What happened to the AWS data centres in the UAE in March 2026?

On 2 March 2026, AWS confirmed two UAE facilities were directly struck by drones and a third in Bahrain was damaged by a nearby strike. EC2, S3, DynamoDB, Lambda, Kinesis, CloudWatch, and RDS were all degraded across the ME-CENTRAL-1 and ME-SOUTH-1 regions.

Why does this matter for Australian businesses?

Enterprises on AWS Middle East faced unplanned, real-time migration decisions with no warning. For workloads subject to SOCI Act or HCF requirements, Europe as a failover does not meet compliance. Domestic sovereign infrastructure is the only compliant answer, and pre-positioning DR configurations in certified Australian facilities is now an urgent priority.

How much are hyperscalers investing in data centres globally?

Total future lease commitments across Oracle, Microsoft, Meta, Amazon, Google, and CoreWeave exceed US$700 billion as of March 2026. Oracle leads with US$261 billion; Microsoft carries US$155 billion; Meta holds US$104 billion.

When the Cloud Burns: Why Australia Must Win the Race for Sovereign AI Infrastructure

On 1 March 2026, Iranian drones struck three AWS data centres in the UAE and Bahrain in the first confirmed military attack on a hyperscale cloud provider in history, taking two of three availability zones in the ME-CENTRAL-1 region offline simultaneously
In the same week, Bloomberg reported global hyperscale lease commitments have crossed US$700 billion, with Microsoft and Meta each adding US$50 billion in a single quarter and Oracle holding US$261 billion in future obligations structured over 15 to 19 years
Sydney's sovereign-certified capacity is effectively sold out; enterprise and government buyers without pre-positioned domestic failover had no compliant recovery path during the AWS incident, making HCF-certified colocation an immediate risk management priority, not a future IT consideration

One Week, Two Forces

The strikes occurred on Sunday 1 March 2026. Two AWS facilities in the UAE were directly hit; a third in Bahrain sustained damage from a nearby explosion. AWS confirmed two of its three availability zones in the ME-CENTRAL-1 region went offline, with EC2, S3, DynamoDB, Lambda, RDS, Kinesis, and CloudWatch all degraded or unavailable. Amazon told customers that recovery would be prolonged "given the nature of the physical damage involved" and advised immediate migration to alternate regions. The Uptime Institute confirmed it as the first military attack on a hyperscale cloud provider in history.

Three days later, Bloomberg reported that Microsoft and Meta each committed approximately US$50 billion in new data centre lease agreements in their most recent quarters, pushing total industry-wide future commitments past US$700 billion. Oracle leads all operators with US$261 billion in future obligations, up 148 percent from August 2025, driven largely by OpenAI's US$300 billion cloud deal. These are forward commitments to capacity not yet built, locked in years before a single rack is installed.

Read separately, each story is significant. Read together, they make the same argument: geography is now the most important variable in infrastructure investment, and geopolitical stability has become a hard commercial requirement.

Why the Strikes Repriced Location Risk

The strikes did not just disrupt services. They destroyed the assumption that commercial cloud is a neutral, protected category of infrastructure. The IRGC stated it targeted AWS facilities specifically because they host workloads supporting US military operations. AWS declined to comment on that claim, but the blast radius was real: two of three availability zones across an entire AWS region went dark simultaneously, a scenario AWS's own redundancy model is not designed to survive.

This matters because the same infrastructure powering enterprise applications, AI platforms, and financial services also supports intelligence processing, logistics coordination, and battlefield decision-making.

The $700 Billion Wave Needs Somewhere Safe to Land

The hyperscalers committing capital at this scale cannot afford to concentrate it in conflict-exposed regions. Oracle's leases are structured over 15 to 19 years. Microsoft brought 1 gigawatt of capacity online in a single quarter. These are generational commitments that require political stability, reliable energy, and sovereign regulatory certainty as baseline conditions.

Australia meets all three. Geopolitical stability, a maturing renewable energy sector, and physical distance from active conflict zones are precisely the attributes this capital wave is seeking. Australia's national data centre capacity has grown fortyfold since 2005 and is projected to double again by 2030. Domestically, Macquarie Data Centres' AUD$350 million IC3 Super West facility at Macquarie Park is on track to open Phase 1 in September 2026, the only new AI-density sovereign facility opening in Sydney's north zone that year. The National Reconstruction Fund's $200 million commitment to Macquarie Technology Group signals that government co-investment is beginning to move in step with that market reality.

The Window Is Open, Not Permanent

Other Indo-Pacific markets are actively competing for the same capital with regulatory certainty, streamlined planning approvals, and coordinated energy commitments. Australia's fragmented state-level planning approvals and the absence of a national data centre strategy create drag at precisely the wrong moment. The decisions made in Australian policy and planning frameworks over the next 12 to 18 months will determine whether global hyperscale capital flows here, or whether it moves to markets that removed the friction first.