What Higher Ed's Cloud Storage Data Is Really Telling Us: 2026 Wasabi Global Cloud Storage Index

Higher education IT is used to doing a lot with not enough. These teams have big ambitions, complicated infrastructure, and budgets that get defended line-by-line in committee rooms. What is the role of cloud storage within this reality?

For the fourth consecutive year, independent analysts surveyed 1,700 IT decision-makers worldwide, including 241 from the education sector, to produce the 2026 Wasabi Global Cloud Storage Index. We recently released a report highlighting the higher ed numbers, which tell a story worth unpacking: persistent fee pressure, a dark data problem AI has made urgent, and a security posture that hasn't kept pace with the threat record. 

The fee problem isn't going away

The Cloud Storage Index found that more than half (54%) of education cloud storage spending goes toward fees, not storage capacity. The global average is 50%, so higher ed is already above the baseline. What’s even more telling, however, is the year-over-year trend: in 2025, 50% and in 2024, 49%. Not only is this trend consistent, but it’s actually worsening over time, with more and more of cloud budgets being eaten up by fees.

Higher ed budgets aren't built for that kind of persistent drain. They're set annually, defended in committee, and not designed to absorb billing models that frequently extract value through egress charges, API calls, and retrieval fees. Nearly half (41%) education respondents say they exceeded their cloud storage budget last year, and 89% of those point to fees as at least part of the reason. That budget overage had to come from somewhere, and the effects can ripple outward fast. A deferred project, a research initiative that doesn't get funded this cycle, or an AI program that scaled back before it starts, all because of storage fees.

AI storage has its own fee dynamics, and higher ed is feeling those too. Thirty-nine percent of education respondents cited high or unpredictable costs as a top challenge for AI-related storage specifically.

Dark data has a new reason to matter

For most of cloud storage's history, dark data, which is the unstructured, unanalyzed data sitting largely untouched in storage, was more of an audit footnote than a strategic concern. Most institutions knew it was there, but they didn't have a compelling reason to do much with it.

AI changed that. The majority of education respondents estimate that between 25-74% of their stored capacity is dark data. They’re sitting on decades of research outputs, academic archives, lab datasets, student records, and institutional files that nobody's looked at recently but nobody’s deleted either. Now, AI can put all that information to work.

This isn't a small number of institutions: 89% of respondents say that operationalizing dark data is a priority, yet the same infrastructure friction driving higher ed budget overruns is now standing directly between institutions and their AI goals. The clearest sign of that tension is that 67 cents of every AI dollar in higher ed goes not toward AI itself, but toward the data and processing power needed to feed and run it.

When the majority of your investment is just keeping the lights on, it's no surprise that only 37% of AI projects are delivering positive ROI today, or that half the sector identifies data storage as the single biggest implementation challenge. The fee problem and the dark data problem aren't separate issues. They're hitting the same pressure point.

Ambitious and honest about the timeline

Higher ed institutions are committed to AI: 98% of education respondents have an AI infrastructure budget for the coming year and, 99% currently use or plan to use LLMs.

And yet the global market is projecting a 19-point jump in AI ROI over the next 12 months, while higher ed is only projecting 10. That gap is especially interesting when you take the context into account. Higher ed IT leaders know their environments well: they have research systems that don't talk to each other, compliance constraints on what data can move where, and legacy infrastructure that was not designed with AI pipelines in mind. They've seen enough initiatives stall out to be skeptical of 12-month turnarounds.

The conservative projection in the higher ed market doesn’t come from a lack of confidence in AI. It's likely an accurate read of how much infrastructure work stands between current state and an environment where AI can actually deliver.

Higher ed is starting from a stronger position than the global average: 37% of education AI projects are currently delivering positive ROI versus 32% globally. But in a sector where budgets are set annually and defended in committee, a 37% success rate is a harder sell than it might be elsewhere. When most AI projects aren't yet delivering returns, the institutions funding them don't have much room for error.

Security confidence doesn't match the threat record

Higher ed has long been an attractive target for cyberthreats. With open networks, valuable research data, student records, and security teams that are historically underfunded relative to all the data they're protecting, education data is vulnerable. The 2025 Microsoft Digital Defense Report puts some hard numbers on how exposed the sector is.

The report ranked research and academia third among all global sectors for cyber threats in the first half of 2025, behind only government agencies and IT companies. Education accounted for 39% of all identity compromise incidents Microsoft observed in that period, which is a disproportionate share of a threat type that surged 32% overall. Part of the reason is structural: higher ed environments have some of the largest and most complex identity systems, with decentralized IT, constant user turnover, and inconsistent MFA enforcement. The same data shows higher ed had the longest average threat actor dwell time of any industry, meaning attackers went undetected longer than anywhere else.

The 2026 CSI data reflects this same trend, given 44% of education respondents have already experienced a cyberattack that cost them access to public cloud data. Only 47% are completely confident they could keep data operational and unaltered after another one. And 41% believe their cloud vendor doesn't have the tools to adequately defend against an attack. Institutions ultimately manage to get through breaches, but the confidence gap that remains is significant.

There is one encouraging data point: immutability adoption climbed 14 percentage points since last year. Today, nearly 63% of education respondents are using Object Lock. Whatever is driving this increase (breach response, compliance pressure, vendor conversations), education is moving in the right direction by ensuring data cannot be edited, altered, or deleted for set periods of immutability.

The bottom line

The 2026 CSI data describes a sector in motion, but not yet in sync. Higher ed is committing real budget to AI, taking security more seriously than it did two years ago, and starting to treat dark data as an asset rather than an afterthought. The fundamentals are moving in the right direction.

What hasn't moved is the infrastructure layer underneath all of it. Fee structures that were already straining higher ed budgets are now compounding against AI cost pressures. Dark data that institutions want to operationalize is sitting behind the same friction that's causing budget overruns. Security confidence hasn't caught up to the threat record. These aren't separate problems. They're the same problem showing up in different places.

The institutions that close the AI ROI gap fastest won't necessarily be the ones with the biggest budgets. They'll be the ones that stopped treating infrastructure as a line item and started treating it as a decision.