Amazon says drone strikes hit 3 facilities in UAE, Bahrain

Amazon Web Services said two data centers in the United Arab Emirates (UAE) were directly struck by drones, causing sparks and fires and taking the sites offline.

A facility in Bahrain was hit by a drone strike nearby, causing physical impacts but not a direct hit.

AWS said the incidents occurred Sunday morning and linked the outages to the ongoing conflict in the Middle East.

The strikes caused structural damage, disrupted power delivery, and in some cases led to water damage from fire suppression.

Key services, including EC2, S3, and DynamoDB, experienced elevated error rates and degraded availability.

AWS said it is working to restore data access and service availability and asked customers to consider backing up data or migrating workloads to other regions.

Separately, Amazon posted notices to its Middle East marketplaces, including Israel, Saudi Arabia, Kuwait, Bahrain, and the UAE, warning of extended delivery times amid missile and drone activity in the region.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

One Availability Zone took the hit, separate power problems spread the outage

The claim that two data centers were struck skips over how the failure unfolded.
AWS said objects hit a data center in one Availability Zone (AZ) called mec1-az2, which caused sparks and fire 1.
An AZ is a cluster of one or more data centers with independent power and connectivity within a geographic Region, designed to isolate failures 2.
AWS also said two Availability Zones in the United Arab Emirates remained impaired, while another update described a “localized power issue” in a separate zone of data centers 1, 3.
AWS said it was investigating connectivity and power issues in Bahrain, with no report of a direct strike there 3.
Together, these problems disrupted services including EC2, S3, and DynamoDB.

Conflict turns cloud risk into a physical problem

The incident sets a precedent where cloud hardware becomes a failure point during military conflict.
Region selection now needs to weigh geopolitical risk along with latency and data sovereignty.
Multiple impaired AZs in one Region strain a common expectation about cloud resilience 1.
AWS told customers to “enact their disaster recovery plans” and recover into alternate AWS Regions, which goes beyond typical outage advice 4.
This pressure may speed up multi-region designs, which raise cost and complexity.