What AWS says cannot be restored
Amazon Web Services says it has exhausted restoration options for resources and data that were still hosted exclusively in its Bahrain region when the region became unavailable. AWS told customers that damage there spanned multiple Availability Zones and exceeded what its regional and multi-AZ services were designed to withstand.
In the United Arab Emirates, AWS says it cannot restore access to resources and data hosted exclusively in the mec1-az2 Availability Zone. Recovery work is continuing for regional resources and for resources in the other affected UAE zones, mec1-az1 and mec1-az3. The company plans further UAE restoration updates in the coming months and another Bahrain update in early 2027.
The failure unfolded across two rounds of disruption
The infrastructure damage began in March during Iranian missile and drone retaliation following US and Israeli attacks on Iran. AWS said at the time that two UAE facilities were directly struck and that a strike near a Bahrain facility caused physical damage. After the first Bahrain Availability Zone was damaged, AWS recommended that customers move workloads to other regions.
Reuters reports, citing AWS's status update, that most customers completed that migration before disruption to a second Bahrain Availability Zone in April took the region offline. AWS says many affected customers later re-established operations elsewhere from backups or copies of data that remained accessible. The unrecoverable scope therefore concerns resources that did not have an accessible copy outside the destroyed failure domain.
Why the incident matters for cloud resilience
The incident provides a rare public example of physical destruction crossing the boundaries normally assumed by cloud availability architecture. Availability Zones are intended to isolate failures inside a region, but the Bahrain damage affected more than one zone and AWS says the event exceeded the design envelope of its regional and multi-AZ services.
That does not establish that every AWS durability promise failed, or that ordinary multi-AZ architecture is equivalent to a cross-region disaster-recovery plan. AWS services have different replication and durability models, and some resources are deliberately zonal. The practical lesson supported by this event is narrower: workloads that must survive destruction of an entire regional failure domain need a recovery path whose data and control plane remain accessible outside that domain.
What remains unclear
AWS has not published the number of customers affected, the volume of inaccessible data or a service-by-service inventory of what could not be recovered. Without that detail, it would be misleading to generalise the incident into a single claim about the durability of EC2, S3, RDS or other AWS products.
The Hacker News discussion that surfaced the update contains broader claims about cloud redundancy, pricing and customer architecture, but those comments are not evidence for the scope of the loss. The established facts are AWS's own recovery determination, the affected Bahrain region and UAE zone, and Reuters's reporting of the company's status update; the service-specific causes and customer recovery outcomes remain more limited than the headline event.