Microsoft is developing a global network of interconnected data centers, dubbed 'superclusters', to support the next generation of artificial intelligence models that will use hundreds of trillions of parameters. The initiative began last October by linking data centers in Mount Pleasant, Wisconsin, and Atlanta, Georgia, enabling them to function as a single high-performance computing unit. The company aims to expand this model worldwide, using high-speed internet connections to integrate multiple centers for large-scale AI training. These facilities, called 'Fairwater', employ direct-to-chip liquid cooling to minimize water use. Microsoft Azure CTO Mark Russinovich said the company’s goal is to scale AI infrastructure across multiple centers for efficiency and flexibility. Analysts note that this approach will allow Microsoft to optimize data center locations based on cost, climate, and energy availability.