Compliance first: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.
Home / Insights / In the news
In the news

What Have AI Companies Paid for Data? The Licensing Deals So Far

What Have AI Companies Paid for Data? The Licensing Deals So Far

From Reddit's reported $60 million-a-year deal with Google to Shutterstock's $104 million in AI licensing revenue, here is what public reporting shows AI developers have paid for training data, and what it signals for business data.

Two years ago, "data licensing" sounded like a niche corner of media law. Today it is a line item in IPO filings and earnings calls. The deals below are public, verified market news, collected so business owners can see how the market for training data has been priced so far. They are not OpenDataPolicies transactions, and none of these companies is presented as a buyer of data through us.

By OpenDataPolicies Editorial · October 6, 2026 · 6 min read

The short answer

Publicly reported AI data licensing deals range from undisclosed amounts to hundreds of millions of dollars. The best-documented figures include a reported $60 million per year for Reddit's data from Google, $203 million in aggregate data licensing contracts disclosed by Reddit, a reported deal worth more than $250 million over five years between OpenAI and News Corp, and $104 million in 2023 AI licensing revenue at Shutterstock.

Most of those dollars went to companies that hold large, unique, hard-to-replicate collections of human-created content. That is the pattern worth noticing.

Reddit and Google: the deal that made headlines

In February 2024, Google struck a deal with Reddit that, according to reports, was worth about $60 million per year. As The Verge summarised it, the agreement gives Google real-time access to Reddit's data for AI training, while Reddit uses Google AI to improve search on its own platform. CBS News, carrying the Associated Press report, described it as a deal allowing the search giant to train AI models on human posts.

Why would anyone pay that much for forum posts? Because Reddit is one of the largest archives of real people asking real questions and getting real answers, in millions of niche communities. It is conversational, specific and constantly refreshed: exactly what generic web crawls lack.

Reddit's IPO filing: $203 million in data contracts

The same week, Reddit filed for its stock market listing. As TechCrunch reported, the filing disclosed that Reddit had entered into data licensing arrangements with an aggregate value of $203 million. Reddit pitched its data as a key element in training leading large language models and as a new revenue channel alongside advertising.

The lesson for everyone else: a company's archive of user-generated and operational content was presented to public-market investors as an asset with a dollar value. Data is no longer just exhaust from running a business.

News publishers: News Corp, the AP and others

News archives were among the first content types to be licensed.

  • News Corp and OpenAI (2024). The companies announced a multiyear agreement in May 2024 without disclosing terms. According to Engadget, citing The Wall Street Journal, News Corp was estimated to receive $250 million over five years in cash and credits. OpenAI gained access to current and archived articles for training and for answering user queries.
  • The Associated Press and OpenAI (2023). In an AP press release, the two organisations said OpenAI would license part of AP's text archive, while AP would leverage OpenAI's technology and product expertise. Financial terms were not disclosed.

Notice the structure: cash plus technology credits, multi-year terms, and access to archives rather than one-off file dumps. Those are the same levers (term, exclusivity, payment form) that appear in business data deals. Our guide to data licensing agreements explains them in plain English.

Stock imagery: Shutterstock's $104 million

Images and video are training data too. According to Business Insider's reporting syndicated on AOL, Shutterstock CEO Paul Hennessy said the company earned $104 million in revenue from AI licensing agreements in 2023, and projected that this revenue could reach up to $250 million annually by 2027.

That is a stock-photo company discovering that its library, built for a completely different purpose, had a second life as AI training material. Many operational businesses are in a similar position with their records today.

What the deals have in common

Look across the reported deals and a few patterns stand out:

  1. Uniqueness drives price. Reddit's conversations, News Corp's journalism and Shutterstock's licensed imagery cannot be scraped freely at the same quality. Unique data is king.
  2. Scale and history matter. Every one of these sellers had years of accumulated content. Archives command premiums.
  3. Rights are clean. The sellers owned or controlled licensing rights to what they sold. Clean provenance is part of the product (see data provenance).
  4. Terms are multi-year. Buyers want continued access and fresh data, not just a snapshot.
  5. Timing matters. Many of these deals were struck in 2023–2024, as researchers warned that public data was running short (see Is AI running out of training data?).

None of these deals were for personal data in the privacy-law sense. They were licences to content and archives, with obligations around how it is used.

The gap these deals leave open: operational data

Here is what is missing from the headline deals. Reddit tells a model how people talk about work. News archives tell it what happened. Stock photos show what things look like. None of them show how work actually gets done, step by step, inside a business.

That is the domain of operational data: the support ticket and its resolution, the change order in a construction project, the dispatch log of a plumbing company, the intake workflow at a law firm. Model builders working on AI agents need exactly this kind of workflow data, and it lives almost entirely inside private companies. We cover why in operational data vs web data.

Large publishers could negotiate their own deals. A 25-person HVAC company or a regional accounting firm usually cannot, which is the gap OpenDataPolicies was built to fill: brokering de-identified operational data from businesses of every size to AI labs and model developers.

Questions to ask before you license anything

The big platforms negotiated with in-house lawyers and bankers. A smaller business should still ask the same questions, even if the paperwork is simpler:

  • Is it a licence or a sale? Most AI data deals are licences: the buyer gets defined rights to use the data, and you keep ownership. See perpetual license and term license.
  • Is it exclusive? Exclusivity usually raises the price but stops you from licensing the same data to anyone else.
  • What can the data be used for? Training, evaluation and fine-tuning may be treated differently in the agreement.
  • How is payment structured? Lump sum, instalments, or a mix; the details are set in the purchase agreement.
  • What happens to the data afterwards? Retention, deletion and onward-transfer limits should be written down.

Our guide to is it legal to sell business data? walks through the rights questions. It is informational, not legal advice, so involve your own counsel before signing.

What a small business might expect

Public deals involving household-name platforms are not a price list for your business, and you should be wary of anyone who implies they are. Your data's value depends on its volume, years of history, industry, structure and uniqueness.

According to founder Jeff Cline, small companies have received over $100,000 for de-identified operational data; bids can come within 72 hours if the data is correct and a header sample is provided; sellers receive offers from multiple buyers so they can choose the highest bid; and deals typically close in about 90 days. Results vary; not a guarantee.

One thing is not negotiable. We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control. Buyers in this market want the shape of the work, not the identities of the people involved.

How to find out what your data is worth

Start with a quick estimate in the valuation calculator, then read what data buyers want to see how your systems compare. If you hold years of records in tools like Gmail / Google Workspace, HubSpot, Jira or ServiceTitan, you may have the kind of archive buyers are looking for.

The headline deals show that buyers will pay for unique data. They also show that the early sellers set the terms. Request a free bid to see what buyers would offer for yours.

Our compliance promise: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Frequently asked questions

How much did Google pay Reddit for data?

Neither company confirmed the figure, but the deal announced in February 2024 was reported to be worth about $60 million per year, giving Google real-time access to Reddit data for AI training.

Do AI companies pay for training data?

Yes. Public reports and filings show licensing deals with platforms, news publishers and stock-media companies, from undisclosed sums to reported values of hundreds of millions of dollars over several years.

Are these companies buyers on OpenDataPolicies?

No. These are public news reports included for market context. OpenDataPolicies does not name the buyers it works with; we describe them generally as AI labs and model developers.

Can a small business license its data like Reddit did?

Small businesses rarely have the scale to negotiate alone, but de-identified operational data can still be valuable. According to founder Jeff Cline, small companies have received over $100,000. Results vary; not a guarantee.

Sources

  1. The Verge — Google cut a deal with Reddit for AI training data (Feb 2024)
  2. CBS News / AP — Google strikes $60 million deal with Reddit (Feb 2024)
  3. TechCrunch — Reddit says it's made $203M so far licensing its data (Feb 2024)
  4. Engadget — OpenAI will reportedly pay $250 million to put News Corp's journalism in ChatGPT (May 2024)
  5. Associated Press — AP, OpenAI agree to share select news content and technology (2023)
  6. Business Insider via AOL — Shutterstock earned over $100 million in revenue (AI licensing, 2023)

Updated November 19, 2026.

OpenDataPolicies.org connects businesses with buyers of de-identified operational data. We are not a law firm, broker-dealer or investment adviser, and nothing on this site is legal, tax or financial advice. Statements about amounts, timing and outcomes are Jeff Cline's statements based on past transactions. Results vary; nothing here is a guarantee of any bid, price, buyer or closing. Every deal depends on the data, its quality, its rights and buyer demand at the time.