The era of low-cost mechanical turks in AI training has definitively closed. Four-year-old data vendor Micro1 scaled its gross annual run rate from $100 million to $500 million in just eight months, according to a person familiar with the company's financials. By contracting credentialed professionals—including physicians, corporate attorneys, and research scientists—to evaluate reasoning traces and fine-tune frontier models, the platform captures a massive 60% to 70% take-rate. That dynamic translates to a net annualized revenue run rate of $150 million to $200 million.
Frontier AI labs are starving for verifiable domain knowledge, transforming expert data annotation and specialized validation into one of the most lucrative bottlenecks across the AI supply chain. Alongside custom reinforcement learning environments, Micro1 sells off-the-shelf pre-training packages and automated synthetic video descriptions, driving gross margins on packaged datasets up to 80% to 90%, as TechCrunch reported.
Even as competitors scale aggressively—Handshake hitting a $1 billion gross run rate and Mercor crossing $2 billion annualized—the high-end training supply chain is splintering along geopolitical lines. Founder Ali Ansari publicly stated on X that Micro1 refuses data sales to Chinese model builders, openly distancing the firm from rivals whose datasets reportedly trained architectures like Kimi K3.