US Data Centers Could Outburn Germany and Japan

The AI boom may turn US data centers into one of the world's largest consumers of natural gas. A new analysis warns that by 2035, American data centers could burn more gas than Germany and Japan combined, driven by the enormous electricity needs of training and running advanced AI models. That projection would have seemed far-fetched a few years ago, but hyperscale campuses are already forcing utilities to rethink load forecasts that were built for flatter demand. The result is a collision between the technology industry's rapid buildout and national goals for cutting emissions. Data centers need reliable, around-the-clock power, and natural gas plants can be built faster than nuclear, solar, or wind projects at the scale required. As a result, utilities are proposing new gas-fired generation to serve AI campuses, even as climate targets call for reducing fossil fuel use. The analysis suggests the sector's gas consumption could become a major factor in US energy planning, grid reliability, and carbon accounting. Supporters argue that gas is a pragmatic bridge that keeps the AI industry competitive while cleaner technologies mature. Critics counter that long-lived gas infrastructure could lock in decades of emissions and slow the transition to renewable power. The debate is also local: communities near planned data centers worry about air quality, water use, and rising electricity bills. Meanwhile, grid operators must manage unprecedented concentration of demand, sometimes in regions with limited transmission capacity. If the forecast proves accurate, AI's energy footprint will no longer be a side note in climate policy. It will be central to decisions about generation, infrastructure, and how quickly the US can decarbonize while remaining a leader in artificial intelligence.