Hardware

Amazon Backs Texas Gas Plant With Record Emission Permit

Amazon is backing a massive new natural-gas power plant in Texas to fuel its AI data centers, highlighting the severe environmental toll of the tech industry's rapid infrastructure expansion.

The Verge AI1 day agoHardware
Image: The Verge AI

Amazon has acquired a site in Pecos County, Texas, and is investing in a massive new natural-gas power plant called GW Ranch to fuel an adjacent data center. According to data tracker Cleanview, the state of Texas has granted the facility a permit to emit up to 33 million tons of carbon dioxide annually. This limit exceeds the emissions of the largest coal-fired power plant currently operating in the United States.

The GW Ranch project will feature 35 natural-gas turbines designed to generate 7.65 gigawatts of electricity. Crucially, the facility will operate off-grid initially, delivering its massive electrical output directly to Amazon's new data center complex rather than connecting to the Texas power grid. While power plants rarely reach the absolute maximum emissions allowed by their permits, the scale of the GW Ranch project underscores the immense energy demands of modern artificial intelligence infrastructure.

This move represents a significant hurdle for Amazon's Climate Pledge, which was championed by founder Jeff Bezos with the goal of achieving carbon neutrality by 2040. Instead, the company's greenhouse gas emissions have risen for several consecutive years, driven largely by the power-hungry nature of AI workloads. Addressing this tension, Amazon spokesperson Margaret Callahan noted that "the world looks different now" compared to when the pledge was established, though she maintained that the company's core commitment remains unchanged.

For AI practitioners and developers, this development highlights a growing ethical and operational dilemma. As tech giants like Google, Meta, and Amazon increasingly turn to fossil fuels to power next-generation model training and inference, software engineers may face mounting pressure to optimize algorithmic efficiency. Practitioners must weigh the computational benefits of massive, resource-intensive models against the real-world carbon footprint of the infrastructure supporting them, potentially shifting the industry's focus toward energy-efficient architectures.

This is our own summary of reporting by The Verge AI

More in Hardware