The AI data market, weekly · 9 Nov 2026
What investors pay for each dollar an AI data company earns
We ranked the AI data companies by the money on record and set each valuation against the revenue before it: Scale AI at 33 times, Mercor at 20. Eminem's and Elvis Presley's rights owners sued Suno, The New York Times sued Brave, and the hearing on Spirit's sale of passenger data to Google moved to 17 November.
Our research this week
- We now rank 91 of the 198 companies that sell AI developers data, expert work, labeling, RL environments or evaluations, by twelve-month revenue, then valuation, then money raised, every figure linked to the filing or article that reported it. Mercor leads on a reported gross run rate of over $2 billion (June 2026), ahead of Surge AI ($1.2 billion in 2024) and Handshake AI (over $1 billion, August 2026). MoreSource: AIData.Broker, the rankings
- Set against the revenue on record in the year before each, the newest valuations put Scale AI at 33 times its revenue, AfterQuery at 32, Arena at 31 and Mercor at 20; micro1's October employee tender valued it at 4 times its run rate. Our arithmetic on the records, each linked to its source. MoreSource: AIData.Broker, AI data company valuations
- Scale AI's valuation, round by round: over $1 billion at its 2019 Series C, over $3.5 billion in December 2020, $7.3 billion in April 2021, $13.8 billion in May 2024 and $29 billion with Meta's investment in June 2025. MoreSource: Scale AI; TechCrunch; SiliconANGLE
Filings and courts
- Eminem's publisher, Eight Mile Style, sued Suno over 269 compositions it says trained Suno's music generator, and Elvis Presley's rights owners filed their own suit over AI imitation of Presley's voice and the use of his songs (6 November). MoreSource: Detroit Free Press, via Yahoo News; Nexstar Media Wire
- The New York Times sued Brave Software (S.D.N.Y., 2 November) over scraping its sites into the Brave Index and the answers of Ask Brave. MoreSource: The New York Times Co. v. Brave Software, complaint (Dkt. 1)
- The hearing on Spirit Airlines' sale of its de-identified passenger data to Google moved from 6 November to 17 November; replies are due 13 November. The sale is not approved. MoreSource: U.S. Bankruptcy Court, S.D.N.Y., No. 25-11897, Dkt. 1779
- A São Paulo court barred Meta from using the news content of ANJ, Abert and Aner members to train AI, pending a final ruling (3 November). MoreSource: The Rio Times
Deals, and what buyers publish now
- The Associated Press licensed its content to Anthropic, Anthropic's first licensing deal with a digital publisher by Adweek's account. Neither side disclosed the terms. MoreSource: Nieman Lab, summarising Adweek
- Mercor's referral listings now state $50 million paid to companies for their data in 2026, and $2 million paid to those who introduced them. MoreSource: Mercor, referral listings
- micro1 raised its published tiers for company records to $100K-500K, $500K-3M and $3M+, from $100K-500K, $500K-2M and $2M+ on 26 October. MoreSource: micro1, data partnerships page
- DataVendor's documentation now says about $30,000 per licence is the floor for a good codebase, and SourceX invited businesses to license their data citing up to $4 million for the largest qualifying enterprise datasets, which it says is not a typical payout. MoreSource: DataVendor documentation; SourceX announcement (AiThority)