This is no ordinary regulatory filing. When the United States decides to ease export controls on advanced chips for a single country, the message is clear: an ally is being rewarded, and a power architecture is being built. UAE-based G42 seized the moment, announcing a 5-gigawatt AI campus spanning the United Arab Emirates and the United States, fueled by those very newly approved chip exports.

The figure is staggering. Five gigawatts is roughly the power consumption of a small city. In AI infrastructure terms, it means hundreds of thousands of cutting-edge accelerators – GPUs and ASICs – capable of training and inference at continental scale. You don’t build a facility like this without deep strategic alignment with Washington, and indeed the export license is no random pass: it’s the fruit of a privileged relationship that the Emirates have cultivated, likely by committing to limit technology ties with China and to keep data under shared security rules.

For those working with on-premise and hybrid deployments, the news redraws the map of available options. Until now, anyone wanting to train or serve LLMs in the Middle East with top-tier hardware often had to rely on major cloud providers with data centers in Europe or the US, facing latency and data sovereignty trade-offs. A 5-gigawatt campus – if opened to colocation or managed services – offers local hyperscale capacity with a US approval stamp that validates it on the export control front. This means financial institutions, energy companies, and government bodies in the region can seriously consider physically close infrastructure for sensitive workloads without forgoing the most advanced cards.

Winners and losers. Beyond G42 and the Emirati tech ecosystem, the entire chip supply chain benefits: NVIDIA, AMD, Intel, and memory and networking suppliers will see an order-of-magnitude demand that tilts the market. Middle Eastern enterprises needing low-latency inference or fine-tuning on local data get a concrete alternative to Western cloud giants. On the losing side, China sees the selective-fence strategy confirmed: while some partners get a longer leash, Beijing faces ever-higher barriers. Global cloud providers could also face competition from a regional pole that combines state-of-the-art hardware with data residency in a single offering.

Structurally, the move signals that export restrictions are not a uniform wall but a granular grid that rewards geopolitical alignment. It’s likely that other countries – India, Japan, some Southeast Asian states – will seek similar deals, creating a mosaic of Washington-approved AI hubs. This fragments the global compute landscape: instead of a single center of gravity in Silicon Valley or Ireland, privileged corridors will emerge where chips travel according to bilateral understandings. For those making infrastructure investment decisions, it means navigating not only TCO and performance assessments but also shifting political maps, where an export license becomes a factor of cost and availability.

The G42 campus is a monument to a new era: one where every watt of Washington-approved compute draws a new border, and where technological sovereignty hinges on alliances as much as on silicon.