# Data centres are now targets > Drone strikes on data centres in Russia and Ukraine show a cloud region is a building. AI in critical systems must keep running when one goes dark. - URL: https://www.sector88.co/blog/data-centres-are-now-targets - Author: Taskin Karlar - Published: 2026-10-09 - Updated: 2026-10-12 - Category: Analysis - Tags: critical-infrastructure, cloud-outages, resilience, edge-ai, sovereign-ai ## Quick answer Data centres have become physical targets in the war between Russia and Ukraine. On 8 October 2026 a drone strike took a Yandex Cloud availability zone offline, and it was still down at the end of the day. Failover kept running workloads alive, but creating new resources was limited. AI in critical systems needs to keep working, and stay under the operator's command, when a facility or link is lost. In the early hours of 8 October 2026, drones hit Yandex's data centre in Sasovo, south east of Moscow, one of the company's five large data centres. [Yandex said](https://www.straitstimes.com/world/europe/drones-hit-yandex-data-centre-in-first-major-attack-on-russian-data-hub) the attack caused a fire and that it fully suspended the centre's operations, and its cloud reported a power failure across an entire availability zone. By the end of the day, Yandex said it [could not yet confirm](https://www.straitstimes.com/world/europe/russias-yandex-says-unclear-whether-data-centre-can-be-restored-after-drone-strike) whether the equipment at Sasovo could be restored. **Update 12 Oct 2026:** ru-central1-b remains down with no restoration date, according to the latest [Yandex Cloud status update](https://status.yandex.cloud/en/incidents/2092). [Yandex has confirmed](https://www.channelnewsasia.com/business/ukraine-expands-drone-strikes-data-centres-owned-russias-yandex-6445596) a second data centre, in Kaluga, was hit on 9 Oct. Reuters called it the first major attack on a Russian data hub. It followed weeks of [strikes on data centres in Kyiv](https://www.reuters.com/business/aerospace-defense/russia-hits-ukraines-largest-mobile-provider-strikes-data-centres-2026-09-27/). In one September attack, about 100,000 households in Kyiv and the surrounding region lost internet, according to Ukraine's digital transformation ministry. Reuters notes that drone attacks by both sides have spread across the economy, from energy and ports to, most recently, data centres. Compute is now part of the battlefield. This post takes no side in the war. It is about what the infrastructure is telling everyone who runs AI in systems that cannot stop. ## Why it matters A cloud region is a set of buildings. Buildings need power, cooling, fibre and a roof, and every one of those can be hit. For twenty years the industry treated the loss of a whole facility as a rare disaster to plan around once. It is now a recurring event in at least one part of the world, and there is no reason to think it stays there. The Yandex incident also shows what failover does and does not buy you. According to the [Yandex Cloud status updates](https://status.yandex.cloud/en/incidents/2092), the provider told customers to move load to other zones, and by morning Moscow time said the switch was complete and services there were running normally. That is good engineering. But overnight it also warned that creating new virtual machines, Kubernetes clusters and databases could be limited even in the zones that were not hit, and it was still saying new resources could be limited that evening. By then it expected the zone to stay down for some time and told customers to use alternative recovery plans. Failover saves what is running. It does not save your ability to act. When the system needs to change, scale or recover, the parts that let you make those changes are often the first to choke. ## What it means for operators For a grid operator, a ground station, a port or a defence program, the question is no longer "do we have a backup region?" It is a harder set of questions: 1. **If the facility or the link is gone, does the AI at the site keep running?** If it depends on a round trip to a distant region, the answer is no. 2. **Who can still change it?** Updating a model, pausing a workload or rolling back should not depend on a control path that crosses the thing you just lost. 3. **Where is the record of what ran?** If the logs sit in the building that burned, you lose your account of what the system did at the worst possible moment. 4. **Have you tested the cut?** Most organisations discover their real dependency map during the outage. Break the network on purpose, on a schedule, and find out first. None of this means abandoning the cloud. It means placing each decision where it can still be made when the network or the building is gone. That is the argument we made about national control in [Location is not control](/blog/location-is-not-control-australias-ai-sovereignty-test), and it is why [sovereign AI](/blog/what-is-sovereign-ai) is about command and continuity, not postcodes. ## Where Sector88 fits Sector88 is one system to deploy, run and govern AI at every site, connected or completely cut off. The operator keeps command of what runs, and a durable record of what ran, at the edge and in the facility. We build it for the day the region goes dark. A region is a building, and buildings can be hit. Plan your AI accordingly. ## Frequently asked questions ### What happened to the Yandex data centre in Sasovo? In the early hours of 8 October 2026, drones hit Yandex's data centre in Sasovo, in Russia's Ryazan region. Yandex said the attack caused a fire and that it fully suspended the centre's operations. Yandex Cloud reported a power supply incident in its ru-central1-b availability zone and moved workloads to other zones. Late that day the zone was still unavailable, and Yandex said it could not yet confirm whether the equipment at Sasovo could be restored. ### Does multi-zone failover protect AI systems from a lost data centre? It protects what is already running, if the system was built for it. It does not guarantee the ability to act. During the Yandex incident, the provider warned that creating new virtual machines, Kubernetes clusters and databases could be limited even outside the affected zone. Capacity to change things is often the first casualty. ### How should critical infrastructure operators plan for losing a data centre? Assume any single building or link can be lost. Keep the AI that a site depends on running at that site, keep the decision path and the record of what ran under the operator's command, and test the cut on purpose so the first real loss is not the first test. ## Sources - [Drones hit Yandex data centre in first major attack on Russian data hub](https://www.straitstimes.com/world/europe/drones-hit-yandex-data-centre-in-first-major-attack-on-russian-data-hub), Reuters via The Straits Times - [Russia's Yandex says unclear whether data centre can be restored after drone strike](https://www.straitstimes.com/world/europe/russias-yandex-says-unclear-whether-data-centre-can-be-restored-after-drone-strike), Reuters via The Straits Times - [Yandex Cloud incident 2092: availability zone power failure](https://status.yandex.cloud/en/incidents/2092), Yandex Cloud - [Ukraine expands drone strikes on data centres owned by Russia's Yandex](https://www.channelnewsasia.com/business/ukraine-expands-drone-strikes-data-centres-owned-russias-yandex-6445596), Reuters via CNA - [Russia hits Ukraine's largest mobile provider, strikes data centres](https://www.reuters.com/business/aerospace-defense/russia-hits-ukraines-largest-mobile-provider-strikes-data-centres-2026-09-27/), Reuters - [Russia puts Ukraine's internet in its sights as attacks on Kyiv infrastructure intensify](https://kyivindependent.com/russias-latest-target-ukraines-internet/), The Kyiv Independent