Amazon Backs Texas Gas Plant Permitted to Emit 33 Million Tons of CO2 Annually
Amazon is funding a massive off-grid natural gas plant in Texas to power an AI data center, securing permits to emit more carbon dioxide annually than any coal plant in the United States.
Amazon has backed construction of a large off-grid natural gas power plant in Pecos County, Texas, to supply electricity to a new artificial intelligence data center. According to reporting from The New York Times and data tracking firm Cleanview, the facility has secured state permits allowing it to emit up to 33 million tons of carbon dioxide annually.
The proposed facility, known as GW Ranch, will feature 35 natural-gas turbines capable of generating 7.65 gigawatts of electricity. Initially, the plant will operate entirely disconnected from the main Texas electrical grid, delivering power directly to Amazon's computing infrastructure. Amazon has confirmed its ownership of the site and its intent to purchase the electricity generated there.
The permitted emissions ceiling of 33 million tons of carbon dioxide exceeds the maximum potential output of any operating coal plant in the United States. Environmental analysts note that industrial plants rarely operate at their maximum allowed pollution limits, but the threshold itself reflects notably permissive state regulatory standards. The project arrives as federal energy policy shifts toward easing environmental restrictions on fossil-fuel power generators.
Amazon is not alone. Rivals including Google and Meta have increasingly turned to natural gas and legacy power assets to meet the intense electricity demands of AI workloads. Computing clusters required to train and deploy complex AI models demand continuous, uninterrupted baseload power, and public electrical grids frequently cannot deliver it without major structural upgrades.
The project also puts pressure on Amazon's existing climate commitments. In 2019, then-chief executive Jeff Bezos co-founded the Climate Pledge, committing the company to net-zero carbon emissions by 2040. Amazon's overall greenhouse gas emissions have nonetheless expanded in recent years alongside capital investments in cloud and AI infrastructure. Addressing the GW Ranch project, Amazon spokeswoman Margaret Callahan said that while the world looks different than when the climate pledge was established, the company remains committed to its long-term goals.
The Cognarah Angle
The rapid expansion of AI is forcing a stark choice between aggressive computational growth and corporate climate commitments. For years, major cloud providers marketed their infrastructure on renewable energy guarantees and net-zero timelines. The decision to build private, gas-fired power plants disconnected from public grids is a signal that immediate demand for raw compute now overrides those roadmaps.
The off-grid model is strategically clever and environmentally convenient. By building generation capacity directly adjacent to data centers, technology companies sidestep lengthy regulatory delays and grid capacity constraints. But this approach privatizes the power supply while distributing the environmental cost broadly. Granting a single data center project a pollution ceiling higher than any US coal plant lays bare the sheer energy appetite underpinning current AI development.
Callahan's statement that the world looks different today is not a climate strategy. It is an admission that clean energy infrastructure has not kept pace with AI power consumption, and that Amazon is unwilling to let that gap slow its expansion. The Climate Pledge was never legally binding, and this project tests exactly how much that matters to Amazon's leadership.
If building the next generation of AI infrastructure requires reviving fossil fuel dependence at this scale, the industry has a serious problem it cannot footnote away with net-zero pledges.
Reporting sourced from The Verge. Analysis and Cognarah Angle are Cognarah's own.
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
