Mercor raises $750 million at a $20 billion valuation to train and vet SI agents for companies
Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .
Money in SI is shifting from training models to vetting them on real work: human expertise becomes the raw material from which companies build their own vetted agents.
Mercor, a startup that began as a staffing platform and became a supplier of expert data for training models, has announced a $750 million Series D at a $20 billion valuation — double the $10 billion of October 2025. Benchmark, Felicis, Fundomo, Apollo entities, G Squared, General Catalyst, Glade Brook, Nvidia, Salesforce, Sapphire Ventures, StepStone, XTX Ventures and Zoom participated; part of the sum went to buying shares from employees. The money will go toward evaluating and fine-tuning agents for specific companies: Mercor builds custom benchmarks for each organisation based on its real processes and exceptions. In 2026 the company acquired Deeptune, whose team recreated hundreds of enterprise applications as environments for training agents. Co-CEO Brendan Foody writes that Mercor already spends three times more on inference than on employee salaries. According to Crunchbase, SI startups received $102 billion in the third quarter of 2026 — 64% of global venture capital.