Yandex's Hardware Dead End: Why Rebuilding Its Servers Will Take Years and Transform the Entire Russian IT Market

My modest analysis of the situation surrounding the Russian IT giant's data centers: why the rapid restoration of mobile apps masks the worst hardware crisis in the history of the Russian internet.

The software illusion and the true scale of the damage

A series of kinetic drone strikes over four days in October 2026 disabled three of Yandex's six known data centers — in Sasovo, Ryazan Region, Kaluga, and Vladimir. At the height of the infrastructure crisis, the company's network traffic plummeted by 99% in its search segment and by 90% on the Yandex Cloud platform. Users across the country and abroad experienced widespread outages affecting taxi booking, money transfers, and delivery services.

Through cascading throttling (including a temporary ban on downloading files from Yandex Disk) and an emergency migration to competitors' infrastructure—VK Cloud, Selectel, and K2Cloud—the basic functionality of consumer apps was restored by October 19. However, this rapid, superficial stabilization of smartphone interfaces creates the dangerous illusion that the worst is already over.

An optimistic six months versus the harsh reality

When it comes to replacing physical equipment and repairing infrastructure, the timeline changes dramatically. The optimistic scenario presented by sources cited by Kommersant suggests that delivery of equipment to replace hardware destroyed in the UAV attacks could take up to six months.

However, industry experts and internal sources estimate that the actual time required to restore the hardware infrastructure will be at least 1–1.5 years in the best-case scenario. Moreover, cybersecurity analysts explicitly doubt that the destroyed server capacity can be restored to its previous level within six months, a year, or indeed at any point in the foreseeable future.

A sanctions dead end and the loss of AI supercomputers

The main obstacle to a swift recovery is not simply the need to purchase standard server racks, but the catastrophic loss of high-performance computing infrastructure. The three sites that were attacked housed all three of the company’s supercomputers—Chervonenkis, Lyapunov, and Galushkin—which were used to train YandexGPT neural networks and autonomous vehicle algorithms.

These high-performance systems were powered by US-made Nvidia A100 graphics accelerators, which are subject to the strictest international sanctions. Attempts to replace the destroyed chips through parallel import schemes are running into a global shortage of AI accelerators and unprecedented price increases. Yandex is being forced to procure additional equipment in “the most expensive market in recent years,” competing for chips with international corporations.

Geographic Shift and Rising Costs for Customers

Even with sufficient funding, rebuilding the data centers at their former locations near Moscow carries a constant risk of further drone strikes. The traditional model of locating data centers within 100–300 kilometers of the capital has proven vulnerable. The government is already discussing the need to redesign the architecture and relocate key computing nodes far beyond the Urals—for example, to regions of Siberia—as well as procure mobile containerized data centers.

The enormous cost of relocation, gray-market imports of scarce chips, and the construction of electronic warfare and passive defense systems will inevitably lead to a significant increase in the price of Yandex Cloud services for end customers. As a result, restoring the hardware at the company’s “iron heart” will take 12–18 months, while the economic repercussions will continue to affect the cost of Russian digital services for a long time to come.