After Attacks on Its Data Centers, Yandex Strikes Deals With Selectel, K2Cloud and VK Cloud

Yandex has agreed with three Russian cloud providers, Selectel, K2Cloud and VK Cloud, to speed up the hosting of its customers' data and services. The trigger was attacks on the company's own data centers. vc.ru reported this on 9 October 2026. In this arrangement, Yandex's cloud-market competitors act as a fallback site, which is one of the most direct signals of how serious the situation turned out to be.

What is known so far

The original report is short. It names three providers and a shared goal: to quickly take on Yandex customers' data and services. Everything else is absent from the published material.

• Who is involved: Yandex as the source of customers; Selectel, K2Cloud and VK Cloud as the receiving sites.
• What is being moved: customers' data and services, with no breakdown by product.
• Cause: attacks on Yandex data centers. The type of attack (cyberattack, physical action or something else) is not specified.
• Not disclosed: the number of affected customers, the volume of data being moved, the timeline, the commercial terms and the list of services.

The lack of these details matters in itself. It is still unclear whether this is preventive diversification or a forced evacuation of part of the workload. The answer determines how to assess the consequences for the whole market.

Why competitors are helping each other

Selectel, K2Cloud and VK Cloud compete with Yandex for enterprise customers. Even so, they agreed to take on someone else's workload, and there is a clear logic to it. Customers left without service are easier to keep inside the national cloud ecosystem than to see them go to foreign solutions or to their own servers. For Russian companies that handle citizens' personal data, the choice is limited anyway by data-localization requirements. The three named providers meet those requirements.

The technical side of the move is less straightforward. Clouds have different APIs, different managed database services, and their own load-balancing and storage schemes. Plain virtual machines and object storage move quickly. Managed services, queues, serverless functions and provider-specific identity mechanisms require rework. So the word "accelerated" in the announcement is best read as a promise of priority and dedicated resources, not a guarantee of instant migration.

AI workloads raise a separate question. Model training and inference depend on GPU clusters and fast networks between nodes, and that kind of capacity is scarce in Russia. The source does not say whether GPU instances and ML platforms are part of the agreements.

What cloud customers should do

The story offers a practical lesson that holds no matter how this particular episode ends. A single provider, even a large one, remains a single point of failure.

• Check where your backups are stored. If they sit in the same cloud and the same zone as production, that is not redundancy.
• Describe your infrastructure as code (Terraform, Ansible, Helm). This cuts migration time from weeks to days.
• Avoid excessive lock-in to proprietary managed services where open alternatives exist: PostgreSQL, Kafka, S3-compatible storage.
• Lower the DNS TTL for critical records in advance, and check who in your company is able to change routing.
• Reread your contract: what availability and recovery obligations it states, who is liable for losses, and how data export is handled.
• If Yandex customers are in your supply chain, ask them for their current recovery plan.

Three things will shape what happens next: Yandex's official explanation of the nature of the attacks, the timeline for moving services, and how smoothly migration goes for the first customers. For now, only the agreement itself is confirmed, and the question of its scale remains open.