Part 3 – Bombproofing the Backbone: Security, Recovery Timelines, and Forward Risk


In Part 1 we established what happened, and in Part 2 we followed the money. This final part is for the engineers, the operators, and anyone else who wants to understand the how: what physically broke, how long it takes to rebuild, and what this precedent means for an industry most of us never think about until the lights go out.


What actually broke

For all the talk of “Russia’s Google going dark,” the failure was less mysterious than the headlines implied, and the technical record is unusually well documented.

The first domino fell at 1:31 a.m. Moscow time on October 8, when the Sasovo facility lost power in what Yandex Cloud designated its ru-central1-b availability zone[^1]. That’s a deceptively dry phrase — an “availability zone” is simply a physically isolated slice of a cloud provider’s infrastructure, engineered so that a failure in one zone shouldn’t take down the others.

The problem is that the attacks didn’t hit one zone. By October 11, a third strike had knocked out the ru-central1-a zone in Vladimir, and Yandex Cloud’s own status messaging shifted to a tone infrastructure companies reserve for emergencies: the remaining configuration was “unstable,” the platform was in “emergency mode,” and the company was actively telling customers to invoke their alternative disaster-recovery plans — outside Yandex’s cloud entirely[^2].

Cloudflare, which watches global traffic patterns, captured the collapse precisely. Yandex operates two distinct networks: one for its search engine (AS13238) and one for its cloud business (AS200350). The search network’s traffic fell 99% below baseline at the peak on October 11; the cloud network fell 90%[^3]. That’s not an outage you ride out — it’s an outage you evacuate.

And then the company did something that should make every corporate lawyer take notice: on October 10, Yandex quietly added a force-majeure clause to its user agreement, explicitly covering drone and missile attacks[^4]. That is the paperwork equivalent of an admission that the old assumptions — “your data is always available” — no longer hold.


How long does recovery actually take?

Here’s the honest answer, and it’s the part most early coverage under-sold: recovery is measured in months for the ordinary stuff and years for the hard stuff.

The key estimates come from two of Russia’s most-cited telecom/tech analysts, reported by the technical forum Habr:

ComponentRecovery estimateWho
New server hardware delivery2–3 monthsEldar Murtazin (Mobile Research Group)
Supercomputers, if destroyed2–2.5 yearsEldar Murtazin
Vladimir data center restoration“weeks”Denis Kuskov (TelecomDaily)

[^5]

Why the enormous spread between “weeks” and “2.5 years”? Because the two numbers describe fundamentally different problems.

Getting a data center back online means reconnecting power and replacing generic servers — hard, but doable in weeks-to-months, even with sanctions slowing parts delivery[^6]. But the Sasovo and Vladimir sites housed all three of Yandex’s AI supercomputers — “Chervonenkis,” “Lyapunov,” and “Galushkin” — the specialized machines that train YandexGPT, the company’s flagship AI model[^5]. Supercomputers are bespoke, astronomically expensive, and — critically — built from components that Western sanctions have made difficult or impossible to source at scale. Replace them from scratch, Murtazin estimated, and you’re looking at two and a half years[^5].

There’s a smaller technical detail worth knowing too, because it explains why the outage felt worse than a simple “servers down”: by one detailed engineering post-mortem, Yandex Cloud’s Kubernetes and managed-database control plane — the layer that tells all the other servers what to do — took roughly 36 hours to become manageable again after the zone failures[^7]. When the control plane goes down, the apps can’t even be moved to working hardware, let alone run.


Why “redundancy” wasn’t enough

Every cloud provider sells redundancy — the idea that your data lives in multiple zones so a fire in one building can’t hurt you. Yandex’s own architecture was built exactly this way. So why did it collapse?

The short answer, from the same Habr post-mortem, is a concept called capacity exhaustion[^7].

Multi-availability-zone design only works if the surviving zones have enough spare capacity to absorb the load of the lost zones. Lose two of four zones in a week, and you run out of spare room — no matter how cleverly the system was architected.

The symptoms were textbook. Yandex Cloud was forced to switch compute and disk quotas to fully manual mode, freezing new resource creation and prioritizing customers whose critical services were already dead[^7]. In other words, even if you had a backup plan, you couldn’t freely spin up the backup resources — because the cloud itself was rationing capacity.

The post-mortem’s bluntest conclusion should be quoted in full: “Multi-cloud has ceased to be an academic theory.”[^7] The lesson for every business that heard this story: storing your backups in “a different availability zone of the same provider” was never true redundancy — it was one provider wearing a different hat. The companies that weathered the outage best (the post-mortem cites the real-estate platform Cian as a case study) were the ones that had capacity with a different company[^7].


The physical-security panic

The strikes did something else: they jump-started a new industry inside Russia — defending data centers as military targets.

The response unfolded fast:

  • The Digital Development Ministry confirmed it was working with the Defense Ministry on measures to protect data centers from drones, and Kremlin spokesman Dmitry Peskov publicly said the government was discussing the fallout[^8].

  • Rostelecom, the state telecom operator, went further — calling for Russia to build “doomsday” backup sites hardened against exactly this kind of attack[^9].

  • Yandex itself removed its data centers from Maps, in an apparent attempt to reduce their vulnerability to open-source targeting[^10].

  • And the most extraordinary proposal came from the coal-rich Kemerovo region: bury data centers 300 meters underground in exhausted open-pit mines, using the standing groundwater as free cooling and the rock as drone-proofing. The governor pitched it in almost comically practical terms — “a massive hole 300 meters deep… put a roof over it, lay down greenery, build a public park on top, and the data center is completely enclosed and hidden.”[^11]

You can read the coal-pit plan as strange, or as alarmingly rational. Either way, it tells you the direction of travel: in a world where data centers are legitimate military targets, physical security — air defenses, siting, hardening, concealment — has become a core engineering requirement, not an afterthought.


What it means going forward

Stepping back, three forward-looking takeaways matter more than any single bomb:

1. Data centers are now, demonstrably, targets. This wasn’t a one-off. The New York Times framed the Yandex/Sberbank strikes alongside Iranian attacks on data hubs in the Middle East as a turning point for the industry[^12]. The Atlantic Council put it starkly: states are now learning from each other — Tehran hit data hubs in the UAE and Bahrain, and the playbook is spreading[^13]. If you thought data-center siting was a purely economic decision, it just became a geopolitical one.

2. Concentration is the real vulnerability, not any single weapon. The lesson Russia just learned — and every other country should be learning — is that a hypercentralized digital economy is a single point of failure disguised as efficiency. When “the Google of your country” is your maps, your taxis, your payments, your cloud, and your courts, then three buildings can take down a meaningful slice of your modern economy[^13]. Expect to see a push toward distribution — multi-cloud, multi-region, multi-country — driven as much by security as by cost.

3. Insurance, liability, and trust are the quiet consequences. The force-majeure clause is a preview of what’s coming: providers will write attacks into their terms of service, customers will demand cross-provider backup guarantees, and the question of who bears the loss when the infrastructure is bombed — the provider, the customer, or the state — will become a genuine legal and financial fight. As one analysis put it, the “unexpected consequences of bombing data centers” ripple far beyond the blast radius into payment systems and regional cloud markets[^14].


Series coda

The story of Russia’s data-center bombings is, at bottom, a story about a modern economy discovering how much of itself it had quietly bolted onto a few concrete buildings. The drones found the weak point. The hard question — for Russia, and for a world watching — is whether anyone else will treat it as a warning before their own lights go out.


Sources

[^1]: vc.ru, “Сбой Yandex Cloud нарушил работу российских сервисов,” (Oct. 8, 2026). Link

[^2]: Habr, “Yandex Cloud работает в аварийном режиме после атак на три дата-центра Яндекса,” (Oct. 12, 2026); BleepingComputer, “Yandex confirms service outage as users lose access to cloud files,” (Oct. 12, 2026). Link

[^3]: Cloudflare, “Yandex outage: data shows that instead of going dark, malicious bots went on the move,” (Oct. 14, 2026). Link

[^4]: Habr, “Yandex Cloud работает в аварийном режиме после атак на три дата-центра Яндекса,” (Oct. 12, 2026). Link

[^5]: Habr (citing Mobile Research Group & TelecomDaily), “Analyst Estimates on Recovery Time,” (Oct. 12, 2026). Link

[^6]: RFE/RL, “How Ukraine’s Strikes On Data Centers Disrupted Taxis, Shopping, And Business In Russia And Beyond,” (Oct. 13, 2026). Link

[^7]: Habr, “Каскадный сбой Yandex Cloud: анатомия событий октября 2026 и выживание бизнеса вне одного облака,” (Oct. 12, 2026). Link

[^8]: The Moscow Times, “Russia Weighs Drone Defenses for Data Centers After ‘Unprecedented’ Yandex Attacks,” (Oct. 12, 2026); Reuters, “Kremlin says government is discussing Yandex’s problems after Ukrainian strikes,” (Oct. 13, 2026). Link

[^9]: Data Center Dynamics, “Rostelecom: Russia needs ‘doomsday’ data center to combat Ukrainian strikes,” (Oct. 15, 2026). Link

[^10]: UNITED24 Media, “Yandex Hides All Four Cloud Data Centers on Maps After Ukrainian Drone Strike on Russia,” (Date not specified). Link

[^11]: The New Voice of Ukraine, “Russia proposes building data centers in coal pits after drone strikes,” (Oct. 16, 2026). Link

[^12]: The New York Times, “Data Centers Become Targets in Wars in Ukraine, Russia and the Middle East,” (Oct. 15, 2026). Link

[^13]: The Atlantic Council, “Ukrainian drone strikes highlight the vulnerability of Russia’s data centers,” (Oct. 16, 2026). Link

[^14]: El País (English), “The unexpected consequences of bombing data centers,” (Oct. 14, 2026). Link

Leave a Reply

Your email address will not be published. Required fields are marked *