The crisis began on February 28, when US-Israeli attacks on Iran set off a war that quickly reached critical digital infrastructure across the Gulf. On March 1, Iranian drones struck Amazon facilities in Bahrain and the United Arab Emirates; AWS urged customers to shift workloads to other regions and use remote backups for inaccessible systems.¹

At the time, the company’s advice suggested disruption, not irreversible loss. But the damage kept mounting. A follow-up strike hit Bahrain on April 1, and the Iranian Revolutionary Guard later claimed the Bahrain site had been targeted over Amazon’s support for the US military.² In July, another missile strike targeted a remaining Amazon data-centre structure in Bahrain, damage later confirmed by satellite imagery.¹

Six months on, AWS has delivered its starkest assessment. In Bahrain, it said the destruction across multiple availability zones “exceeded what our regional and multi-AZ services are designed to withstand,” leaving it unable to restore access to resources and data hosted exclusively in the region.¹ In the UAE, data held in one availability zone, mec1-az2, is also beyond recovery, while AWS continues trying to restore the other two affected zones.¹

Amazon says it has suspended billing in the affected regions, replaced damaged infrastructure and remains committed to customers there, with a UAE update due in coming months and a Bahrain update promised in early 2027.² Yet the admission lands far beyond AWS’s own outage: Gulf governments and technology companies had envisioned the region as a major AI-computing hub. Reuters reported that the UAE is now weighing dispersed or underground facilities, blast-resistant construction, and drone and missile interceptors for its planned Stargate campus.²

The episode is a brutal qualification to cloud computing’s promise of redundancy. Multiple facilities can protect against routine failure; they are far less certain protection when a war strikes an entire region.