UAE scatters its 5-gigawatt AI buildout into a defended network after the Iran war

The UAE is breaking its planned 5-gigawatt AI data-center complex into a distributed network of smaller, defended sites scattered across the country, six people familiar with the matter told Reuters, a redesign forced by the Iranian drone and missile strikes that hit Gulf infrastructure during the Iran war.

The original blueprint called for a single 10-square-mile (26-square-kilometer) campus on the outskirts of Abu Dhabi, one of the largest AI data-center projects outside the United States. The shape now under consideration replaces that concentration of compute, power, and cooling with a topology of smaller facilities, some hardened with air defenses and some built partially underground, that can keep operating if individual sites are hit.

The project at stake is the UAE-US AI Campus, led by Abu Dhabi-based G42 and announced during President Trump's visit to the Emirates last year. The first phase, a $30 billion, 1-gigawatt compute cluster branded Stargate UAE, began construction in 2025; the first 200 megawatts of capacity are scheduled to come online this year. Stargate itself is a U.S. joint venture between OpenAI, Oracle, and SoftBank to build AI infrastructure, and the UAE committed under the same U.S. deal to invest in Stargate's American data centers.

G42 is controlled by Sheikh Tahnoun bin Zayed al Nahyan, the UAE's national security adviser and brother of its president. At the project's launch, Oracle chairman and CTO Larry Ellison said the end-state ambition is to connect "every UAE government agency and commercial institution" to the world's most advanced AI models.

The redesign is a direct answer to the threat model the Iran war made real. The same drone and missile risk that hangs over Gulf energy and industrial facilities now applies to a centralized 10-square-mile concentration of GPU clusters, power substations, and high-bandwidth fiber, the kind of target a single coordinated strike could put out of service for weeks. Distributed sites reduce the blast radius of any one attack. Underground and air-defended facilities raise the cost of a successful hit. A topology that spreads load across the country can reroute around a damaged site the way a content-delivery network reroutes around a failed node.

The UAE secured access to Nvidia AI chips and other advanced U.S. technology in a deal that required it to cut AI ties with China and to invest in the Stargate initiative's American data centers. The defended, distributed shape sits more comfortably inside that framework than a single foreign mega-site would, and Reuters' sources did not address how the redesign will change the project's cost or its U.S. obligations. Distributed, hardened, and partially underground capacity is, by any measure, more expensive to build and run.

The numbers Reuters could not pin down are the ones that matter most to operators. The wire's six sources, a U.S. official, a Western diplomat, and industry executives briefed on the deliberations, with four speaking on condition of anonymity, could not say how close authorities are to a finalized new masterplan, how the distributed topology will change the project's total cost, or whether the first-phase 200-megawatt milestone will land on schedule. The 200-megawatt delivery is the most concrete near-term test: if it ships on time on a single defended campus, the security-driven redesign thesis is wrong. If the site footprint has visibly fractured into multiple locations, the kinetic-risk story is the one to follow.

The wider signal is structural. The same threat model that pushed the UAE to scatter its compute now applies to every Gulf AI mega-project in the pipeline, and to hyperscaler regions elsewhere planned as single concentrated sites. Siting AI infrastructure the way militaries site runways, distributed, hardened, redundant, is no longer a fringe idea in the Gulf. It is the plan.