Academic publishers are rushing to sign licensing deals with artificial intelligence (AI) companies, carving out a new revenue stream as US research funding cuts dim their outlook.
Informa Plc’s Taylor & Francis signed a US$10 million (S$12.8 million) deal with Microsoft Corp. in 2024 to provide the technology giant access to part of its library to train large language models, or LLMs.
Bloomsbury Publishing Plc is looking to “monetise academic content through AI deals”, it said in its latest set of results. John Wiley & Sons Inc. announced partnerships with Amazon Web Services and Perplexity earlier in 2025.
The deals provide a boost to earnings for publishing giants. Informa sold US$75 million of non-recurring data access in 2024 through its partnership with Microsoft and another undisclosed partner, which helped to accelerate underlying revenue growth in the Taylor & Francis unit to 15 per cent from 3 per cent the year prior. Excluding AI deals, organic growth would have been closer to 3.5 per cent, according to Berenberg analyst William Larwood.
This could offset some of the impact from the US administration’s decision to cut federal spending on academic research, Mr Larwood said in a note in May. About 11 per cent of Taylor & Francis’ subscription revenues could be at risk from US funding cuts in 2026, he calculated.
A 43 per cent cut to the National Institutes of Health’s budget for fiscal 2026 – what US President Donald Trump is pushing for – would result in a £33 million (S$57.4 million) revenue headwind for Taylor & Francis. A less drastic cut of 25 per cent would mean a shortfall closer to £19 million, Mr Larwood estimated.
Even for those who are steering clear of licensing deals for now, AI is unavoidable internally. Springer Nature AG & Co KGaA, for instance, has not signed any licensing agreements so far, but is using AI to tighten workflows. “It’s not our first priority,” its chief executive officer Frank Vrancken Peeters said on the most recent earnings call.
“The main emphasis in terms of AI for us has been transforming the publication process, making it faster, more researcher-friendly and more reliable in terms of research integrity,” Springer Nature said in an e-mailed response to questions.
The recent flurry of licensing deals in the academic world is part of a broader trend of publishers both fighting against and joining forces with powerful AI companies.
The New York Times Co. recently reached a deal to license its editorial content to Amazon.com Inc. for use across its artificial intelligence platforms, after years of battles in court with OpenAI and its partner Microsoft over copyright infringement. The Atlantic and Vox Media signed similar deals, as publishers try to prevent tech companies from scraping the web to train chatbots without providing fair compensation.
“Licensing deals could be the most straightforward path to resolving copyright disputes between the developers of large language models and the owners of the content upon which those models were trained,” wrote Ms Tamlin Bason, a legal analyst at Bloomberg Intelligence.
Compensation remains a key point of contention in academic publishers’ AI deals. Some authors whose work was licensed out said they were not given the opportunity to opt out of the partnership and were not appropriately compensated.
Microsoft offered HarperCollins US$5,000 per title, split 50-50 with the author. There is no ballpark in terms of what writers can expect, according to Ms Mary Rasenberger, the CEO of the Authors Guild, whose position is that authors should be asked for permission and given 75 per cent to 85 per cent of the revenue from the AI deal. An author was recently paid just US$97 by Taylor & Francis to give over a book for LLM training, she told Bloomberg.
“Proceeds from these partnerships are being shared with authors and other rights holders in accordance with the licensing terms and royalty statement periods in their contracts,” a spokesperson for Taylor & Francis said.
“Academic authors don’t have expectations of making that much money from their books,” Ms Rasenberger said. “It’s more that they don’t want to lose control over their work.”
Beyond souring relationships with their authors, the bigger concern for publishers is becoming obsolete in a world of AI-generated content.
“It’s definitely a risk, without a doubt,” Bloomberg Intelligence analyst John Davies said. The key question is “how can you give certainty to your authors and your shareholders that you’ve not chucked the baby out with the bath water”, he added.
Publishers’ experience with selling rights to their data should be key in implementing the appropriate guardrails, Mr Davies said. Taylor & Francis has negotiated clear rules around display rights, repurposing and contiguous word count – verbatim extracts of more than a set amount of words – its CEO Stephen Carter said in an earnings call in 2024.
Berenberg’s Mr Larwood is not too worried either, pointing to the intellectual property rights embedded in the Taylor & Francis agreements. “This suggests that it will look to sign more contracts in the future, providing upside to estimates,” he said. BLOOMBERG
Join ST’s Telegram channel and get the latest breaking news delivered to you.
