xAI is moving to scale up its artificial intelligence infrastructure in the southern US, buying a data centre footprint anchored in Memphis.
The expansion centres on the purchase of a third facility in the region, a move that significantly lifts the company’s planned computing capacity and reinforces its ambition to run one of the world’s largest AI training operations.
The development was disclosed by Elon Musk via a post on X.
The site acquisition brings xAI’s total planned power capacity close to 2 gigawatts, a scale that places the project among the most energy-intensive AI data centre clusters currently under development.
Expanding footprint
xAI has already completed one major data centre in Memphis, known as Colossus. A second facility, referred to as Colossus 2, is under construction nearby.
The newly acquired building, called MACROHARDRR, is located in Southaven and directly adjoins the Colossus 2 site, according to earlier reporting by The Information, which cited property records and a person familiar with the project.
By clustering multiple facilities across state lines, xAI is building a contiguous computing campus rather than a single standalone site.
This layout allows the company to concentrate energy supply, cooling infrastructure, and high-speed data connections in one region, a structure increasingly favoured by firms training large-scale AI models.
Power and scale
Musk indicated that the third building would take xAI’s training compute to almost 2 gigawatts.
At that level, the electricity demand would be comparable to that used by hundreds of thousands of US households, underlining the strain that next-generation AI infrastructure can place on local power grids.
Such a scale reflects the rapidly rising computational demands of frontier AI systems.
Training large language models requires vast amounts of parallel processing, typically delivered through dense clusters of advanced graphics processing units and specialised networking hardware.
Chips and capital
Earlier this year, Musk outlined plans for Colossus 2 to eventually house around 550,000 chips supplied by Nvidia Corp.
At prevailing market prices, such an installation would involve hardware spending running into the tens of billions of dollars, before accounting for buildings, power systems, and ongoing operating costs.
The Memphis expansion signals that xAI intends to compete at the very top end of the AI infrastructure race, alongside technology groups and cloud providers that are also investing heavily in custom-built data centres.
By owning and controlling its physical sites, xAI gains direct oversight of compute availability, a critical factor as demand for high-end AI training resources continues to tighten.
Strategic implications
The choice of the Memphis area reflects a broader industry shift toward regions that can support large land parcels, high-voltage power access, and favourable logistics.
For xAI, consolidating multiple facilities in one geographic cluster could streamline operations while accelerating deployment timelines.
As construction progresses on Colossus 2 and integration begins at the newly acquired MACROHARDRR building, the Memphis region is set to become a central hub in xAI’s long-term AI training strategy.
The post Elon Musk’s xAI buys third data center to expand its supercomputer clusture appeared first on Invezz