The Largest AI Data Labs and Human Data Companies in 2026, by Revenue on Record

Originally published at aidata.broker, where the full ranking is kept current.

By the revenue on record, the largest AI data labs are Mercor at over $2B (run rate, Jun 2026), Surge AI at $1.2B (FY2024) and Handshake AI at over $1B (run rate, Aug 2026). They pay experts and contractors to write the tasks, demonstrations, rubrics and judgments AI labs train and test their models on. AIData.Broker ranks every company that says on its own pages that it sells this work, 103 of them, by the money on record, and links every figure to the filing or article that states it.

The ten largest

1. Mercor: over $2B (run rate, Jun 2026)
2. Surge AI: $1.2B (FY2024)
3. Handshake AI: over $1B (run rate, Aug 2026)
4. micro1: $1B (run rate, Oct 2026)
5. Scale AI: about $1B (FY2025)
6. Snorkel AI: $375M (run rate, Sep 2026)
7. Innodata: $350M (12 months to Sep 2026)
8. Turing: about $300M (run rate, Mar 2025)
9. TaskUs: over $250M (12 months to Sep 2026, AI Services)
10. Appen: $249M (12 months to Jun 2026)

The other 31 ranked companies, and the 62 with no money on record yet, are on the full list.

Who is growing fastest

Where a company published two revenue figures about a year apart, the change between them, never extrapolated to a year:

• Snorkel AI: from about $20M (run rate, Sep 2025) to $375M (run rate, Sep 2026), up about 19 times in 12 months
• micro1: from over $60M (run rate, Sep 2025) to $1B (run rate, Oct 2026), up to 17 times in 13 months
• Mercor: from $500M (run rate, Sep 2025) to over $2B (run rate, Jun 2026), up more than 4 times in 9 months
• Turing: from $100M (run rate, Dec 2023) to about $300M (run rate, Mar 2025), up about 3 times in 15 months

Every company with two such figures is on the growth ranking, and what investors pay for that revenue is on the valuations page.

How the ranking works

Each company is ranked on the first of these it has on record: its revenue over twelve months (a run rate or a year as published, or its last four quarters added up; an AI segment where the rest of its business is not AI data), then its newest valuation, then the money it disclosed raising. No company pays for a place.

A projection, a target or guidance is never revenue, a valuation a company is only in talks to raise at is not used, and every claim of what a company sells is a quote from its own site, checked word for word. The method has every rule and each change to it.

The data

The ranking is open data under CC BY 4.0: CSV and JSON, with the ids of the records each figure comes from. The same rule ranks RL environment companies, data labeling and annotation companies and 15 other lists, all on one page.