A third Yandex data center shutdown has taken another major Russian computing site offline after a drone attack damaged the company’s Vladimir facility on October 11. Yandex said no employees were injured, but the center stopped operating and some services became unavailable.

 

The incident followed attacks on Yandex facilities in Sasovo and Kaluga on October 8 and 9. The sequence turns cloud infrastructure and AI computing capacity into direct operational risks for a company whose consumer services and business platforms depend on a limited number of large sites.

 

Yandex has confirmed three separate disruptions:

  • Sasovo was seriously damaged and fully stopped.
  • Part of the Kaluga infrastructure failed.
  • The Vladimir data center was shut down completely.

 

Related Research

 

Yandex Data Center Shutdown Reaches Vladimir

Yandex updated its company notice on October 11 to confirm that the Vladimir site had been attacked overnight. The company said the facility’s work had stopped and that emergency teams and its own specialists were dealing with the consequences.

 

The official Yandex Cloud incident page separately reported damage to the Vladimir infrastructure after an unmanned-aerial-vehicle attack. Reuters confirmed that the outage made some Yandex services unavailable, although neither source disclosed the physical scale of the damage or a restoration timetable.

 

That uncertainty matters because Vladimir was not an isolated failure. Yandex said its Sasovo center had been seriously damaged three days earlier and had stopped operating. The Kaluga attack then knocked out part of a second facility before Vladimir became the third affected site.

 

Yandex reported no injuries at any of the three sites. That limits the confirmed human impact, but it does not resolve the operational consequences for customers or the replacement cost of damaged computing equipment.

 

Sasovo Damage Puts AI Computing in Focus

The Sasovo center links the attacks directly to Yandex’s AI capacity. Reuters reported that the facility housed two of the company’s three supercomputers used to develop its AI model, built around Nvidia A100 processors. Yandex has not said whether those systems were damaged.

 

The company also has not disclosed whether model training has stopped or shifted to other infrastructure. Any claim that YandexGPT development has been disabled would therefore go beyond the available evidence. The confirmed fact is that a center containing critical AI hardware remains offline while specialists assess whether its equipment can be restored.

 

Training systems are harder to relocate than ordinary web workloads. Large clusters depend on dense accelerator connections, high-capacity storage, specialized cooling and stable power. Even when data is backed up, rebuilding equivalent compute elsewhere can take far longer than moving a conventional application to another server region.

 

Cloud Customers Face a Concentration Test

Yandex supports navigation, music, video, food ordering and search products used by millions of people. Its cloud platform also hosts business applications and data. A disruption can therefore spread from Yandex’s own services to companies that depend on its availability zones, databases and recovery tools.

 

Reuters reported that one application operator could not reach its backup database in Yandex Cloud after the earlier attacks. Yandex responded by arranging accelerated transfers with other cloud providers, an emergency measure that can reduce downtime but also reveals the limits of keeping primary and recovery systems inside one provider’s regional footprint.

 

Cloud resilience is usually designed around equipment failure, software faults and localized utility outages. Three physical attacks in four days present a different scenario: several facilities may become unavailable before operators can finish shifting workloads from the first one.

 

The episode strengthens the case for geographic separation, independent backup credentials and recovery copies that remain reachable when a primary cloud control plane is impaired. For AI operators, it also raises the cost of concentrating scarce accelerators in a small number of highly visible locations.

 

Restoration Timelines Remain Unclear

Yandex said it was still evaluating the condition of its infrastructure and could not give a reliable schedule for restoring equipment. Its immediate priorities are maintaining services, supporting affected customers and determining which hardware can return to operation.

 

The next verifiable milestones will be service-status updates, confirmation that the Vladimir and Kaluga sites can resume work, and a damage assessment for Sasovo’s computing equipment. Any disclosure about the two AI supercomputers would clarify whether this remains primarily a cloud-availability crisis or has become a longer-term constraint on Yandex’s model development.