In the early hours of 8 October, a drone strike set fire to a Yandex data centre in Sasovo, in Russia’s Ryazan region. Yandex said the facility, one of its five major data centres, had stopped operating. The company was still assessing the damage and could not confirm whether the equipment could be restored.
The consequences extended beyond a building full of servers. Yandex operates services used for cloud computing, file storage, documents, search and other everyday digital functions. An outage at a single facility can travel through the systems built on top of it, affecting customers who may have no idea where their data is physically stored.
According to the Sasovo city administration, Yandex had invested roughly six billion rubles in the facility over eleven years. The exact financial loss from the strike remains a separate question. The more consequential point is that a physical attack on one site can disrupt services far beyond its perimeter.
Whether Ukraine deliberately selected the data centre or struck the industrial site around it remains unclear. Kyiv has not publicly provided a detailed account of the operation. President Volodymyr Zelenskyy said Ukraine was responding in kind to Russian attacks on Ukrainian data centres.
A warhead that can burn a factory can burn the server hall inside it. The people planning the next strike know that now.
Russia and Ukraine have been expanding their attacks on each other’s infrastructure. In September, Russian strikes damaged Ukrainian data centres and telecommunications facilities, including sites associated with MiroHost, Parkovy, Data Group, Utels and Pautina. Ukrainian operator Datagroup confirmed damage to its Kyiv data centre, while other providers reported disruption to their networks. In some cases, customer services continued operating because backup systems were available elsewhere.
Moscow claimed that Ukrainian data centres were being used to process or transmit intelligence for the armed forces and warned Ukraine of a “digital detox”. Such claims require evidence, which was not publicly supplied in the reporting cited here. But the underlying argument is familiar: if infrastructure serves a military purpose, an attacker may argue that it has become a legitimate target.
That argument becomes particularly difficult when a single facility serves several kinds of customer. The same data centre might host a government department, a bank, an online retailer and a logistics company. An attacker may describe the strike as a blow against military capability. The civilians whose payments fail or whose services disappear may have been nowhere near the military operation being invoked to justify it.
The problem is not confined to Eastern Europe. In March 2026, drone strikes damaged Amazon Web Services infrastructure in the United Arab Emirates and affected a facility in Bahrain. AWS reported physical damage, power and connectivity problems, and significant impairment to two of the UAE region’s three availability zones. Customers experienced disruption to cloud services, and recovery was expected to take time.
The Associated Press reported on the damage to AWS facilities, which illustrated how quickly a military conflict can become a problem for commercial cloud customers. Businesses and public institutions may buy a service that appears virtual and borderless, but the machinery that delivers it still occupies physical buildings with power supplies, cooling systems, fibre connections and roofs that can be damaged.
Governments have spent years treating cloud migration as a procurement decision. Which provider offers the best price? Which has the strongest security certifications? How quickly can a department move its systems? These are reasonable questions, but they do not settle the question of what happens when the building hosting those systems becomes a target.
Iran’s strikes on AWS infrastructure made that question impossible to ignore. The reported military rationale for targeting the Bahrain facility was disputed and should not be treated as independently established fact. The damage itself, however, was publicly acknowledged by AWS. Governments that have moved essential services onto commercial cloud platforms now have a practical reason to examine their assumptions about resilience.
The engineering principle behind cloud redundancy is straightforward. Providers divide infrastructure into separate availability zones so that a failure in one location need not take down an entire region. These zones are designed to withstand failures involving equipment, power, cooling and connectivity. Their usefulness depends, however, on the nature and scale of the incident, the independence of the sites and whether workloads can actually move elsewhere.
A military attack can expose weaknesses that ordinary disaster planning does not fully address. Several sites may be hit within a short period. Power and communications infrastructure beyond the data centre may also be damaged. Staff may be unable to reach facilities safely. Recovery capacity may be constrained by the same conflict that caused the outage.
Yandex customers face a related question. The company has offered redundancy and backup options, but the protection customers receive depends on how their particular systems are configured and where their copies of data are stored. Redundancy within a region can be valuable against an equipment failure or local accident. It is less reassuring when multiple facilities in that region are exposed to the same military campaign.
This is where the marketing language around cloud resilience needs to meet the physical geography. A customer can have several copies of a database and still be in trouble if those copies depend on the same electricity network, communications links, administrative controls or vulnerable region. A backup that cannot be restored in time is a comforting line on a contract, not a functioning recovery plan.
My view is that governments are behind on this because they have treated digital infrastructure as a technical issue rather than a national security concern. Defence budgets account for airspace, borders, ports and energy facilities. Yet the shared buildings where public records, payment systems, communications and hospital services may reside have often been discussed as ordinary commercial infrastructure.
A country that cannot keep essential banking, health or government services available during a sustained attack has a security problem larger than the inconvenience of an app failing. The same is true when a disruption prevents emergency services from accessing records or businesses from processing payments. Digital resilience is part of the country’s ability to function under pressure.
Three changes follow. First, governments should distinguish critical sovereign and military workloads from ordinary consumer services, with separation proportionate to the risks involved. That does not mean every government system must have its own building. It means a provider and a government should understand whether a disruption affecting one class of workload could disable the other, and design accordingly.
Second, critical systems need geographic redundancy that crosses meaningful risk boundaries. For some services, that may mean separate regions in different countries, independent connectivity and backups that are protected against compromise or destruction at the primary site. The right arrangement will vary, but it must be tested through actual recovery exercises rather than assumed to work because a contract says “high availability”.
Third, governments and legal experts need to confront the status of commercial digital infrastructure during armed conflict. A data centre can host military workloads and civilian services simultaneously. That does not automatically settle whether the facility is a lawful military objective, nor does a belligerent’s public assertion settle the matter. The applicable law of armed conflict, including the principles of distinction, proportionality and precautions in attack, remains relevant. Civilian harm cannot be dismissed simply because a server building has some military connection.
Commercial providers also have responsibilities. They should be candid with customers about the geographic concentration of their workloads, the limits of regional redundancy and the conditions under which recovery might fail. Governments, in turn, should know which essential services depend on which providers and what the fallback arrangements actually are.
The cloud is not above the battlefield. It is a collection of buildings, cables, power systems and people, and a determined attacker can reach them.
Until governments plan for that reality, data centres will remain critical infrastructure that society depends on but often treats as though its physical location does not matter. Sasovo and the Gulf have demonstrated the cost of that assumption. The next strike should not be required to teach the lesson again.
Be First to Comment