Updated 8 Nov 2026

How the rankings are made.

179 companies on the register, 78 ranked on the money on record, in 18 lists. Every figure links to the filing or article it comes from, and every rule that decides a place is below, with each change we have made to it.

The method

What is counted, from where, and what is left out.

Who is on the register.
179 companies that sell AI developers training data, expert work, labeling, RL environments or evaluations. Each is there because its own pages say so: every claim of what it sells, what it works on and where its data comes from is a quote from its own site, fetched and checked word for word by script before it is recorded, with the day it was read. A claim whose quote is not on the page is left out.
What it is ranked on.
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.
Revenue.
What the company states (its filings, results, announcements) or what a named outlet reports it earns, each quote checked on the page it comes from. Twelve months means a run rate or ARR as published, a year as published, twelve months to a date as stated, or the last four quarters or two halves added up. A projection, a target or guidance is never revenue. Where a company's AI data business is a segment of a larger one, only that segment is read, under the name its own reports give it.
Valuation.
The newest valuation of a completed round or investment on record. A valuation a company is only in talks to raise at is not used.
Money raised.
Disclosed funding rounds and direct investments in the company. A vehicle that invests in a round already counted (an SPV), a share sale between existing holders and a loan are not money raised.
Sources we refuse.
Sacra, Latka, Crunchbase, PitchBook, Tracxn, CB Insights, Dealroom, Growjo, Owler, ZoomInfo, LinkedIn and Reddit posts. A figure on these sites is an estimate, a copy of someone else's report, or a post anyone could write: never the record itself.
Who pays.
No company pays for its place. Some companies take companies' data through AIData.Broker's introductions and pay us a referral fee when one goes ahead; the tables say so beside them, and it never moves a place: the order is the money on record.
Competitors and head to heads.
A company's competitors are the companies whose own pages say they sell at least two of the things it sells, in the order of the same rule (78 pages). A head to head is a pair Google itself suggests comparing, read on 8 Nov 2026 (37 pairs); a revenue page exists for each company ranked on revenue (19).

Changes to the rule

Each change to how the records are read, newest first, with what it moved.

  • 8 Nov 2026A company whose main business is something else (security testing, customer experience, compute) is ranked only on the revenue of its AI data segment; without such a figure it is listed and not ranked, since its valuation and money raised are the whole business's. TELUS Digital, Prime Intellect, Welocalize, V7 and Bugcrowd are listed without a place; their figures stay on their pages.
  • 8 Nov 2026Revenue stated for twelve months to a date ("trailing twelve months", "12 months to 30 September") is read as twelve months, as a run rate or a year already was. TaskUs is ranked on its AI Services revenue for the twelve months to September 2026 ($250M) instead of its 2025 figure.
  • 8 Nov 2026Money raised counts direct investments in a company (a strategic stake, a SAFE) as well as funding rounds, and never a vehicle that invests in a round already counted (an SPV). Scale AI's disclosed funding includes Meta's 2025 investment.
  • 8 Nov 2026A valuation a company is only in talks to raise at is not its valuation; the newest completed one is. Surge AI shows no valuation (its reported talks never closed on record); Mercor shows its $10B round of October 2025.
  • 8 Nov 2026Where a company's AI data business is a segment of a larger one, its entry names that segment and only that segment's revenue is read; every other company is read whole. Before, any segment with "AI" in its name was preferred. Appen is ranked on its whole revenue (a small slice it reports had hidden it); TaskUs and Handshake keep their AI segments.

A figure you think is wrong or out of date: write to hello@aidata.broker with the page that states the right one. We check it against that page and correct the record; a newer figure from the company itself replaces an older one.

The data

The whole ranking as open data, under the same terms as every table on this site.

We ask once, before your first download, so we know who uses the data. The files stay free to reuse under CC BY 4.0. Downloading means you accept our terms and privacy policy.

Your is downloading.

The companies that build AI training data, ranked by what is on record.

Every source behind the figures: the source list