Micro1 Hits $1B Annualized Revenue

Micro1’s jump to a $1B annualized run rate shows that expert human labor has become core AI infrastructure, not a side service. The company went from $7M at the end of 2024 to $125M at the end of 2025, then to $1B by October 2026, which means a very small vendor became a major conduit for frontier lab training spend in less than two years. In practice, Micro1 is selling fast access to vetted specialists who grade outputs, compare answers, and generate high quality RLHF data for labs racing to improve reasoning models.1 2 3 4

• This is now a crowded but enormous market. Handshake was estimated at $1.1B annualized gross revenue in April 2026, Surge at $1.2B in 2024 revenue, Scale AI at $1.5B in 2024 revenue, Turing at $300M in 2024, and Prolific at about $350M annualized in April 2026. Micro1 moved from emerging player to peer set member very quickly.2 3 5 6 7
• The workflow is concrete. Labs pay platforms to find credentialed workers, route prompts and model outputs to them, collect rankings and written feedback, run QA, and handle payouts. Invisible’s model shows the underlying economics, with labs billed hourly for raters while the platform keeps the spread and operational fee, which helps explain why gross revenue can scale so fast once a few big lab contracts land.4
• The strategic constraint is concentration. Micro1’s own company profile flags heavy dependence on a small number of well funded AI labs. That fits the broader market pattern, where a handful of foundation labs and hyperscalers drive most spend, vendor switching is relatively easy, and sudden budget shifts can rapidly reallocate hundreds of millions of run rate across suppliers.8 2

The next phase is a shift from labor marketplaces into deeper data infrastructure. The winners will bundle expert supply with evaluation tools, RL environments, and workflow software, so labs buy an integrated training stack instead of just contractor hours. If Micro1 keeps compounding, its path is toward becoming part marketplace, part operating system for human feedback in frontier model development.2 4