Every ChatGPT-style query has a hidden cost, and Amazon’s newest project makes that cost impossible to ignore. Deep in Pecos County, West Texas, the company is building an AI-focused data center campus at a site called GW Ranch — and the numbers behind it are staggering enough to reshape the entire conversation about clean computing.
At the heart of the campus sits a 7.65GW natural gas power plant, driven by 35 gas turbines that will feed more than 5 GW of electricity straight into the server halls. To smooth out the peaks and valleys, Amazon is layering in 1.8 GW of battery storage and 750 MWac of solar capacity. On paper, that solar and battery mix nods toward sustainability. In practice, the gas turbines do the heavy lifting.
Here’s the headline figure regulators had to sign off on: an air permit authorizing up to 33 million tons of CO₂ emissions annually. That’s not a rounding error — it’s enough to push this single facility past the emissions of America’s largest existing polluter. A data center built to power artificial intelligence could, all by itself, become the country’s dirtiest single point source of greenhouse gases.
The irony is hard to miss. Amazon has spent years touting its clean-energy commitments, positioning itself as a corporate leader in renewables. A gas-fired megaplant emitting tens of millions of tons of CO₂ a year sits awkwardly against that reputation — a reminder that the compute demands of modern AI are colliding head-on with climate promises.
The project is very much real and moving fast. Satellite imagery from late July 2026 showed active land-clearing across the site, and in early August 2026 three separate construction permits for data center buildings were filed in Texas. First power is targeted for Q1 2027. As of now, the campus is under construction rather than operational — but the pace of work suggests Amazon intends to hit that timeline.
Why gas, and why so much of it? The answer is brutally simple. Training and running large AI models is enormously power-hungry, and the grid can’t always deliver gigawatts on demand where and when hyperscalers want them. Building your own power plant on-site sidesteps that bottleneck. It also means owning the emissions outright.
What GW Ranch really represents is the physical footprint of the AI boom made visible. Behind the sleek chatbots and image generators are 35 spinning turbines, a solar field, a battery bank the size of a small town’s demand, and a permit for 33 million tons of annual carbon. The cloud, it turns out, has a very heavy shadow — and West Texas is where it’s landing.
