Common Questions

How many gas turbines does xAI use at Colossus?

Tennessee authorities approved 41 natural gas turbines totalling 1.2 GW for xAI's Colossus complex in Memphis in March 2026, after the turbines had already been operating under temporary permits.Source: Lowdown data-centres briefing

Why is xAI's Memphis data centre controversial?

xAI's Colossus facility used temporary gas turbine permits before formal approval, raising air-quality concerns in the surrounding majority-Black neighbourhood and questions about regulatory oversight of behind-the-meter power generation.Source: Lowdown data-centres briefing

What is Colossus xAI?

Colossus is xAI's AI supercomputer cluster in Memphis, Tennessee, described as the world's largest AI training facility by GPU count. It runs on a combination of onsite natural gas generation and grid power.Source: xAI

Background

xAI drew regulatory attention in March 2026 when Tennessee authorities approved 41 natural gas turbines totalling 1.2 GW to power its Colossus supercomputer complex in Memphis. The approval was controversial: the turbines operated under temporary permits before formal approval, and air-quality advocacy groups raised concerns about localised pollution in a majority-Black neighbourhood.

Colossus is the world's largest known AI training cluster by GPU count, built at speed that the grid cannot match. xAI's behind-the-meter gas strategy is the most prominent example of a growing industry pattern: hyperscalers sourcing their own generation to escape grid-connection queues rather than waiting years for utility power. GE Vernova, whose backlog of gas turbine orders reached 80 GW by December 2025, cited xAI-style demand as a primary driver of that backlog stretching into 2029.

xAI was founded by Elon Musk in 2023 after his departure from OpenAI's board. Colossus is located in Memphis, Tennessee, and uses a combination of onsite gas generation and grid power. The gas turbine volume — equivalent to a medium-sized power station — places xAI alongside hyperscalers in a debate about whether AI infrastructure expansion is compatible with electricity decarbonisation timelines.