With two major AI licensing agreements now signed in under two years, Rupert Murdoch's company has stopped asking whether to deal and started asking how much. The rest of the industry is following.
There is a version of this story in which News Corp (NASDAQ:NWSA) is the cautionary tale: a media company that spent years positioning itself as the toughest negotiator in the room, only to end up licensing its journalism to the same platforms it had publicly condemned.
Chief executive Robert Thomson's description of his publications as an "input company", analogous to semiconductors or data centres, does not obviously sound like a victory for editorial independence.
But that framing misreads what is actually happening. The News Corp CEO is not waving a white flag. He is making an argument about where durable value sits in an AI-driven economy, and he is cashing that argument into revenue before the window closes.
The holdout position was always temporary
The question was never whether publishers would license their content to AI companies. It was always who would go first and at what price. The first movers set the market. Everyone who follows negotiates against a benchmark already established by others.
News Corp understood this early. Its 2024 deal with OpenAI, worth $250 million over five years, was the largest publicly disclosed licensing agreement in the sector at the time. It told every other publisher what a premium catalogue was worth to the leading AI platform. The Meta deal, at up to $50 million a year, adds a second data point.
By the time the Guardian, the Washington Post, Axios and others signed their own agreements through 2025, the shape of the market was set. The deals on offer to latecomers are structured differently, and the financial terms, where disclosed at all, are generally less favourable.
What Thomson's "input" framing actually means
The semiconductor analogy deserves more scrutiny than it has received. Thomson is not being modest when he describes News Corp as an input company. He is making a claim about position in a supply chain.
Semiconductor manufacturers do not compete with the devices their chips power. They sit upstream, supplying something irreplaceable to everyone downstream. If you accept that real-time, verified, professionally reported information is genuinely scarce, and that AI products degrade without it, then the publisher's position in the AI supply chain is more defensible than the platform era suggested.
The platform era, by contrast, treated content as abundant and interchangeable. Publishers competed for placement in feeds that could be algorithmically adjusted at any moment. The bargaining power was entirely with the platforms.
AI changes that calculus, at least partially. Large language models trained on low-quality or synthetic data produce worse outputs. Publishers with deep archives and strong verification processes have something that cannot be easily replicated. Thomson's bet is that this scarcity is durable enough to support recurring licensing revenue rather than a one-off payment.
The $2.9 billion question
The broader market data support the idea that this was inevitable. Across the industry, AI companies have committed an estimated $2.9 billion in multi-year content licensing agreements with publishers. OpenAI alone has signed roughly 18 deals globally. Meta, before the News Corp agreement, had already contracted with CNN, Fox News, USA Today and People Inc.
The average of the publicly disclosed deal runs to around $24 million. News Corp's agreements, at the values reported, are multiples of that average. Size, reach and brand credibility command a premium.
What this market does not yet settle is whether the terms being agreed now will look adequate in five years. Critics of the licensing model argue that AI companies are paying for content at rates that bear no relationship to the cost of producing it, and even less relationship to the value they extract from it.
Investigative journalism, which is expensive to produce and uniquely difficult to replicate, may be the most underpriced asset in any of these bundles.
The holdouts are running out of runway
The New York Times remains the most prominent organisation still pursuing litigation rather than licensing. Its suits against OpenAI, Microsoft and Perplexity are working through the courts and could, if successful, force a significant repricing of the entire market. That outcome would benefit every publisher, including those that have already signed deals.
But the litigation path is slow, expensive and uncertain. Each publisher that signs a commercial agreement marginally weakens the collective leverage of those who have not. The holdout position becomes harder to sustain as the industry norm shifts and as AI companies demonstrate that they can build functional products with or without any single publisher's content.
The Guardian's move to partner with OpenAI in early 2025 was, symbolically, the most significant defection. A publisher with strong editorial independence credentials and a history of confronting concentrated media power concluded that commercial engagement served its interests better than resistance. That decision will have been noted in every remaining holdout's boardroom.
The real debate has already moved on
Whether News Corp's deals represent a sellout or a smart positioning exercise is becoming a less interesting question than what comes next. The first wave of agreements is approaching renewal. AI companies are beginning to shift focus toward commerce and transactions, where the relationship between content quality and commercial value may look quite different to the current training-data model.
Thomson's "input company" thesis holds only as long as what publishers produce remains genuinely hard to replace.
If AI systems become capable of generating credible-sounding current information without relying on licensed journalism, the leverage disappears. The deals being signed now are, among other things, a race to monetise that leverage before it erodes.
For the moment, the race appears to favour those who moved early and priced high. Whether that position can be held through the next cycle of negotiations is the question the industry has not yet answered.