Yandex Data Centers Hit Twice in 48 Hours

This is a stark reminder that “cloud” infrastructure remains physically concentrated and vulnerable, especially when a provider’s AI compute is clustered in a few sites.

• –Sasovo hosts two of Yandex’s three Nvidia A100-based supercomputers used to train YandexGPT.
• –Yandex warned of disruptions to cloud services, including power and Object Storage issues.
• –Developers relying on a single availability zone or provider face serious recovery risks.
• –Rebuilding advanced AI infrastructure may be difficult amid sanctions and limited access to imported chips.
• –The incident elevates geographic redundancy and disaster recovery from best practices to strategic necessities.