Micro1 has become another marker of how quickly AI training data has turned into core infrastructure for model builders. TechCrunch reports that the four-year-old startup expanded from $100 million to $500 million in gross annual run rate over the past eight months, citing a person familiar with the company.

The company works in the same broad category as data-labeling and expert-network businesses that supply model labs with specialized human feedback, evaluations, and training examples. Micro1 describes itself as talent and data infrastructure for AGI, saying it partners with AI labs and enterprises to vet human intelligence, assemble expert teams, and turn that work into training data.

The revenue number is gross rather than net. TechCrunch reports that Micro1 retains roughly 60% to 70% of the gross figure, putting net annual run rate between $150 million and $200 million. That distinction matters because these businesses often pay large pools of contractors, including domain experts such as doctors, lawyers, engineers, and scientists.

The broader signal is that the data layer is becoming a budget line next to compute. As frontier labs run into scarcity around high-quality examples and evaluations, vendors that can source, filter, and package domain-specific data are moving from annotation services toward AI supply-chain infrastructure.