Micro1 recruits and vets subject-matter experts who create training data and judge AI-system outputs for model developers and large companies. Forbes reports that the San Francisco startup has raised more than $100 million at a $4 billion valuation, citing two people familiar with the deal. Micro1 declined to comment, so the financing remains reported rather than company-confirmed.
The workflow begins with the recruiting software Micro1 built before its pivot. Its AI interviewer screens engineers, doctors, lawyers and other specialists, then the company assigns those experts to projects where they write difficult examples, grade model responses and supply feedback used to improve AI systems. Customers are buying organized human judgment as a managed input to model development.
Micro1 is also packaging that work into simulated environments where agents can practice real-world tasks. Its Realm product generates data from those exercises, while Cortex helps teams evaluate how agents perform in production. For an AI agent, the exercise is closer to rehearsing a job inside a controlled replica than studying another static collection of text.
Reported growth is large, but the accounting label matters. TechCrunch reported in August that Micro1’s gross annual run rate had climbed from $100 million to $500 million in eight months, citing a person familiar with the company. It estimated a net annual run rate of $150 million to $200 million after expert payouts, making the gross figure a measure of business flowing through the platform rather than revenue Micro1 keeps.
Micro1’s previous financing was a company-confirmed $35 million Series A at a $500 million valuation in September 2025. The new reported valuation is eight times that mark in roughly a year. Forbes says Microsoft, Amazon, robotics company 1X and frontier AI labs are customers; people familiar with the deal also said several customers and two xAI cofounders participated in the new round.
The business began shifting away from conventional recruiting after a data-labeling company asked Micro1 to find hundreds of engineers. It now applies the same sourcing system to robotics: TechCrunch reported that hundreds of people record everyday object interactions at home, producing demonstrations that robotics developers can use to train their systems. The underlying asset is a recruiting and oversight system for producing new, specialized data—not one fixed dataset.