Micro1, a four-year-old startup that supplies AI training data, has grown its gross annual run rate from $100 million to $500 million in just eight months, according to a person familiar with the company. The surge reflects a near-bottomless appetite among top AI labs and large enterprises for unique, human-generated training data.
Like its peers, Micro1 hires domain experts — doctors, lawyers, and scientists — on a contract basis to produce high-quality datasets. The company retains between 60% and 70% of that gross figure, putting its net annual run rate between $150 million and $200 million.
Micro1 still trails its larger rivals: Mercor reached $2 billion in gross annualized revenue this summer, while Handshake crossed $1 billion earlier this year. Yet the company's trajectory suggests there is more than enough demand to support multiple suppliers of human expertise for model training.
The boom may have room to run. Some researchers hypothesize that future AI spending on data could rival spending on compute. Micro1 is also leaning into synthetic data — automatically generating descriptions of video content, for example — and "off-the-shelf" datasets it can sell to multiple customers carry gross margins as high as 80% to 90%, the person familiar said. Contract sizes, meanwhile, keep growing.
