AI companies are building huge natural gas plants to power data centers. What could go wrong?

Major tech companies are making a massive, multi-gigawatt bet on natural gas to power the AI boom. Microsoft is partnering with Chevron and Engine No. 1 on a potential 5 GW plant in West Texas. Google is building a 933 MW facility with Crusoe in North Texas. Meta is adding seven new plants to its Hyperion data center in Louisiana, bringing its total capacity to 7.46 GW—enough to power the entire state of South Dakota. This concentrated push into the natural gas-rich southern U.S. is a direct response to the unprecedented power demands of AI model training and inference.

The rush has created a severe equipment shortage. According to consultancy Wood Mackenzie, turbine prices are projected to rise 195% by year-end compared to 2019, and new orders face six-year delivery timelines. Tech companies are gambling that AI's exponential power needs will persist and that natural gas is essential for success. However, this strategy carries significant risks. While U.S. gas supplies are currently plentiful, production growth in key shale regions has slowed. Tech firms may face price volatility and supply crunches, especially during extreme weather events that spike residential demand.

Furthermore, the industry's massive consumption could drive up electricity prices nationwide, as natural gas generates about 40% of U.S. power. While companies can initially shield themselves with "behind-the-meter" plants connected directly to data centers, scaling this approach could draw criticism from other industries and consumers. The move also represents a stark trade-off, locking in fossil fuel dependency for decades to come in pursuit of AI advancement, with potential repercussions for energy markets and climate goals.