What is a data monetization company?
A data monetization company pays a business for data that business already holds, or helps it license that data to someone who will. In 2026 the buyers are mostly AI developers. Labs training models and agents need real records of how work gets done, so they pay for team chat, email archives, CRM history, support tickets, code with its tickets and documents, call recordings and finance exports. One label covers companies with very different roles: some buy outright, some resell to labs at a markup, some run a marketplace, and a few work for the seller.
This page lists every company we have found that buys, licenses, brokers or resells business data for AI training. It leaves out consumer data brokers, ad-tech audience sellers and analytics vendors that help you use your own data internally. Those are a different market that happens to share the name. For the price ranges and the sources behind them, see what company data sells for and our sources and methodology.
The six types of data monetization companies
The type filter above uses the same six groups. Each one buys differently, pays differently and carries a different risk.
1. AI data vendors (13)
Vendors such as micro1, Handshake AI, Mercor, Polyshares, Telegraph Lab and Ooak Data license an operating company’s records and supply them to AI labs. They are the most active buyers of Slack, email, CRM and ticket history, and the ones with published ranges pay $100K to $4M per license. Most de-identify the data themselves after it leaves you, and several publish perpetual, exclusive terms with offers that expire in 14 days. Scale AI, Appen, Turing and Surge AI are established labelling vendors that have added company data to what they buy.
2. Data marketplaces (6)
On Grepped, DataVendor (HUD), Protege, Troveo, Wirestock and Human Native AI you list the dataset, set or accept a price, and the platform takes a cut when a buyer pays. The two that publish a split keep 20% to 40%. Troveo, Protege and Wirestock focus on video, audio, images and health imaging; Grepped and DataVendor list chats, codebases and documents. Payment usually follows the lab’s payment, sometimes by 45 to 60 days.
3. Wind-down buyers (2)
SimpleClosure and Sunset buy or broker the archives of startups that are closing, with published deals of $10K to $100K per company for code, chat, email and documents. Both are also the shutdown service provider, so the company handling your dissolution is the same one making the offer, and end buyers are not always disclosed to founders. Coverage in MIT Technology Review and Fast Company describes the market.
4. AI labs (8)
Google, xAI, OpenAI, Meta, Anthropic, Amazon, Microsoft and Apple are the end buyers. Their disclosed deals are mostly with publishers, from $5,000 per book title to more than $250M over five years. Direct purchases of company records are rare: Google’s $10M bid for Spirit Airlines’ records in its 2026 bankruptcy auction, reported by CSO Online and TechRadar, is the public example. Most labs buy company data through the vendors above or through partnership programs with an interest form.
5. Brokers (2)
A broker works for the seller and never takes a position in the data. Bright Info runs sealed bid rounds among the vendors, marketplaces and labs on this page, negotiates term, exclusivity and use, and is paid a commission from the closed deal by the buyer the owner chooses. Miro Advisory describes a similar service but publishes no fee, no team and no track record, and its domain was registered in April 2026.
6. Big tech infrastructure (1)
Cloudflare’s Pay Per Crawl lets a website set a per-request price that AI crawlers pay to read its pages. It monetizes public web content rather than internal records, so it sits in its own category.
How much do data monetization companies pay?
Three ranges have been disclosed in 2026, by the companies themselves, by the press or in court filings.
| Seller | Range | Basis |
|---|---|---|
| Closing startups | $10K–$100K | Per company, for workspace and code archives sold through wind-down services. Largely self-reported by the intermediaries. |
| Operating companies, 20–500 staff | $100K–$4M | Per license, from the partnership ranges AI data vendors published in 2026. |
| Large enterprises | $7.5M–$12.5M | Bids for Spirit Airlines’ records (about 100 million emails, 500 million Teams messages and 30 million lines of code) in its 2026 bankruptcy auction, from court filings. |
Within a range, four things move the number: how many years the record covers, how many systems it spans, how much visible reasoning it contains (decisions, escalations and handoffs rather than notifications), and whether the license is exclusive. The Spirit auction also showed what one competing bid does: a second bidder raised the offer from $10M to $12.5M, a 25% increase.
How to choose a data monetization company
Six questions separate a good offer from an expensive one. The “Watch for” column records where public reporting has raised each of them.
- Who de-identifies the data, and when? Names, customers and accounts should be removed before the record leaves your systems, by a party you can verify, not by the buyer after transfer.
- Is the license bounded? Look for a fixed term, a field of use limited to AI training, and resale rights that stay with you. A perpetual exclusive license means you can never sell the same record again.
- When do you get paid? Terms range from seven days after signing to 60 days after the lab pays the marketplace. Milestone and revenue-share structures move the risk to you.
- Is the end buyer disclosed? Most vendors resell to labs at an undisclosed markup. You should at least know the category of the buyer who ends up with your data.
- What does the company disclose about itself? Fees, team, security history and ownership. A March 2026 breach, a 49% stake held by an AI lab, or an anonymous operator are all on this page.
- How many offers will you see? One buyer sets one price. A broker or a competitive process puts several offers side by side and lets you pick, or walk away.
The security and data handling page describes how Bright Info applies each of these in its own process.
How we built this list
Every row comes from the company’s own site or pricing page, from press reporting, or from court filings in the Spirit Airlines case. Ranges marked “reported” are third-party figures that the company has not confirmed. “Watch for” notes are drawn from public reporting and are not our own claims; “None found” means we found nothing, not that nothing exists. We review the table monthly and when a deal is disclosed, and the date at the top of the page changes when a row does. Bright Info is a sell-side broker and appears in the table as such; the other 31 companies are the buyers it invites to bid. To correct a figure, write to hello@brightinfo.com.
Primary sources for the market figures on this page:
- CSO Online, TechRadar and AI Weekly on the Spirit Airlines auction: Google’s $10M winning bid, Mercor’s $7.5M backup bid and micro1’s late $12.5M offer, from court filings.
- MIT Technology Review, Fast Company and Gizmodo on wind-down archive sales of $10K to $100K per company.
- TechCrunch on the agent-training environment market.
- Each company’s own site or pricing page, linked from its name in the table, for published ranges and terms.
Full list and the method behind the $100K–$4M operating-company range: sources and methodology.