Drone strikes hit Yandex data centers, disrupting Russia’s AI infrastructure

Introduction

A Ukrainian drone campaign has brought one of Russia’s largest technology companies directly into the expanding digital front of the war. According to reporting cited by Ars Technica AI, including Reuters, drones struck Yandex data centers in Russia’s Sasovo and Kaluga regions on October 8 and October 9. Yandex is often described as “Russia’s Google” because it operates the country’s most popular search engine and provides a broad collection of online, cloud, and artificial intelligence services.

Key points

• Two of Yandex’s five data centers were hit in the reported strikes.
• The Kaluga facility is described as Yandex’s largest data center.
• The Sasovo site houses two of the company’s three supercomputers used to train YandexGPT.
• Yandex said no employees were injured, while acknowledging interruptions to some services.
• Russian users reported access problems involving media websites, streaming, banking, rail, real estate, dating, and other applications.

The Ukrainian company Firepoint said its FP-1 one-way attack drones were involved in the strike on the Kaluga site. Ukrainian President Volodymyr Zelenskyy characterized the operation as a symmetrical response to Russian attacks on data centers and other infrastructure in Ukraine. The available reporting does not establish the precise condition of the affected supercomputers, nor does it show that YandexGPT was destroyed or permanently taken offline. The clearest conclusion is that important facilities were damaged and service continuity was affected.

Why a data center matters

A data center is more than a warehouse for websites. The same physical and network infrastructure can support search, payments, transportation services, media distribution, enterprise applications, and AI workloads. Damage to power systems, cooling, connectivity, or compute clusters can therefore produce cascading failures across services that appear unrelated to one another.

The reported outages illustrate that concentration. Users experienced problems with a video platform, a digital bank, a railway service, and several consumer applications. The incident does not necessarily mean that each service was physically located in the damaged facilities. It does show how dependent online platforms can be on shared infrastructure, regional networking, and common cloud or hosting providers.

For an AI company, high-performance training infrastructure is particularly difficult to replace quickly. Training a large model depends on specialized processors, data pipelines, storage, networking, and reliable power. Even if model files remain intact, damage to a training cluster can delay new versions, interrupt experiments, or reduce the capacity available for user-facing inference. The source material does not provide enough information to determine whether YandexGPT training stopped or how large the long-term loss may be.

A widening class of wartime targets

The strikes show how the definition of critical infrastructure is expanding. Russia has repeatedly attacked Ukrainian cities, energy systems, transportation networks, and data centers since its full-scale invasion began. Ukraine has responded with longer-range drone operations aimed at military sites, supply routes, energy facilities, and other Russian-linked infrastructure. The reported Yandex strikes fit into that broader escalation.

For technology companies, resilience may increasingly require geographic redundancy, distributed computing, independent backup systems, and stronger separation between critical workloads. These measures are often framed as reliability or cybersecurity investments; in an active conflict, they become physical security requirements as well.

The consequences also extend beyond the companies involved. Disruptions to banking, railway, media, and public-facing applications can affect ordinary users and essential economic activity. The Yandex case does not prove that an AI system can be easily disabled by a single strike, but it demonstrates that companies operating large compute clusters are now strategic infrastructure operators. Their facilities connect artificial intelligence to electricity, telecommunications, logistics, and geopolitics in ways that are increasingly difficult to separate.

Source: Ars Technica AI