The race to build the infrastructure behind artificial intelligence has usually been framed around three things: chips, electricity and money. A major rethink underway in the United Arab Emirates adds another factor to that list — physical security.
Reuters reported on September 11 that the UAE is revising plans for its enormous 5-gigawatt AI data centre programme after attacks in the Gulf raised concerns about the vulnerability of critical technology infrastructure. The project had originally been envisaged as a roughly 10-square-mile campus in Abu Dhabi, but Reuters, citing people familiar with the discussions, says later phases are now likely to be distributed across multiple locations.
That matters because this is not an ordinary data centre project. The wider UAE–US AI Campus is one of the most ambitious AI infrastructure developments outside the United States and sits at the centre of a technology partnership involving some of the biggest names in the industry.
What happened?
The security reassessment follows attacks that damaged US-linked technology infrastructure in the Gulf earlier this year. Reuters reported that officials are considering measures including underground construction, blast-resistant materials, additional backup power and cooling, and protection against drone and missile threats.
The report does not mean Stargate UAE has been cancelled. In fact, G42 told Reuters that work on the AI campus is progressing as planned, while noting that project details are continually reviewed against the security, resilience and operational standards expected of critical infrastructure at this scale.
OpenAI similarly said it was making progress with the UAE on the infrastructure and adoption priorities needed to deliver the country’s AI ambitions.
Stargate UAE is only part of a much bigger project
OpenAI announced Stargate UAE in May 2025 as its first international Stargate deployment. The company said the 1-gigawatt cluster in Abu Dhabi would be built with G42, Oracle, NVIDIA, Cisco and SoftBank, with the first 200 megawatts expected to come online in 2026.
G42 has described Stargate UAE as part of the broader 5-gigawatt UAE–US AI Campus. NVIDIA is expected to supply advanced AI systems, while Oracle is involved in operating the Stargate cluster alongside OpenAI. Cisco is also part of the infrastructure partnership.
The scale is difficult to overstate. AI companies increasingly need enormous amounts of specialised computing capacity, power and cooling to train and run advanced models. Projects such as Stargate are effectively becoming industrial infrastructure for the AI era.
Why the redesign matters
The interesting part of this story is not simply that one data centre may be redesigned. It is what the change says about the next phase of the global AI build-out.
Until now, much of the debate around AI infrastructure has focused on access to GPUs, electricity supply, water use, construction costs and semiconductor export controls. The Gulf situation demonstrates that hyperscale AI facilities may also need to be treated like other strategic infrastructure.
Concentrating enormous amounts of compute in a single campus creates obvious efficiencies. It can also create a single point of vulnerability. Spreading capacity across multiple sites may make an AI network more resilient, but it could introduce additional costs and complexity in networking, power delivery, cooling and security.
If that thinking spreads, the architecture of future AI campuses could begin to look different. Instead of simply asking where power and land are cheapest, developers may increasingly ask how infrastructure can continue operating through geopolitical disruption, cyber incidents or physical attacks.
What does it mean for readers?
For everyday users, a redesign of an AI campus thousands of kilometres away may sound remote. But the infrastructure underneath services such as ChatGPT and other AI platforms ultimately affects how quickly new capabilities can be deployed, where data can be processed and how much it costs to provide increasingly compute-intensive services.
More resilient infrastructure could reduce the risk that a regional disruption takes large amounts of computing capacity offline. The trade-off is that hardened and geographically distributed facilities can be more expensive to build.
There is also a broader lesson here. AI is rapidly moving beyond software. The industry now depends on semiconductor factories, power grids, fibre networks, cooling systems and enormous physical facilities. As those assets become strategically important, governments are likely to treat them increasingly as critical national infrastructure.
The AI infrastructure race is entering a new phase
Stargate UAE remains an ambitious attempt to build frontier-scale AI capacity outside the United States. The first 200MW phase is still expected in 2026 according to the project’s original timetable, and the companies involved have not announced that the project is being abandoned.
But the reported rethink reveals something important about where the AI boom is heading. Building the world’s largest computing systems is no longer just an engineering challenge. It is becoming a resilience and national-security challenge too.
That may ultimately prove to be one of the defining shifts in the next generation of AI infrastructure.
Sources: Reuters reporting published September 11, 2026; OpenAI’s official Stargate UAE announcement; G42’s official Stargate UAE and UAE–US AI Campus announcements.
