Nobody Cut This

Report
Written By Arthur Knott


Published 31 August 2026. iDigBio's media servers at the University of Florida are scheduled to shut down on 31 August 2026. Status verified 28 August 2026; no extension had been announced. This report will be updated on 1 September with what actually happened.
cell phone with GBIF landing page sits on a bed of moss next to a stone path in a garden, dark leaves silhouetted in the background.

The third grant phase was written in 2021 with an end date on it, and the end date is arriving on schedule. What breaks is a storage layer that a great many institutions treated as permanent because it kept working.

On 31 August 2026, iDigBio shuts down its media servers at the University of Florida. Image files physically stored on those servers are deleted. Any specimen record whose image link contains iDigBio in the path stops resolving, which means GBIF and other aggregators may show you the record and not the picture.

Read quickly, that is a national biodiversity database going dark. It isn't. The accurate version is duller and more useful: a grant-funded project became part of the working infrastructure of American science while one layer of its storage stayed explicitly temporary, and that layer is now being switched off on the schedule it always had.

What is actually at risk

Only files sitting on the retiring Florida servers, and records still pointing at them.

A collection has four possible positions and only one of them survives. Files still on iDigBio's servers are deleted. Files moved elsewhere with the links not updated become unreachable. Files moved and relinked but not republished to the aggregators also become unreachable. Files moved, relinked, and republished come through intact. iDigBio has been explicit that it cannot perform any of those steps on a collection's behalf.

The University of Kansas team supporting Symbiota portals has said it will attempt transfers and link repair where it can. Everyone else reviews their own records.

What is not at risk

GBIF is not shutting down. iDigBio reports that more than 98 percent of its data providers are now registered with and serving specimen data through GBIF. A new GBIF-hosted portal is intended to become iDigBio's public face, and the legacy portal and API are expected to stay up through 2026, though data ingestion stopped in May.

The number circulating is 65 million images. That figure describes media records reachable through the portal, not files verified as sitting on the servers being retired. iDigBio has not published a total for the affected subset. The immediate loss is probably narrower than the number implies. The structural problem is considerably wider.

Follow the money, because the money explains the date

The central project took roughly $45.5 million from the National Science Foundation across three five-year phases: $10 million in 2011 to stand up the coordinating hub at the University of Florida with Florida State, $15.5 million in 2016, and just under $20 million in 2021 for a third phase aimed at sustaining what had been built. None of that includes the separate awards to Thematic Collections Networks that paid institutions to digitize particular groups and push the data into the system.

iDigBio holds a no-cost extension that carries some work into 2027. A no-cost extension adds time, not money — the recipient keeps spending funds already awarded. The servers are still going off on the 31st.

This is the part worth being careful about. The phase that ends now was written in 2021 with a defined duration. Nothing was cancelled. If you want a villain, the honest answer is that there isn't one available on the current evidence, and inventing one would make the actual problem harder to see.

The political context, stated precisely

The timing is charged. The president's fiscal year 2026 request proposed $3.9 billion for NSF against the agency's $9.06 billion enacted level for fiscal 2025. Congressional committees then considered substantially higher figures, which is the reminder that a presidential request is not an appropriation.

So: it is accurate that this transition is happening during severe pressure on federal science budgets. It is not accurate, on the evidence available, to say new budget cuts caused the 31 August shutdown. Both of those sentences are true at once, and a report that prints only the first one is doing advocacy.

The thing that should worry you

Grants are supposed to end. That is what makes them grants.

The problem is that a project became infrastructure without anyone establishing permanent stewardship for every layer that access depended on. A biodiversity database is not a server. It is a working relationship among physical specimens, collection staff, digitizers, standards, file hosts, stable identifiers, aggregators, and public money, and no single component holds the network up by itself. When a layer everybody understood to be provisional turns out to be load-bearing, removing it propagates outward as broken links across systems that are otherwise running perfectly.

The specimens are all still there. Nothing in a cabinet in Gainesville or Lawrence or Berkeley has moved. What is at stake is the layer that decided who could look at them without booking a flight, and that layer was funded on a five-year clock while the collections it served operate on a scale of centuries. That mismatch is the story, and it does not resolve on the 31st — it recurs every time an award with a terminal date starts producing permanent expectations.

If you publish to GBIF, the concrete task is querying the GBIF API for records containing iDigBio addresses, filtered by your publishing organization. iDigBio's notice walks through it and warns that the search throws false positives, so verify against your collections management system rather than the query alone. The searchable terms for the next layer are stable identifier, persistent identifier, and digital preservation, and the question underneath all three is not who hosts a file but who has agreed to keep hosting it after the money stops.

Sources


Arthur Knott
Previous

The 1:4 Rule
Next

What Synergetics Explains, and What It Only Describes