiring domain experts to train models.
The reported daily payouts put tangible scale behind Mercor’s claimed “new category of work,” following its reported $500M ARR milestone and contractor payouts in 2025. The significance is less the marketplace’s headcount than the price being placed on scarce, high-consequence expertise.
First-order effects
- Mercor is directing roughly $2M a day to about 30,000 contributors, giving professionals an immediate paid channel for applying domain knowledge to model training.
- Specialist workers, particularly in fields such as radiology, can command far higher rates than the reported average, making expert review and training a meaningful source of flexible professional income.
Second-order effects
- Other AI-data and contractor marketplaces face pressure to recruit, vet, and retain qualified specialists at comparable pay and with credible project flow.
- For model developers, the cost of improving performance in narrow professional domains becomes more visibly tied to expert labor, not only to compute or generic data.
Third-order effects
- If sustained, specialist AI training marketplaces could become a durable labor layer between knowledge professionals and model builders, with screening and access to expert supply becoming a strategic asset.
- The pattern shifts attention toward the economics of a useful AI task: as generic capabilities improve, differentiated model quality may increasingly depend on costly, domain-specific human input.
The trend: AI development is evolving from broad data collection toward paid, structured access to scarce professional judgment for specialized model improvement.