OpenAI's five-year licensing deal with News Corp is worth more than $250 million, according to the Wall Street Journal. That figure, reported when the deal was announced in May 2024, remains the largest disclosed AI content licensing agreement.
It is not an outlier. Reddit $RDDT signed a content licensing deal worth about $60 million per year with Google $GOOGL to train Gemini AI models, an agreement reached in early 2024, just before Reddit's IPO filing. In its IPO prospectus, Reddit disclosed data licensing arrangements with an aggregate contract value of $203 million and terms of two to three years, according to TechCrunch. Reddit CEO Steve Huffman had argued the platform's data should not be "given to some of the largest companies in the world for free."
These numbers map a market that barely existed three years ago. The global AI training dataset market was valued at $3.59 billion in 2025 and is projected to grow to $4.44 billion in 2026, according to Fortune Business Insights. A separate estimate from MarketIntelo values the dataset licensing segment at $4.8 billion in 2025, projected to reach $22.6 billion by 2034.
The era of free AI training data is over. What is replacing it has a price list.
What publishers are charging
The rates vary by orders of magnitude depending on the content, the publisher, and the buyer. At the top, News Corp's deal averages about $50 million per year across its portfolio of Wall Street Journal, New York Post, and global titles. Amazon $AMZN's first AI-focused licensing deal with the New York Times is worth $20 million to $25 million per year, according to the Wall Street Journal, as reported by eWeek.
Below that tier, the numbers fall but remain significant. Axel Springer's deal with OpenAI for Business Insider, Politico, and other titles was reported to be about $13 million per year over three years. The Financial Times signed with OpenAI at a reported $5 to $10 million per year. Dotdash Meredith's deal with OpenAI was worth at least $16 million, according to Digiday.
Academic publishers have extracted their own payments. Wiley executed a $23 million AI deal with an undisclosed tech company in fiscal 2024, followed by a second agreement worth $21 million, Publishers Weekly reported. Informa, the parent company of Taylor & Francis, signed a $10 million initial data-access agreement with Microsoft $MSFT and expects total AI-related revenues to exceed $75 million for the year, according to Inside Higher Ed.
Shutterstock leads all media companies in AI licensing revenue: CEO Paul J. Hennessy told Bloomberg the company expected $138 million in 2024, up from $104 million in 2023. Its CFO said individual Big Tech deals ranged from $25 million to $50 million each.
Pricing structures are shifting
The initial wave of AI content licensing deals in 2023 and 2024 focused on flat-rate annual fees, most of which were for training. That structure is already evolving.
Publishers now get paid when AI systems fetch or ground their content in real time, and Reddit anticipated this second wave of usage-based pricing. Reddit is discussing with Google and OpenAI a future deal structure that could enable dynamic pricing, where the platform earns more as its data becomes more valuable to AI-generated answers. Reddit CEO Steve Huffman said during a July 2025 earnings call that "every variable has changed since we signed those first deals" and that the company's corpus is "bigger" and "more essential," putting it in a strong position to renegotiate.
Meanwhile, revenue-sharing models are on the rise, according to Yulia Petrossian Boyle, founder of YPB Global, who noted that publishers earn a portion of subscription revenue or performance-based compensation. ProRata, an AI search startup, pays 50% of all revenues from its Gist.AI product to publishers, according to AI Watch.dog.
Personal data: Valuable in theory, cheap in practice
The deals above involve institutions. For individuals, the economics remain lopsided.
Vana, a startup spun out of MIT, has attracted more than a million users who contribute personal data across various pools, with typical individual payouts still small at roughly a few dollars, Vana told Axios. Vana CEO Anna Kazlauskas has acknowledged the scale problem: "My data on its own isn't that valuable, but a data pool with tens of thousands or millions of people is really valuable," she said in an MIT News profile.
"A lot of the fear around AI comes from the lack of proper attribution and economics," Kazlauskas told Quartz. "If you teach AI how to do your job, you should actually own that AI model."
The gap between institutional and individual compensation reflects a structural asymmetry. News Corp can negotiate from a position of legal leverage, archives, and editorial infrastructure. An individual's Spotify $SPOT listening history or Instagram engagement data has no comparable bargaining power unless pooled.
The approaching scarcity
These prices reflect a market under pressure. Epoch AI, a research organization that tracks AI scaling constraints, projects with 80% confidence that the stock of publicly available human-generated data will be fully utilized between 2026 and 2032, according to its published research. The indexed web contains about 500 trillion words of unique text, with a projected 50% increase by 2030. That sounds large, but frontier models consume data faster than the internet produces it.
The scarcity extends beyond volume. As AI systems move into robotics, augmented reality, and health care, they need data types that were never on the public internet. Drone imagery, fitness tracker logs, and enterprise telemetry sit behind walls of ownership, consent, and technical formatting.
Publishers signing deals in 2023 and 2024 were pricing an asset with no established market against a buyer with far better information about how much it needed them. That information gap is narrowing. Prices are rising. And the sellers, for the first time, know what they have.