Is the era of ephemeralization finally over or just beginning?

For years, libraries have attempted to navigate inflationary pressures and unsympathetic budgets by following the maxim of “doing more with less,” in part by developing efficiencies and leaning into automation. Architect and futurist Buckminster Fuller described doing more with less as “ephemeralization” and predicted a future where technology would advance to the point that more and more could be done with less and less until eventually everything could be done with almost nothing. As libraries, vendors, and publishers rush to integrate artificial intelligence (AI) into their offerings and workflows, optimists in libraryland see AI as the next logical step in the continuing ephemeralization of library products and services. Others are more pessimistic: with long held assumptions about the value of public knowledge and higher education changing, some librarians may find themselves resigned not to doing more and more with less, but doing less and less a little more. 

ECONOMIC CONTEXT

Overall, the U.S. economy remained robust this year. Strong consumer spending, tax reform, and AI-driven growth have some economists predicting an even stronger economy in 2026. Per the Bureau of Economic Analysis, the nation’s gross domestic product (GDP) increased at an annual rate of 2.2 percent in 2025, while according to the Bureau of Labor Statistics, the Consumer Price Index (CPI) all items index increased 2.7 percent. Still, concerns about inflation and economic growth persist, with the Consumer Confidence Survey and the Michigan Consumer Sentiment Index both reporting year over year drops. Slower hiring, major shifts in social and economic policy nationally, continued political instability globally, and fears of an AI-driven stock market bubble mean that many public institutions will choose to budget extremely conservatively.

NASBO, the National Association of State Budget Officers, reports that state spending on higher education increased approximately 3.9 percent in fiscal year 2025. That increase covers inflation, but not by much. With most states’ revenue slowing, states will be under pressure in future years just to maintain existing levels of higher education spending. In addition to facing increased costs for both everyday operations and basic transportation infrastructure, states will also need to assume responsibility for a greater share of Medicaid and Supplemental Nutrition Assistance Program costs as federal changes enacted in the One Big Beautiful Bill Act take effect in the coming years.

ROCKY SEAS FOR HIGHER EDUCATION

Higher education continues to be buffeted by challenges. The only certainty for the next few years appears to be uncertainty, which makes budgeting—particularly long-term budgeting—difficult. The National Association of College and University Business Officers noted top challenges facing higher education include managing increasingly unreliable funding sources, handling rising operational costs, navigating political and policy disruptions, and effectively demonstrating the value of higher education. Many institutions are opting to freeze, shrink, or cut programs entirely, focusing instead on programs with higher impacts. For higher education, an industry used to expanding services rather than reducing them, the market correction may prove to be painful.

As college enrollment continues to plateau—whether due to demographic changes and/or increased public skepticism about the return on investment in higher education—a disquieting fact becomes more apparent: there are too many schools and not enough students willing to pay to attend them all. Loan cap changes in the One Big Beautiful Bill Act, while intended to pressure colleges to lower tuition and reduce student debt, could have a significant impact on enrollment, driving some students away from high-cost graduate programs. According to the National Student Clearinghouse Research Center, the four percent increase seen in FTE enrollment growth at higher education institutions in 2024 dropped to a one percent increase in fall 2025. Flagship universities and community colleges were the main benefactors of that small increase, while enrollment at private four-year colleges continued to decline, as did international graduate student enrollment. Notably, vocational-focused community colleges saw a 16 percent enrollment surge, a sign that Generation Z is continuing to see better job security in blue-collar jobs than in white-collar ones, many of which are threatened by the growth of AI. The growing interest in trades, combined with an economy increasingly driven by skills in data science, information technology, and business analytics, means higher education will continue to pivot its offerings accordingly in order to remain relevant. Similarly, as students, educators, and practitioners increasingly incorporate AI into their daily work and the need for AI skills becomes increasingly important, institutions will continue to develop comprehensive AI and machine learning courses that accommodate both technical and non-technical students, maintaining academic rigor while “addressing industry demands.”

AI? AIEEEE!!

While many in higher education and scholarly communications are embracing AI, what’s not clear is how quickly AI-related services will bring the operational efficiencies and reduced staffing levels that many public institutions are counting on in a future of doing less with less. In turn, while many AI companies are counting on public institutions being a reliable and lucrative source of income for their products and services, it’s also unclear when these companies will begin to see a return on their massive investments in building out AI-related infrastructure like data centers. As a percentage of GDP, the $670 billion that Alphabet, Amazon, Meta, and Microsoft are expected to spend in 2026 to build out AI-related infrastructure is more than the capital laid out in the Golden Age of U.S. railroad expansion in the 1850s, and second only to the Louisiana Purchase in 1803.

For libraries looking to emphemeralize operations, embracing AI remains a double-edged sword. On the one hand, AI has already changed how and how quickly users expect to find and retrieve information. The speed with which AI agents can ingest information and create catalog records means backlogs of uncatalogued materials could be made discoverable for a fraction of the time and cost of conventional original cataloging. On the other hand, concerns remain about hallucinations and the accuracy of search results, and the potential biases resulting from applications built with large language models. Metadata creation in libraries relies heavily on critical thinking and analysis, whereas most AI systems at this point are largely very efficient pattern-matchers. To date, most experiments with using AI to create bibliographic records show that human review and oversight will continue to be required. Many libraries will find themselves in a chicken-and-egg situation: building effective AI tools to streamline operations, developing the accompanying guidelines and strategies, and evaluating AI-generated answers and output will require time and resources. For libraries, the obvious route to doing less with less may still require doing more.

Although many libraries are incorporating AI literacy into their instructional offerings, most are just beginning to develop guidelines for using it in library work. Only a few institutions have developed guidelines for handling AI-related clauses in licenses, clauses that are becoming more common as publishers try to predict how their content will be used after it has been ingested into a large language model.

Large publishers have already begun to take advantage of AI, introducing AI products that can assist researchers by sifting through and summarizing the thousands of articles on their platforms. Many publishers are also using AI to speed up the manuscript submission, review, and acceptance process, as well as to maintain the integrity of the scholarly record by catching papers that may have problematic text, references, or images. The International Association of Scientific, Technical & Medical Publishers (STM), for example, now offers an AI tool originally developed by SpringerNature to identify AI-generated nonsense text in manuscripts. STM is also in the early stages of developing a global standard for disclosing the use of AI in research articles.

As parent institutions begin to develop their own AI agents, libraries may face pressure to negotiate the rights to use subscribed content into these agents. But such access may come at a price, further pinching libraries looking to do less. Similarly, as researchers become more reliant on using AI agents to streamline search and summarize thousands of articles in a results list, libraries may also face pressure to subscribe to the AI products now being rolled out by most of the large commercial publishers.

A bigger question might be how AI will affect open access (OA) publishing. While open access purists may cringe at the idea that freely available academic content is being used to train large language models owned by commercial AI companies, excluding high-quality content may lead to AI outputs that are more inaccurate or more biased. Regardless, proponents of a sustainable and affordable scholarly communications ecosystem may be justified in wondering whether AI will further accelerate the path to open or whether it will further concentrate dissemination of scholarly output in the hands of a few large commercial entities. 

OA GROWTH STABILIZES

The big five commercial publishers (Elsevier, Sage, SpringerNature, Taylor & Francis, Wiley) continue to dominate OA publishing. However, the rapid growth of OA journals seen over the past decade appears to be over, with the Open Access Scholarly Publishing Association and DeltaThink both reporting that OA growth has begun to level off. Similarly, Rick Anderson, writing in the Scholarly Kitchen, suggests that in the absence of any university requiring faculty to publish openly or the government requiring open rather than public access, the global transition to fully OA has begun to stall. For the foreseeable future, Anderson notes that scholarly publishing will likely continue to disseminate output using a mix of models, an approach that probably results in a healthier publishing ecosystem. 

In an era that sees doing less with less colliding with the rapid adoption of AI, some publishers may decide to coalesce their OA efforts around a single publishing model. An OA model that remains popular with libraries because it does not require comprehensive read and publish agreements or utilize article processing charges is Subscribe to Open or S2O. With S2O, if enough institutions subscribe or renew, that year’s content is opened. This year, both Royal Society Publishing and BioOne announced that they had met S2O funding targets for opening their subscription journals. The flipside of the S2O model is that if participation is not sufficient, that year’s content remains gated, potentially resulting in titles with a patchwork of gated and open access, depending on the year. The uncertainties associated with relying on the S2O subscription model to fund operations and expansion may be too great for some publishers. At the end of 2025, for example, AIP Publishing ended the S2O pilot for two of its journals after concluding that, while S2O offered a level of stability, it was not a sustainable model for them in the long term. Similarly, Wiley’s decision that its Knowledge Unlatched subsidiary (KU) would be a better fit under a nonprofit publisher like Annual Reviews, suggests that the crowdfunding model KU uses to open content may have struggled to mesh with Wiley’s more conventional use of article processing charges.

A new model that holds promise is the Open Journals Collective (OJC). OJC promotes itself to libraries as an “exit ramp” from costly Big Deals and hopes that by pooling resources, libraries, scholarly societies, and small nonprofit publishers can create a more visible and sustainable ecosystem for diamond open access publishing. While such a model is unlikely to shift commercial publishers’ share of the market in the short term, it offers a more acceptable model for those academics and researchers who, justifiably or not, feel that the dominant model of scholarly publishing is too focused on the needs of shareholders, rather than efficiently disseminating research output. Yet another option for smaller publishers wanting to expand their reach but unable to offer open access or to afford costs associated with bringing their platform in compliance with accessibility requirements is to have their content hosted in an aggregator database. Both models may be appealing for smaller publishers who may be faced with the daunting task of making their online content compliant with Title II accessibility requirements.

THE 2026 FORECAST

The sub–6 percent price increases seen for several years after both the 2008 recession and 2020 pandemic are inching back toward the 6–7 percent annual price increases seen for much of the early 2000s and again in the early 2010s. For the 2027 renewal season, prices are expected to increase by 6 percent. Price increases were also analyzed for 7,700 e-journal packages handled by EBSCO Information Services. For 2025, the average rate of increase was nearly 5 percent, up from 4 percent in 2024, but in line with the 5 percent seen in 2023. For 2025, Read & Publish packages processed by EBSCO showed an average increase of 4.8 percent, a very small increase over 2024’s increase of 4.22 percent. 

Ephemeralization has its roots in the word ephemeral, meaning “transitory” or “lasting for a very short time.” Libraries, of course, are not ephemeral and will continue to play an important role in terms of curating and delivering content. Increasingly, however, libraries may determine that, even with the help of AI, a service or workflow has been ephemeralized as much as it can be. At that point, librarians may recall the words of economist Herb Stein: “If something cannot go on forever, it will stop.” The trick for libraries, faced with doing less with less, will be figuring out what to stop doing.

THE VALUE OF JOURNALS

A journal’s prestige, number of cites, price, cost per use, and relevance to local users all play a role in determining a journal’s value to the institution. Table 8 examines the relationship between price and value by grouping serial titles indexed in the Scopus, Clarivate, and EBSCOhost indexes into price bands, and then determining the average Impact Factor and Eigenfactor for all titles in those bands. This year, the average price ($10,830) for the most expensive journals was 57 times higher than the average price for least expensive ($189) journals, while the Impact Factor only increased by 2.7. The chart accompanying Table 8 provides a graphical view of the correlation between journal price and value: while all value metrics tend to increase with the increase in price, the rise in average price (represented by the blue line) is not accompanied by a corresponding rise in Impact Factor (represented by the burgundy line). 

Table 9 reviews the ratio of citations to periodical costs by subject, as well as the percentage of citable gold OA articles. While average prices for STM journals tend to be high in comparison to other subjects, this scenario changes if the costs are divided by the numbers of citations. Journals in chemistry and physics continue to have the highest average prices but sit in the middle of the pack when it comes to cost per citation. If citations are considered an indicator of value, then chemistry and physics journals, despite their high average prices, are still relatively high-value journals. Conversely, journals in philosophy and religion, music, and history, while less expensive, are also less frequently cited, so journals in those areas show the highest cost per citation. 

Commercial publishers continue to have higher per citation costs than other types of publishers. Commercial publishers showed a cost per citation of 29¢ and an average price of $2,711, while university presses showed 21¢ and an average price of $493, and societal publishers showed 9¢ and an average price of $923. But commercial publishers also continue to lead in the percentage of citable OA gold articles: approximately 23 percent of articles from commercial publishers were citable OA articles, compared to 18 percent for both university and society publishers.

METHODOLOGY

Survey tables are created by combining historical serials pricing information supplied by EBSCO Information Services with sets of serial titles from various indexes, including Scopus, Clarivate Citation Indexes, EBSCOhost Academic Search Ultimate, and EBSCOhost MasterFILE Complete. Reviewing pricing information for several sets of serial titles provides us with a better picture of how pricing has, on average, changed year over year, and predict where it might be for the 2027 renewal season.

Table 1 shows the average 2026 price for the top 15 scientific disciplines by using current year pricing information to determine the average price for a journal in a subject discipline, based on broad Library of Congress (LC) class.

Tables 2, 3, 4, and 10 are cost histories. Table 2 examines historical pricing costs for journals indexed in Scopus by eight broad LC subject areas. Table 3 examines historical pricing costs for online journals indexed in Clarivate Indexes by LC Subject. Prices are lower in this sample as many of the large publishers use custom quotes for online prices; consequently, some high-priced journals are not in the mix. Table 4 examines historical pricing costs for journals indexed in EBSCOhost Academic Search Ultimate by LC Subject. Table 10, which is new this year, provides a cost history for S2O titles handled by EBSCO Subscription Services. That the overall increase in 2026 pricing for S2O titles is higher than the overall increase seen last year for conventional subscriptions may reflect two things. First, the S2O title set used is very small—approximately 325 titles—so large price increases for a few titles will affect the overall total increase. Second, large swings may reflect that many S2O publishers have yet to pin down a pricing and customer retention strategy that is sustainable in the long term.

Tables 5, 6, and 7 show cost projections for 2027. Table 5 predicts the average price increase for periodical titles indexed in EBSCOhost MasterFILE Complete, a multidisciplinary database and index used by many public and school libraries. Although Table 5 averages numbers over three years, the sample size is still relatively small, meaning a large price change for a handful of titles impacts the overall price more than if a large sample size was used. Table 6 predicts the average price increases for periodical titles indexed in EBSCOhost Academic Search Ultimate, a multidisciplinary database and index used by many academic libraries. Table 7 predicts the average price increase for periodical titles indexed in the three discipline-based Clarivate Indexes (Arts & Humanities, Social Sciences, Science) for U.S and Non-U.S.; together, these tables provide guidance as to where we can expect prices to be for the 2027 renewal season.

Except for Tables 2 and 3, the price survey uses a print-preferred pricing model based on the standard retail price for the titles in the selected indexes. Despite the prevalence of online subscriptions, print pricing is still used for consistency because not all publishers make their online-only pricing available, or because they do not have a standard online-only retail price. The index contains pricing for print + online and online only if those were the only rates offered. 

Siôn Romaine is Director of Acquisitions & Rapid Cataloging Services, University of Washington Libraries. Cythnia M. Elliott is Collection Management Librarian, University of Arizona Libraries. Barbara Albee and Jamie Gieseck-Ashworth are Account Services Managers, EBSCO Information Services.