
AI data centres add 56 GW gas capacity
- At least 46 AI-focused data centres are pursuing 56 GW of on-site natural gas generation to avoid grid connection delays.
- The trend is driving a surge in demand for gas-fired power plants, turbines and pipeline infrastructure.
- Developers say dedicated power generation can accelerate deployment, while critics raise concerns about long-term fossil fuel use.
Data-centre developers are pursuing at least 56 GW of on-site natural gas generation across 46 projects as artificial intelligence infrastructure expands faster than electricity grids can accommodate new demand.
The strategy, known as behind-the-meter generation, allows operators to bypass utility interconnection queues that can take several years to complete.
The trend has attracted major technology companies, with xAI operating gas turbines at its Colossus facilities and OpenAI's Stargate project in Texas targeting more than 1 GW of capacity supported by natural gas generation.
Texas has become the centre of the buildout, with approximately 80.6 GW of gas-fired power capacity under development and about 40 GW earmarked for data-centre projects.
Developers argue that dedicated power infrastructure can accelerate AI deployment, while environmental groups have warned that new gas plants could lock in fossil fuel use for 20 to 30 years; following the developments the share price reactions of the companies involved were mixed and not disclosed.
The rapid growth in demand has contributed to a near tripling of US gas-fired generation capacity under development to approximately 252 GW, with more than one-third linked to data centres.
Recent examples include MARA Holdings, which announced a US$1.5 billion acquisition of Long Ridge Energy that includes a 505 MW gas plant and more than 1,600 acres for future AI and data-centre expansion.