Micro1 plans $1 billion spend on workplace data for AI training

Micro1 is acquiring internal company data such as Slack DMs and Google Docs to license to AI model developers.

Micro1 targets workplace data for AI model training. The San Francisco startup buys entire data corpora from midsize companies and packages the material for labs including OpenAI and Anthropic. It pays between $100,000 and $2 million per dataset, with larger and more complex collections commanding higher prices.

The company announced plans to spend $1 billion on data acquisitions over the next year. Chief revenue officer William Almeida told Business Insider that the firm explains to prospective sellers why their records now hold value for building simulations that improve AI performance.

Three licensing models are offered to AI developers. Micro1 can anonymize and license raw data directly. It can also supply an in-progress AI agent for customer feedback. The third option freezes a company's data at a chosen timestamp and provides the resulting environment for model training.

CEO Ali Ansari leads the effort. The approach aligns with a shift in AI data sourcing, moving from public internet text to internal workplace interactions that capture employee workflows.

Executives report receiving substantial offers. Ilya Brotzky, CEO of recruiting startup VanHack, told Business Insider he was surprised by the size of Micro1's proposal after completing a data survey. He ultimately selected Micro1 over a competing bid from data startup Mercor after reviewing its privacy protections.

Industry spending on AI data is projected to grow rapidly. Gartner estimated that annual outlays will rise from less than $1 billion last year to nearly $6.5 billion in 2027. Micro1 focuses on datasets that preserve connections, diversity, and traceability across multiple tools and file types.