Ukrainian Drones Disable Yandex AI Data Centers in Russia

Ukrainian drones disable Yandex data centers

Ukrainian drone strikes have disabled two of the five data centers operated by Yandex, the Russian technology giant often described as the country's equivalent to Google. The attacks occurred on October 8 and October 9 in the Sasovo and Kaluga regions, according to reporting from Reuters. Ukrainian President Volodymyr Zelenskyy characterized the operation as a symmetrical response to a Russian drone campaign that has targeted Ukrainian data centers since late September, disrupting mobile and internet services for thousands of civilians.

The Sasovo facility houses two of the three supercomputers Yandex uses to train its YandexGPT large language model, making the strike a direct blow to the company's artificial intelligence ambitions. The Kaluga site, meanwhile, represents Yandex's largest data center. The Ukrainian firm Firepoint confirmed that its FP-1 one-way attack drones participated in the Kaluga operation. Yandex acknowledged service disruptions but reported no injuries to personnel.

The impact rippled well beyond Yandex's own platforms. Russian users reported outages affecting the Ivi video streaming service, the digital banking platform T-Bank, and the state-owned Russian Railways system. Additional disruptions hit a dating application, a real estate aggregator, a self-publishing platform for books, and the website of the most successful club in the Russian Premier League, according to Reuters and the independent Russian outlet Meduza.

Data centers become wartime targets

This escalation marks a significant development in the growing trend of data centers becoming legitimate wartime targets. In February 2026, following U.S. and Israeli strikes on Iran, Iranian drones disabled multiple Amazon data centers in Bahrain and the United Arab Emirates. Russia has conducted sustained missile and drone attacks against Ukrainian energy infrastructure, transportation hubs, and residential areas since the full-scale invasion began in February 2022. Ukraine, facing shortages of interceptor weapons, has steadily expanded its own long-range drone campaign against Russian military bases, supply lines, shipping in the Sea of Azov and Black Sea, and energy infrastructure such as oil refineries. Earlier in 2026, Ukrainian drones struck warehouses belonging to Wildberries and Ozon, Russia's primary e-commerce platforms.

The targeting of AI-specific infrastructure introduces new considerations for anyone building or deploying machine learning systems at scale. The supercomputers housed in the Sasovo center represent concentrated, irreplaceable compute capacity for training frontier models. Their destruction or degradation cannot be quickly remedied by spinning up cloud instances elsewhere, given export controls on advanced semiconductors and the specialized cooling and power infrastructure such clusters require. Organizations that rely on a single geographic region for model training or inference now face a demonstrated risk that physical infrastructure can be eliminated by state actors with relatively inexpensive drone technology.

AI infrastructure geographic redundancy

For companies developing AI products, the incident underscores the importance of geographic redundancy for critical compute workloads. Training runs that span months and cost tens of millions of dollars can be interrupted or lost entirely if the underlying hardware is physically destroyed. Checkpointing strategies, distributed training across multiple sites, and disaster recovery plans that account for kinetic threats, not just hardware failures or natural disasters, are becoming necessary components of responsible AI engineering.

Enterprises consuming AI services should evaluate the geographic concentration of their providers' infrastructure. A model hosted in a single data center campus, or even a single national jurisdiction, carries heightened operational risk in an era where data centers are strategic targets. Contracts and service level agreements may need to reflect the possibility of prolonged outages caused by conflict rather than technical faults.

Civilian AI infrastructure legal protections

Policymakers and standards bodies will likely face pressure to define protections for civilian AI infrastructure under international humanitarian law, though the dual-use nature of large-scale compute, serving both commercial and military applications, complicates any legal framework. In the meantime, the practical lesson for builders is clear: the physical layer of the AI stack is no longer a background assumption. It is a vulnerability that must be designed around.

Source: Ars Technica