Data licensing deals for model training are announced often and priced rarely. The research reads the parties' own filings and releases, sourced reporting, one tracker and two research papers, all viewed on 8 November 2026. Of 73 public deals collected in a January 2026 paper, 57 disclose no revenue [30]. Where money appears, it is usually a licensor's revenue line, a figure from an unnamed source or a newspaper's estimate [7, 15, 24]. Reddit's registration statement is the one filing here with a contract value and a term: $203.0 million over two to three years [1]. This is research, never legal advice.
What data licensing deals disclose
An announcement names the parties and what is licensed. None of the five below states a price, and only News Corp's gives the length, as "multi-year" [5, 6, 13, 22, 26].
| Deal, as announced | Date | What is licensed | Price or term |
|---|---|---|---|
| Reddit and a search and model developer [5] | 22 Feb 2024 | Reddit's Data API, so the developer can "display, train on, and otherwise use" Reddit content | Neither |
| Reddit and a model developer [6] | 16 May 2024 | Reddit's Data API for the developer's products, with the developer also a Reddit advertising partner | Neither |
| News Corp and a model developer [22] | 22 May 2024 | Current and archived content from News Corp mastheads, to display in response to user questions and "to enhance its products" | "Multi-year", no value |
| The Associated Press and a model developer [26] | 13 Jul 2023 | Part of AP's text archive | Neither |
| Dotdash Meredith and a model developer, in IAC's results [13] | 7 May 2024 | Attributed content in the developer's chat responses, also used to improve the developer's models | Neither |
Collections of public deals find the same gap. Three of the 23 entries in Ithaka S+R's tracker of publisher deals carry a deal size [29]. The January 2026 paper counted 16 of 73 deals with an amount disclosed, and the 15 with a quoted revenue volume sum to $677.3 million [30].
Reddit's filings: a contract value, then revenue
Reddit's S-1, filed 22 February 2024, says it entered data licensing arrangements in January 2024 "with an aggregate contract value of $203.0 million and terms ranging from two to three years", and expected at least $66.4 million of that as 2024 revenue [1]. It lists "Model Training" next to "Data API Access" [1]. Delivery runs through "continuous access to our data API as well as quarterly transfers of Reddit data", for a fixed or usage-based fee recognized "generally ratably over the license period" [1].
After that, the money sits inside other revenue, a line that also holds products sold directly to users [3]. Other revenue was $15.2 million in 2023, $114.7 million in 2024 and $140.0 million in 2025 [2, 3]. Reddit's contracted revenue still to be recognized, mostly from long-term content licensing contracts, ran down from $252.9 million at 31 December 2024 to $143.7 million a year later and $62.4 million at 30 September 2026 [2, 3, 4].
The filings name no licensee [1, 2, 3, 4]. They say that "substantially all of the contract value associated with our licensing revenue is derived from two of our partners" [2, 3], and the quarterly report filed on 30 October 2026 adds that those agreements "are approaching the end of their respective current terms" [4].
Data licensing revenue in publishers' filings
Publishers and image libraries report licensing income in results statements.
| Licensor | Figure as filed | Period | What the line covers |
|---|---|---|---|
| John Wiley & Sons [15, 16] | "AI licensing revenue" of $23 million, then $40 million, then "AI revenue" of $49 million | Fiscal years to April 2024, 2025 and 2026 | Agreements with unnamed large technology companies, then a recurring stream "anchored by landmark partnerships" with two named companies [15, 16] |
| Informa, for Taylor & Francis [18, 19] | "$75m+ of data access revenue which is non-recurring" | 2024 | Non-exclusive access to archive content |
| Shutterstock [8] | Data, Distribution, and Services, $203.3 million | 2025 | Metadata licenses plus Giphy, Shutterstock Studios and other services |
| IAC, for People Inc. [14] | Licensing and Other revenue up $32.9 million, or 28% | 2025 | Apple News+, syndication partners and a full year of revenue from a model developer |
| The New York Times Company [25] | Licensing revenues up $14.4 million | 2025 | "commercial agreements with third-party digital platforms" |
Only IAC and Wiley name counterparties in their revenue discussion, IAC a model developer [14] and Wiley two partners for fiscal 2026 [16], and neither gives an amount per partner.
The revenue in these lines arrives in lumps. Shutterstock's data offering fell 91% in the third quarter of 2026, and the company says revenue there varies "based on the delivery timing of metadata licensed assets" [9]. Getty Images told lenders in October 2026 that about $40 million from multi-year licensing deals signed in 2025 was recognized early, in the fourth quarter of 2025 [12].
Where reported prices come from
Every price tied to an identified buyer in this set comes from reporting, an interview or a tracker. Reuters reported on 22 February 2024 that Reddit's deal with the search and model developer is "worth about $60 million per year, according to one of the sources", none of them named [7]. On 22 May 2024, Engadget reported a Wall Street Journal estimate of $250 million over five years in cash and credits for News Corp's deal [24]. News Corp owns the Journal, and its own release disclosed no value [22].
Shutterstock's chief executive told Bloomberg in an interview reported on 4 June 2024 that its "licensing business with AI companies produced $104 million in revenue" in 2023 [10]. The filed line for that year, which includes Giphy and Studios, was $137.3 million [8]. The Conversation reported on 23 July 2024 that Informa was to be paid more than £8 million for initial access in its deal with a software company [21], and Ithaka S+R's tracker records the same deal as $10m [29].
Publishers Weekly reported on 18 November 2024 that HarperCollins licensed "select nonfiction backlist titles" to an "undisclosed AI company", and that authors who opt in receive $2,500, citing an agent's email that an author shared on Bluesky [27]. The tracker records $5,000 per book split evenly between author and publisher, over three years [29].
How the deals are structured
The record shows these structures:
- Access over a term: an API feed plus quarterly transfers for two to three years, at Reddit [1].
- Archive plus new content: non-exclusive access to historical content, with recurring annual fees over three years for newly created knowledge, at Taylor & Francis [18].
- Display in the buyer's products: News Corp content shown in response to user questions [22], and attributed Dotdash Meredith content in chat responses [13].
- Cash plus credits, at News Corp as reported [24, 31].
- Equity as payment: Shutterstock held $30.5 million of equity instruments received from customers at 30 September 2026 [9].
- Per title and opt-in, for three years, at HarperCollins as the tracker records it [29].
- Retrieval and citation inside a product, in at least five tracker entries, among them NEJM Group, Springer Nature and De Gruyter Brill [29].
A second paper, documenting 78 deals from 2020 to 2025, finds that buyers in large enterprise deals "pay fixed fees for time-limited access to curated corpora or feeds" [31]. Payment to the people who made the content is rare in the record. Six of 73 deals mention a revenue share with contributors [30], and 7 of 24 major deals compensate the original creator [31].
What disclosure rules require
SEC rules set what a filer has to disclose about a contract. A contract made in the ordinary course of business need not be filed as an exhibit unless, among other cases, the business is "substantially dependent" on it, and a filer may redact terms it treats as confidential when they are not material [32]. Form 8-K's Item 1.01 asks for the date, the parties and the material terms of a material agreement made outside the ordinary course [33]. In the filings reviewed, the deals appear as revenue lines, risk factors and results releases, and none as an Item 1.01 agreement [1, 2, 3, 4, 8, 9, 11, 12, 13, 14, 15, 16, 17, 23, 25]. For how results releases reach markets, see company disclosure timing, and for SEC exhibits with redacted terms, what makes a GPU capacity contract financeable.
The European Commission's template for public summaries of training content requires "limited disclosure of information" for licensed data: a yes or no, and the kinds of content [34]. California's AB 2013 asks developers for a high-level summary that includes "The sources or owners of the datasets" and "Whether the datasets were purchased or licensed by the developer" [35]. The EU template asks for no counterparty, and neither asks for a price [34, 35].
What a buyer can take from the record
The record shows that deals exist, who made some and what a few were said to be worth. It rarely shows the permitted uses, the term or onward sharing, so a buyer settles those in the agreement itself. The clauses are covered in what a training data license grants, and the checks before signing in buying training data. Through its training data acquisition and licensing work, Rillor finds the publishers and data holders that hold the data a team needs, agrees the use, the term and onward sharing in writing before delivery, and delivers every record with the license it came under.
Read the full report
The full report adds the filed figures by period, the sourcing behind each reported price and the disclosure rules in their own words.