Publication date: 11 October 2026
A drone attack that shut a Yandex data centre in Russia’s Vladimir region has sharpened a difficult reality for technology leaders: digital infrastructure is now part of the physical battlespace. Yandex said the facility’s infrastructure was damaged and operations stopped, with some services unavailable. Reuters reported that no one was injured.
The 11 October incident followed attacks on Yandex facilities in Sasovo and the Kaluga region over the preceding days. In a company statement published on 9 October, Yandex said its Sasovo data centre had been seriously damaged in an overnight drone attack. Reuters reported that the Sasovo site housed two of the company’s three supercomputers used to train its YandexGPT large language model.
The pattern matters more than any single outage. Data centres support cloud applications, communications, business systems and increasingly national AI capability. They also depend on concentrated supplies of electricity, cooling, network connectivity and specialised hardware. This makes them both economically valuable and physically vulnerable.
AI capacity is becoming strategic infrastructure
Governments have long treated energy networks and telecommunications as strategic assets. Large AI data centres now belong in the same discussion. Advanced chips are difficult to replace, model training can require long uninterrupted computing runs, and service disruptions can spread beyond the facility that was struck.
The Yandex incidents also blur traditional categories. A data centre may serve civilian customers while supporting technologies considered important to national power. That dual-use character complicates questions about targeting, proportionality and the downstream effect on people and organisations that have no role in a conflict.
For technology companies, conventional disaster recovery is no longer enough. A backup in a nearby region may share the same threat environment, power grid or network route. Resilience plans need geographic separation, tested service failover, spare hardware strategies and the ability to operate with reduced computing capacity.
What leaders should do now
Boards and executives should identify which services depend on a single physical region, cloud control plane or scarce chip cluster. They should also distinguish between data redundancy and operational redundancy. Copies of data are valuable, but they do not keep a service running if replacement computing, identity systems or network access are unavailable.
Governments face a parallel task. Critical-infrastructure rules often focus on utilities, finance and telecommunications, while the systems that deliver AI and cloud services sit across several regulatory categories. Authorities need clearer thresholds for when large data centres should receive enhanced physical protection, incident-reporting duties and coordinated recovery support.
The immediate facts remain limited, and independent assessment of the damage is difficult during an active conflict. What is already clear is that computing capacity can no longer be treated as an abstract resource in “the cloud”. It is housed in identifiable buildings, connected by vulnerable infrastructure and increasingly central to economic and strategic power.