If artificial intelligence systems transform the global economy, then the data centres that train them are the factories of the future. Governments around the world see AI-capable data centres as a strategic resource — one they are racing to control. The idea of high-powered computing as strategic is nothing new. During the cold war, the US permitted supercomputer sales to the Soviet Union only if they were used for weather forecasting, not nuclear simulations. These rules were enforced by requirements that the Soviets accept permanent foreign monitors and even hand over supercomputer data for analysis by US intelligence. Like supercomputers, the AI systems being developed today have both civilian and military capabilities. They may optimise food delivery apps but they can also analyse satellite photos and direct drone strikes. It’s not unreasonable to bet that control over AI data centres will have political as well as economic implications. All advanced AI systems are developed in data centres full of high-end chips such as Nvidia graphics processing units and high-bandwidth memory semiconductors. Cutting-edge AI chips are already subject to US export controls and advanced memory chips may soon be added to the list. Adversaries like China have received blanket bans that prevent them from accessing restricted US chips and are developing their own. It should not therefore come as a surprise that more countries want guaranteed access to AI technology via data centres that are built on their own soil. Saudi Arabia and the UAE have made no secret of their ambition to become AI hubs by investing in vast data centre infrastructure. Kazakhstan wants to build an AI data centre and train Kazakh-language models. Malaysia is experiencing a data-centre boom, with huge new investments by both US and Chinese companies.

Full opinion : How US safeguards on the Microsoft-G42 deal, aiming to protect critical data center and AI infrastructure, could shape future international data center projects.