Three data centres in four days. Overnight on 10–11 October drones hit Yandex’s site in Vladimir, and this time the damage reached the tools SEOs and advertisers use every day. Yandex Metrika and Yandex Webmaster went blank, the Direct interface locked advertisers out, and Search and Alice spent much of Sunday working only partly. By Sunday evening Yandex said the most-used services were coming back. As of Monday morning (UK time) it had still given no date for a full recovery.
This follows the Sasovo fire on 8 October and Friday’s news that Yandex couldn’t confirm the Sasovo hardware could be saved. That Friday morning a second site, in Kaluga, was hit too. Vladimir makes it three.
What Yandex and the reports actually say
- Vladimir is fully stopped. In a statement quoted by Vedomosti, Yandex said the attack damaged the site’s infrastructure and its operation was “completely stopped”, with no casualties. Yandex Cloud added a line no cloud provider wants to write: “We consider the remaining configuration of resources unstable. The platform is operating in emergency mode.” It told customers to use “any other alternative disaster recovery plans”.
- Two availability zones are down. SecurityLab reports a power failure in zone ru-central1-a at 04:12 Moscow time on 11 October (02:12 UK). With ru-central1-b already lost at Sasovo, both zones were unavailable. A newer Vladimir building serving ru-central1-e opened in March 2026, and there were no reports that it had been damaged.
- Kaluga is part-damaged, Sasovo is out. Kommersant puts Sasovo at about 12 MW, home to two Yandex supercomputers, and Kaluga at more than 3,800 racks and 63 MW, partly out of action. According to The Insider, Vladimir housed the third supercomputer, “Galushkin” (16.02 petaflops on the June 2026 TOP500). That would mean all three of Yandex’s ML supercomputers sat on sites that have now been hit. Whether any of them were damaged hasn’t been disclosed.
- The knock-on effects were wide. Fontanka listed Magnit, Lenta, Pyaterochka, Kinopoisk, Lamoda, Dodo Pizza, Vkusno i Tochka, Auto.ru and others among the services that had problems. The BBC said some Yandex services were unavailable, and noted Yandex had said earlier in the week that recovery times were “very difficult to predict”.
The part that matters for search marketers
A notice to Yandex’s business users on Sunday, reposted by agency blogger Mark Larin, set out the position in two lines. Access to the Yandex Business and Direct interfaces is temporarily limited, but ad campaigns keep serving. The Metrika interface is unavailable, but data collection continues and processing may take longer.
Read that carefully. Your Direct campaigns are probably still spending, and you may not be able to log in to pause them, change bids or see what they’re doing. On SearchEngines.guru, webmasters were posting by Sunday morning that Metrika, Webmaster and the RSYa partner dashboard had all stopped loading. One noted that ads were still appearing on sites while the RSYa account itself wouldn’t open. Ad serving had been kept up, while the reporting and control screens were the parts that couldn’t be reached.
Yandex also changed its terms. The Insider reported that on 10 October, the day before the Vladimir strike, Yandex added a force majeure clause to its user agreement covering drone and missile attacks, shelling and explosions. In practice, don’t expect compensation for downtime or for spend you couldn’t control.
How long will this take?
Nobody at Yandex has said. Its own statement on 9 October, before the Vladimir strike, said it was “very difficult” to name a date for restoring the hardware. Experts quoted by Izvestia suggested weeks if only power and cooling were damaged, months if server modules were destroyed, and a year and a half to two years to replace damaged supercomputers. One analyst estimated combined losses at Sasovo and Kaluga at around 15 billion roubles. Those are outside estimates, not figures from Yandex, so treat them with care. Markets are already pricing in the risk. Vedomosti says Yandex shares fell 7.67% after the first two attacks, to about 3,411 roubles, and are down 25.3% this year.
What I’d do this week
- Annotate 8–12 October in every report. Metrika data for these days may arrive late or incomplete. Don’t mistake a gap for an algorithm update or a penalty, and don’t start rewriting pages because of it.
- Use server logs as your source of truth. Logs show whether YandexBot is still crawling and whether Yandex referrals have actually dropped. They don’t depend on anyone’s data centre.
- Review Direct spend as soon as you can log in. Check the Expenses report against what you meant to spend, especially for campaigns you wanted to pause over the weekend.
- Run a second analytics tag on Russian-market sites. If Metrika is your only analytics, this weekend showed what that costs. Polites News made the same point in its guide for site owners.
My view: for years the risk in Yandex SEO was the algorithm. Now it’s also infrastructure. Three sites hit in four days, a cloud platform its owner calls “unstable”, and no restoration dates. Any brand relying on Yandex Search, Direct and Metrika for Russian-market traffic needs a contingency plan this quarter, not next year. We’ll keep following this in Yandex News.