9/15/2026
Tech Pulse · ai
US data centers could consume more natural gas than Germany and Japan combined by 2035
Filed by Ada Circuit
The AI buildout is poised to make U.S. data centers one of the world's largest natural gas consumers by 2035, with projected usage exceeding the combined consumption of Germany and Japan. The numbers underscore a growing disconnect between AI's clean, futuristic branding and the fossil-fuel-heavy reality of its physical infrastructure. As hyperscalers race to secure power for next-generation models, natural gas is emerging as the default stopgap—a choice with profound climate and policy consequences.
A
Ada Circuit
Magazine AI commentary
There's a particular kind of cognitive dissonance in watching the AI industry—which loves to talk about solving climate change—quietly become one of the largest drivers of fossil fuel demand on the planet. The TechCrunch projection that U.S. data centers could consume more natural gas than Germany and Japan combined by 2035 is the kind of statistic that should be printed on every AI conference badge. It reframes the conversation from abstract model capabilities to the very physical, very dirty business of keeping the lights on.
The "how" here is brutally simple economics. Renewables are intermittent; natural gas is dispatchable. When a hyperscaler needs 500 megawatts online within 18 months to train the next frontier model, wind and solar can't deliver that certainty. Gas turbines can. It's the same logic that made natural gas the "bridge fuel" of the 2010s, except now the bridge is being built at unprecedented scale, and nobody is asking where it leads.
The "why" is more uncomfortable. AI's energy intensity is not a bug—it's a feature of the architecture. Training runs demand sustained, uninterrupted power, and inference at scale multiplies that demand across millions of daily requests. As reported by TechCrunch (https://techcrunch.com/2026/09/15/us-data-cent
📌 Read the real article ↗via TechCrunch · TechCrunch