For years, Microsoft 365 storage for universities was unlimited and free. That changed this year. Microsoft capped Education tenants at 100 terabytes plus 50 to 100 gigabytes per licensed user, and turned on overage pricing worth roughly $360,000 per petabyte per year once that quota is reached. Universities holding two or three petabytes of data, which is common at large research institutions, are looking at bills in the hundreds of thousands to over a million dollars a year for storage they never budgeted for.
"Shock. Horror. Panic." That's how Jason Lohrey, CEO and founder of data management company Arcitecta, describes the week an IT department first sees that number. "They are powerless and wedged unless they can get their data out in time," he says. Some are considering deleting data in bulk just to avoid the cost, a move he calls risky given how much of that data is research output, intellectual property, or records under retention rules.
The math gets worse before it gets better. Microsoft also throttles how fast data can leave a tenant, at 400 gigabytes an hour, tenant-wide, across every backup tool and application token combined. At that rate, moving one petabyte out takes roughly 105 days. Universities holding several petabytes are looking at the better part of a year, and overage pricing starts hitting renewals from December 2026 onward. For institutions just now realizing the size of the problem, Lohrey is blunt: "it's likely too late for many to escape unscathed." A second clock is running underneath that one. Microsoft moves accounts inactive for 12 months to read-only, archives them soon after, and permanently deletes them starting mid-2027, putting former students' and staff's research data at risk right along with them.
This is the backdrop for Mediaflux Connect 365, Arcitecta's new platform for managing OneDrive, SharePoint, and Teams data, announced August 25. The announcement itself is less interesting than what it implies operationally, and why the same cost pressure now forcing this issue also happens to be pushing university data toward being usable for AI.
Lohrey draws a direct comparison to self-managed storage gone wrong: data goes in, nobody looks at it again, and years later nobody remembers what's there or why they're still paying for it. "Active data management" is meant to be the opposite: a system that keeps evaluating data against a goal, like cost, and acts on it automatically. Data past its useful life gets flagged for deletion. Data nobody has touched in months moves to cheaper storage on its own. "Active systems will use business rules and AI to ensure the value embodied in the data is realised," Lohrey says.
Doing that without breaking things for end users starts with a scan cross-referenced against Active Directory or LDAP, to find accounts belonging to people who've already left the university. Those move first, since nobody's coming back Monday morning to look for that data. Everything that moves stays reachable: if a drive gets relocated and someone still needs it, the data isn't gone, it's just sitting in Mediaflux instead of OneDrive.
The 400GB-an-hour cap doesn't get solved by moving data faster. It gets solved by moving less of it. Lohrey describes prioritizing what's actually worth the bandwidth: skip duplicate copies, skip anything that's easily regenerated, move what matters most first. "It's easier to be intelligent when you understand what you are dealing with," he says, which is why the analysis runs before anything else moves.
That analysis step is also where the AI story shows up, and Lohrey is candid that cost, not AI ambition, is what actually got universities to pay attention. "In reality, it's cost alone that appears to have been the catalyst," he says. "Embracing the opportunity, we have introduced the perfect triple: cost reduction, data management, and AI." In practice, the same metadata used to decide where a file should live also gets indexed as searchable vectors inside XODB, Arcitecta's own database engine, built specifically to handle billions of files at petabyte scale. A dataset that gets pulled off OneDrive to cut costs ends up more searchable, and more usable for AI workloads, than it ever was sitting in a SharePoint folder.
None of this is unique to Microsoft, and Lohrey doesn't pretend otherwise. Google made a similar move on free storage years ago. "Anything that looks 'free' is not free," he says. "It's a hook to get you locked in." Arcitecta says the reactions it's hearing from prospects range from resigned to something closer to relief. One university, Lohrey says, told them this was "a chance to get out of Microsoft" entirely.
For IT teams watching this play out at their own institution, the real lesson isn't about Microsoft 365 licensing specifically. It's a prompt to ask a more basic question about every platform holding institutional data: who actually controls it, and what happens the next time the terms change.