Micro1 Reportedly Reaches $500M Gross Run Rate as AI Data Demand Surges

AI training data startup Micro1 has reportedly reached a $500 million gross annual run rate, up from about $100 million eight months ago, as demand for specialised data continues to expand across leading AI labs and corporations.

Micro1 retains roughly 60% to 70% of its gross run rate after payments to contract experts and other costs, according to a person familiar with the company’s finances. That puts its net annual run rate between $150 million and $200 million.

AI training data demand is creating multiple large businesses

Micro1 remains smaller than some of its biggest competitors. Mercor reached roughly $2 billion in gross annualised revenue this summer, while Handshake’s AI training business reached about $1 billion earlier in the year, according to The Information.

The growth reflects rising demand for specialised human expertise used to improve AI models. Companies such as Micro1 recruit contractors, including doctors, lawyers, scientists, and engineers, to evaluate model outputs and generate training material for increasingly complex tasks.

Some researchers argue that data could become an even larger constraint as models improve. Former OpenAI researcher Will DePue has argued that AI development is becoming increasingly data-limited, raising the possibility that future spending on high-quality proprietary data could approach the enormous sums already spent on computing infrastructure.

Micro1 is expanding beyond human-generated data

The startup is also increasing its use of synthetic data, including automated descriptions of video content. Some datasets can be sold to multiple customers rather than created for a single client, with gross margins on this off-the-shelf data reportedly reaching 80% to 90%.

That business model has also attracted scrutiny. Forbes reported that several U.S. data companies have sold training datasets to Chinese AI developers, prompting debate over whether the specialised data helping American labs could also strengthen competing models abroad.

Micro1 founder Ali Ansari has publicly distanced his company from that practice. In a post on X, Ansari said Micro1 does not sell its data to Chinese model developers and criticised companies that supply training material to countries competing with the United States in AI.

Micro1 evolved from AI recruiting into training data

Micro1 originally focused on AI-powered recruiting. Ansari expanded into data labelling after seeing customers use thecompany’ss recruiting technology to identify and vet engineers for AI annotation work.

The company has since broadened its work to include model evaluations and robotics training data. One project involves having participants record everyday interactions with objects in their homes to create pre-training material for robotics systems.

Micro1 raised a $35 million Series A in September 2025 at a $500 million valuation. Its reported jump to a $500 million gross run rate less than a year later underscores how quickly spending on AI training data has grown alongside investment in increasingly capable models.