# Ukrainian drones knock out three Yandex data centers in four days, disrupting Russia's largest cloud

The Sasovo site, now fully shut down, held two Nvidia A100 supercomputers used to train YandexGPT, and Yandex is moving cloud customers to rival Russian providers.

- Published: 2026-10-12T06:16:38.065Z
- Canonical: https://polylog.news/ai/2026-10-12/ukrainian-drones-knock-out-three-yandex-data-centers-in-four
- Publisher: Polylog (AI desk)
- Section: geopolitics
- Sources: [Polylog editors](https://polylog.news), [Hacker News / Cloudflare Radar](https://radar.cloudflare.com/bots/as13238?dateRange=7d#bot-traffic), [Al Jazeera](https://www.aljazeera.com/news/2026/10/11/russias-yandex-says-data-centre-in-vladimir-shut-down-after-drone-attack), [Meduza](https://meduza.io/en/feature/2026/10/08/ukraine-strikes-yandex-data-center-housing-two-of-the-company-s-supercomputers), [Yandex](https://yandex.com/company/news/2026-10-09)

Yandex said on Sunday that a drone attack had [completely shut down its data center in the Vladimir region](https://www.aljazeera.com/news/2026/10/11/russias-yandex-says-data-centre-in-vladimir-shut-down-after-drone-attack). It is the third major Yandex facility taken offline since October 8. On the night of October 7 to 8, a drone strike started a fire at the company's site in [Sasovo, in the Ryazan region](https://www.themoscowtimes.com/2026/10/08/ukrainian-drones-hit-yandex-data-center-after-wave-of-russian-attacks-a93915), and that site has stopped operating entirely. A strike on Kaluga on October 9 [disabled several modules there](https://www.kyivpost.com/post/86594). According to [Meduza](https://meduza.io/en/feature/2026/10/08/ukraine-strikes-yandex-data-center-housing-two-of-the-company-s-supercomputers), the Sasovo hub held tens of thousands of servers and two of Yandex's A100-based supercomputers, which the company uses to train its YandexGPT models.

The outage reached developers directly. The Ukrainian outlet [UNN reported](https://unn.ua/en/news/russia-reports-a-major-yandex-cloud-outage-following-an-attack-on-a-yandex-data-center) that more than 80 Yandex Cloud services failed, among them compute, managed databases, storage and the YandexGPT API. Russian-language reports covered the same failures. [Cloudflare Radar](https://radar.cloudflare.com/bots/as13238?dateRange=7d#bot-traffic) data shared on Hacker News shows bot traffic from Yandex's autonomous system down about 99 percent over the past week. That drop implies the company's search crawler has largely stopped running.

A widely read Russian AI channel [explained why recovery is slow](https://t.me/ai_machinelearning_big_data/11110). Restoring service after several large data centers are damaged at once is much harder than shifting load to backup servers, the channel wrote, because Yandex's components are tightly coupled and serve millions of users. The underlying problem is familiar to engineers: geo-redundancy plans usually assume that one region fails. They rarely assume that three sites fail within a few days. The Moscow Times reported that the first strike came [after a wave of Russian attacks](https://www.themoscowtimes.com/2026/10/08/ukrainian-drones-hit-yandex-data-center-after-wave-of-russian-attacks-a93915) on Ukraine. Ukrainian outlets such as [United24](https://united24media.com/war-in-ukraine/russias-yandex-under-fire-as-drones-hit-third-data-center-in-four-days-23321) describe the campaign as a deliberate effort to target Russian digital infrastructure. Russian officials have not publicly explained the targeting, and claims from either side in wartime deserve caution.

## What this means

Ukraine has added commercial AI compute to the list of strategic targets it can reach with cheap, long-range drones. Yandex absorbs the direct cost through lost capacity, migration concessions and new spending on physical defense. Russia's sovereign AI effort, which depends on a small number of large domestic GPU clusters that sanctions make hard to replace, suffers the deeper loss. The closest precedent is Ukraine's refinery campaign of 2024 and 2025. Those strikes continued for months, forced Russia to restrict gasoline exports and pulled air defenses away from the front. The campaign undermines the desk's thesis that Russia dominates its neighborhood, because Moscow cannot protect high-value civilian assets 300 kilometers from its capital. It supports the thesis that compute is now a strategic asset, since Kyiv is treating it as one. Markets have priced in the idea of AI as a contest over chips. They have not yet priced in physical attack risk for data centers in or near conflict zones. The likelier outcome is more strikes on Russian data centers within weeks, followed by Russian strikes on Ukrainian telecom and data infrastructure. That pattern depends on Ukraine maintaining its supply of long-range drones. A negotiated pause in deep strikes would prove this read wrong. For Russian users of Yandex Cloud, the case for spreading workloads across several providers now outweighs the cost of doing so.

## What to watch

- Whether Yandex discloses how much training capacity it lost and whether it delays upcoming YandexGPT releases. A delay would show how dependent Russian AI is on a few concentrated GPU sites.
- Whether Selectel, VK Cloud and K2Cloud report sharp increases in new customers. A lasting migration would redistribute Russian cloud market share for years.
- Russian strikes on Ukrainian data centers or telecom hubs in the coming weeks. These would turn a single campaign into a mutual exchange of attacks on compute infrastructure.