xAI ordered 5 new gas turbines to power up its AI supercomputer

xAI Founder and CEO, Elon Musk, has confirmed the purchase of 380MW natural gas turbines from Doosan Enerbility to generate on-site power for AI data centers.

xAI has been aggressively expanding its computing infrastructure to advance AI development. A key challenge in the space is securing sufficient power for large-scale data centers, as traditional grid connections often face delays due to high demand and regulatory hurdles.

To address this challenge, xAI has brought on-site power generation using natural gas turbines, allowing faster deployment than relying on utility approvals. The AI company kick-started with smaller, mobile solutions such as truck-mounted gas turbines. For instance, at its Colossus-1 facility in Memphis, Tennessee, xAI installed 12 SMT-130 turbines.

Concerns from locals arose over noise and environmental impact in Tennessee, which led xAI to expand operations across the state border into Southaven, Mississippi, adding 7 Titan-350 turbines.

This cross-border approach minimizes regulatory obstacles by distributing infrastructure while maintaining short transmission lines to the main data centers. By late 2025, this setup generated around 460 MW of power, positioning xAI to achieve the world’s first gigawatt-scale data center at a single location.

Memphis’ location at the intersection of Tennessee, Mississippi, and Arkansas offers strategic flexibility. If issues arise in one state, xAI can relocate turbines to another nearby jurisdiction, providing uninterrupted progress. This setup also utilizes engineering expertise from Musk’s other companies, including Tesla and SpaceX, for rapid scaling.

Building on this foundation, xAI has improved the power capacity by procuring larger turbines. Previously, the AI company acquired a new H-class turbine design from Doosan Enerbility, a South Korean company providing integrated solutions for power plants.

In the latest edition, the xAI has purchased 5 additional 380MW natural gas turbines from the same supplier. The first two units will be delivered by the end of 2026. After delivery of all units, the power output equivalent of running 600,000 Nvidia GB200 or more than 350,000 VR200 units in AI clusters.

These turbines are critical for AI training, which requires high computational resources. Related to this, natural gas delivers a quick-to-deploy and reliable alternative to renewables like solar, which requires extensive land, batteries, and longer timelines.

This expansion underscores xAI’s focus on vertical integration in infrastructure. Getting more control over power generation avoids grid bottlenecks that have stalled other AI firms.

The total additional capacity from these 5 turbines amounts to 1.9 GW, potentially making xAI’s Memphis facility the largest AI data center by 2027. It will also help the company scale computing to make its Grok AI more intelligent and capable of providing fast and real-time information and services to the end user.

(source)