Micro1, a four-year-old AI data-labelling startup, has seen its gross annual run rate surge from $100 million to $500 million over the past eight months, reflecting soaring demand from AI labs and corporations for high-quality training data. The company, which contracts domain experts such as doctors, lawyers and scientists, retains roughly 60-70% of that revenue, giving it a net run rate of $150-200 million. It still trails larger rivals Mercor ($2 billion) and Handshake ($1 billion), but its rapid growth underlines a market big enough to sustain multiple data providers as some researchers predict future AI spending on data could rival spending on compute.

Micro1 is increasingly producing synthetic data without human involvement, such as automated video descriptions, and reselling some "off-the-shelf" datasets to multiple clients, which can push gross margins as high as 80-90%. That practice has drawn criticism elsewhere in the industry over concerns that shared datasets help Chinese AI developers close the gap with US models, though founder Ali Ansari has said Micro1 does not sell data to Chinese model makers. Originally an AI recruiting platform, the company pivoted into data labelling after noticing clients using it to vet annotation staff; it raised its Series A at a $500 million valuation last September and may have since closed a new round at a substantially higher valuation.