Eight days after the first drone strike on a Yandex data center, the picture on October 16 is no longer one of widespread service failure. People can book taxis, run searches, send email and use some Yandex Pay payment methods again. But restoring the applications people use is not the same as restoring the physical infrastructure behind them. Yandex is still shifting workloads, restricting some operations and moving systems onto backup capacity.
The crucial distinction is between an app being available and a data center being back in operation. Critical services can sometimes be brought back within days using surviving hardware. Repairing a facility may require replacing power equipment, networking, servers and cooling systems. As of October 16, Yandex had not announced that either of its two fully shut-down sites had reopened, and it had given no date for a return to normal operations across all its services.
Three strikes, three data centers: what was damaged?
Overnight into October 8, a drone attack was followed by a fire at a Yandex data center in Sasovo, in Russia's Ryazan region. The company confirmed serious damage and a complete shutdown of the facility. On October 9, a data center in Kaluga was hit. Some of its infrastructure failed, but Yandex did not say the entire site had shut down. Overnight into October 11, an attack on the Vladimir data center forced that facility offline as well. Yandex said no employees were injured in the three incidents.
None of this establishes that every server at all three sites was destroyed. What has been publicly confirmed is the shutdown of two data centers, damage to part of a third and the resulting loss of access to computing infrastructure. Yandex has not disclosed how many servers were irreparably damaged, which racks survived or what proportion of the hardware is now usable. Public accounts also differ on the total number of Yandex data centers. Without a consistent inventory, assigning a precise percentage to the company's overall losses would be misleading.
FACTUM.PRESS reported the initial Sasovo fire and later documented the broader outages following the three attacks. The immediate question has shifted from how extensive the first wave of disruptions was to how long the company can operate on a reduced and reconfigured infrastructure footprint.
Which supercomputers were at risk?
According to Yandex's previously published technical information, two named supercomputers, Chervonenkis and Lyapunov, were based in Sasovo. The third, Galushkin, was located in Vladimir. They were designed for demanding computational work, particularly machine learning. The original published configurations contained 1,592 NVIDIA A100 accelerators in Chervonenkis, 1,096 in Lyapunov and 1,088 in Galushkin: a combined 3,776 GPUs.
That is a historical inventory of three identified machines, not an inventory of every GPU Yandex owned in October 2026. The company may have had other clusters and additional equipment at separate locations.
Specialists interviewed by T-invariant believe all three supercomputers suffered major damage and were taken out of operation. They highlight the threat posed by smoke to air-cooled servers: soot, moisture and combustion by-products can damage electronic boards and contacts even where flames never reached a rack. Their conclusions are expert assessments, however, not publicly released inspection reports. Yandex itself has not confirmed that all three machines were physically destroyed or published an itemized assessment of the hardware.
Our October 8 technical analysis examined the hypothetical cost of replacing the Sasovo computing systems. Those numbers are not an estimate of confirmed losses: that would require knowing which nodes, GPUs and associated systems can be reused. FACTUM.PRESS has also reported the later expert assessment of all three supercomputers.
What is working again, and what remains restricted?
Search, Maps and taxis. Core consumer functions are available again. Yandex Go is accepting rides, although RBC was still reporting isolated errors in the app on October 14. Search continued operating through much of the incident. Maps is accessible, but Yandex warned that some features would return gradually. Service availability in one city is not proof that individual users elsewhere are no longer experiencing problems.
Yandex Pay. By October 13, some users had regained access to account balances, transfers and payments. According to the company, payments using physical bank cards and payment stickers had been restored. Other transactions could still fail. Yandex said that account records were held in multiple copies and that it had found no loss of accounts holding money. That is the company's statement about data integrity, not an independent audit.
Mail, Disk and other Yandex 360 services. According to a detailed Yandex 360 update dated October 15, email is available, but users may have to refresh pages to see changes. Push notifications are not working, and certain mail-collection and draft functions remain limited. On Yandex Disk, automatic photo and video uploads and downloads of entire folders or albums are still disabled; some individual files may fail to open or download. Telemost still has restrictions on chat import and export, AI-generated meeting summaries and content previews. The company has not declared its entire 360 suite fully restored.
Alice and smart-home devices. Most functions of the Alice AI chat service, smart speakers and YaOS televisions were operational again, Yandex told RBC on October 14. Calls on smart speakers were not yet working, and problems remained with music playback across several devices, fairy-tale playback and the baby-monitor feature. Restoring an existing AI assistant does not establish that the large computing clusters used to train future models have recovered.
Yandex Direct. On October 15, Yandex reopened portions of its advertising dashboard. Advertisers can view settings and reports, switch existing campaigns on and off, add funds and adjust some payment settings. Creating new campaigns, making certain changes to active placements, using the API and working through Direct Commander remain unavailable. Conversion statistics may also be incomplete. Yandex acknowledged a separate attribution issue affecting conversions after October 11 associated with earlier ad clicks. Advertising impressions continued between October 11 and 14, even while campaign-management tools were inaccessible.
Yandex Business. In an October 16 announcement, Yandex said that access to the Business dashboard had been restored, but in read-only mode. Users still cannot edit business listings, upload photos, reply to reviews, add products or launch new advertising campaigns. Previously purchased promotion continues, while reporting may be incomplete. Saying simply that Yandex Business is “back” would therefore overstate the recovery.
Yandex Webmaster and Metrica. The industry publication SEO.RU reported on October 16 that the Webmaster dashboard had become accessible again on October 15. That is not confirmation that every function and data-processing workflow is operating normally. Although Metrica supplies analytics used in Direct reporting, we found no comprehensive public technical update demonstrating that Metrica's dashboard, data collection and historical reporting had all returned to normal by the time of writing. Restored access to some reports should not be confused with full service recovery. FACTUM.PRESS covered the earlier disruption of both tools.
Yandex Cloud: a more important test than any single app
As of October 15, Yandex Cloud's official status page still described the Russian platform as operating under restrictions. Early in the incident, problems affected virtual machines, networking, object storage, databases and Kubernetes. Some components were only partially accessible, and customers faced limits on creating resources or increasing quotas. For business clients, the impact can extend beyond a website or application being offline. Even when existing systems recover, deploying new servers or scaling up may remain difficult.
Immediately after the Vladimir attack, Yandex Cloud warned that its remaining infrastructure was in an unstable configuration and advised customers to consider alternative sites for disaster recovery. Its public language had become less alarming by October 15, but the restricted-operating status remained. Returning a service on reserve hardware does not, by itself, mean that the cloud has regained its previous degree of geographic redundancy.
Yandex made Russian Yandex Cloud services free from October 8 through October 31 for customers who were already using them on October 7. Yandex 360 announced similar support through the end of the month. These are compensation measures, but they also indicate that the company expects the consequences of the incident to persist beyond its first few days.
How Yandex is rebuilding its operations
On October 14, Russian Digital Development Minister Maksut Shadayev told RBC that the government had helped Yandex identify additional facilities. He said some critical systems had been moved to prepared reserve capacity and that migration of the remaining systems was continuing. Neither the precise number of sites involved nor their available capacity or the migration timetable has been disclosed.
Recovery therefore involves three different tasks. First comes bringing essential applications back on surviving or leased hardware. Second is restoring normal functionality, performance and redundancy, which requires spare computing resources and data-integrity checks. Third is rebuilding the damaged physical infrastructure, including power distribution, cooling, network equipment and compute nodes. Each task has a different timetable. An app becoming usable again tells us little about when an affected data center itself will reopen.
When could full recovery happen?
There is no official deadline. Yandex has not published a date for restoring every service or for repairing the equipment at the damaged facilities. Some already functioning services may regain their remaining features gradually, but the available evidence does not justify naming a day when everything will be back to normal.
Experts interviewed by T-invariant suggested that complicated cleaning, testing and repairs to damaged computing equipment could take up to six months. If extensive replacement of hardware and imported accelerators is required, they said rebuilding the affected supercomputer capacity might take one to two years. These are scenarios outlined by outside specialists, not an approved Yandex recovery schedule. The Financial Times has also reported that restoring full infrastructure capacity could take months. Restrictions on access to advanced computing hardware can make replacement harder, but they do not establish a calendar date by themselves.
Claims that “everything will be fixed in a week” and that “Yandex services will be down for two years” make the same mistake: they treat different recovery tasks as one. Payments, taxi dispatch and email can operate on migrated workloads long before a damaged server rack or dedicated training cluster is physically restored.
What the outages reveal
The consequences have spread well beyond Yandex. Outside businesses depend on its cloud infrastructure and software interfaces, and users in neighboring countries also felt the disruption. In a separate FACTUM.PRESS analysis of Kazakhstan, we examined what happened when locally stored personal information remained available but an essential foreign computing component did not.
The strikes have also renewed debate in Russia over geographically distributed backup facilities. After attacks on data centers, Rostelecom floated the idea of a so-called “doomsday” facility for restoring services from backup copies, as FACTUM.PRESS reported in Russian.
The central issue is not simply how quickly damaged buildings can be reconnected to electricity. A large digital platform should be able to withstand the simultaneous loss of several geographically separated facilities without interrupting its most critical functions. That demands additional capacity, independent data copies, regular failover testing and spending that may not be justified by ordinary short-term commercial calculations.
As of October 16, the clearest assessment is this: most everyday consumer services are accessible again, but individual features remain restricted. Advertising dashboards are reopening in stages. Yandex Cloud has not returned to unrestricted normal operation. Two data centers have been completely shut down, and part of a third was damaged. Yandex has not released a full inventory of affected servers or confirmed the physical condition of its named supercomputers. There is no verified date for complete recovery. For now, the company is restoring service availability while still operating with some of its physical computing capacity unavailable.
This analysis is based on publicly available reporting and company statements as of October 16, 2026. Individual service statuses may change throughout the day.