At 08:05:43 on 4 March 2022, the last successful internet measurement reached the Zaporizhzhia Nuclear Power Plant. Every other network in the district stayed online. Scroll to see what that looked like in the data.
KASPR Datahaus probes millions of individual internet endpoints around the world, continuously, at high frequency. Ukraine in February 2022 was no exception: hundreds of thousands of addresses answering — or failing to answer — around the clock.
Each measurement is small. Taken together, they are a live map of which infrastructure is still functioning and which is not.
On the left bank of the Dnipro, 200 km upstream from the Black Sea, sits a purpose-built company town of 50,000 people. Its reason for existing is the facility on its northern edge.
Within the district's address space, one narrow block of IPs belongs to the plant itself. That separation is what makes the next twelve days legible.
The Zaporizhzhia Nuclear Power Plant is the largest in Europe: six VVER-1000 reactors, 5.7 gigawatts, roughly a fifth of Ukraine's electricity.
In the early hours of 4 March 2022 it became the first operating nuclear power station in history to be captured by military force.
What follows is the raw measurement record for the Enerhodar district between 20 February and 7 March 2022. Nothing is smoothed, aggregated or modelled — every mark is a single probe that got an answer.
The upper band holds measurements to IP addresses in the Zaporizhzhia Nuclear Power Plant's own range — the facility's network, not the town's. Запорізька атомна електростанція.
The rhythm is unremarkable, which is the point. Day after day the plant answers, a few hundred successful probes per day, spread evenly across the clock. This is what normal looks like.
Beneath the plant band, the same probes to every other active IP in the surrounding district — homes, businesses, the town of Enerhodar, roughly fourteen thousand endpoints answering continuously.
Two populations, one method, one window. Whatever happens next, we can ask a question that a single time series cannot answer: did it happen to the facility, or did it happen to the region?
Russian forces took control of the site in the early hours of 4 March, after a night of shelling that set a training building alight. At 08:05:43 the last probe to a plant address came back. There was never another.
The white band carries straight through it. The town is still online, the district is still answering. The break is confined to one address block — a site-specific event, not a regional outage.
The district holds for two more days, then degrades and collapses. By 04:00 on 6 March the dense band has frayed to a scatter; by 13:00 almost nothing in the district answers at all.
The ordering matters. Had both failed together, this would read as a power or transit failure. Instead the facility went dark first and alone — and the region followed only once the occupation extended past the fence line.
Underneath, aggregate connectivity for the whole district against its own historical expected range. It tracks inside that band for the first ten days, slips below it on 2 March, and falls to zero on 6 March.
This is the view an operations desk sees: one line, one threshold, an alert the moment it breaks. The scatter above is the evidence; the index is the trigger.
The occupation of Zaporizhzhia was reported within hours. Most infrastructure shocks are not. What this record demonstrates is that the shock had a measurable digital signature — one that was available in real time, from outside the country, with no access to the site.
That distinction is what turns telemetry into intelligence. It rules out the mundane explanations — regional power loss, upstream transit failure, a submarine cable cut — and leaves the ones that matter.
We resolve the IP ranges belonging to strategic assets — power stations, ports, data centres, refineries — and monitor them as named entities rather than as anonymous traffic.
Every asset is scored against the region that surrounds it. The comparison separates a site-specific shock from a wider network event without waiting for confirmation on the ground.
Deviation from an asset's own historical envelope raises an alert in minutes, delivered to the desks that need it — energy and utilities, government and defence, risk and insurance.
Zaporizhzhia is an extreme case, and that is what makes it a useful one: the ground truth is unambiguous and independently documented, so the digital signal can be checked against it. The same method applies wherever a physical asset has a network footprint.
A substation that stops answering while its feeder region does not. A port whose terminal operations network degrades hours before a berth closure is announced. A data centre whose reachability fractures ahead of any status-page update. In each case the question is the same, and so is the way you answer it.
Internet connectivity was for a long time treated as plumbing — technical, uninteresting, someone else's problem. It is better understood as a continuously updating record of which infrastructure is still working, available to anyone with the measurement capacity to collect it.
Tell us which facilities matter to you. We will show you what their digital footprint has looked like for the past twelve months — and what it looks like right now.
Request a Briefing More Case Studies →KASPR Datahaus · Critical Infrastructure Monitoring. Analysis by Simon Angus.
Satellite imagery and event timeline reproduced for illustrative purposes. Event times in UTC.