Amazon’s planned Texas data centre would be powered by a gas plant permitted to emit up to 33 million tons of CO2 a year.
TechCrunch reported that Amazon is investing in an on-site power plant for a planned data centre in Pecos County, Texas, and that the project could become the largest source of climate pollution in the United States. The Verge similarly reported that the Amazon-owned site has a permit allowing releases of up to 33 million tons of carbon dioxide per year.
In plain terms, this is not just a new warehouse-sized computing building. The reported plan pairs a large data centre with its own dedicated gas-fired electricity supply. That matters because AI and cloud facilities need vast, continuous power, and the sources say this project shows how meeting that demand can shift questions from computing capacity to fuel, emissions, and local infrastructure.
What Amazon is building in Texas
According to The Verge, the project centres on a new West Texas data centre at an Amazon-owned site in Pecos County. It says Amazon has confirmed that it bought the site and plans to buy power from a project called GW Ranch.
The reported power arrangement is unusual in one important way: rather than mainly relying on electricity already flowing through the state network, the facility would use new generation built on site. The Verge says the gas-burning plant would have 35 natural-gas turbines and produce 7.65 gigawatts of electricity, primarily for the new data centre. It also reports that, at least initially, the plant would not be connected to the Texas power grid.
Amazon’s own public position, as quoted by TechCrunch, is that the data centre will “be powered by new on-site generation that won’t raise electricity costs for Texas families.” That frames the project as separate from household electricity supply, although the two reports focus mainly on the emissions implications rather than on how the wider power market may respond.
Why the emissions permit is drawing attention
The core concern in both reports is the scale of the plant’s permitted pollution. The Verge says Texas granted GW Ranch a permit allowing emissions of up to 33 million tons of CO2. TechCrunch, citing The New York Times, says that figure would be higher than for any other power plant in the US.
That does not mean the plant will necessarily emit the full amount allowed by its permit. The Verge explicitly notes that plants rarely emit as much greenhouse gas as their permits allow. Still, it says the limit remains notable because it shows how much pollution regulators would permit from the site.
Both reports therefore describe the project carefully as a potential or possible top polluter, not a confirmed one in operation. The distinction matters. What is established by the sources is the permit level, the gas-based design, Amazon’s investment, and Amazon’s plan to purchase power from the site. What remains uncertain is the eventual real-world emissions level if the project proceeds.
How this connects to AI data centre demand
The articles place the Texas plan in a broader shift in how large computing sites are being powered. The Verge says many companies, including Meta and Google, have started building their own power plants and are turning to gas, coal, and other non-renewable sources as the demands of AI data centres have risen sharply.
Here, “AI data centre” effectively means a computing site designed to run and support artificial intelligence systems, which need dense clusters of servers and substantial electricity. When those facilities grow faster than existing power supply can keep up, companies may seek dedicated generation close to the site. In this case, the reported answer is natural gas rather than a renewable source.
TechCrunch adds that data centres are facing growing political opposition for several reasons, including their effect on electricity costs. Its report also says Amazon’s carbon emissions rose 16% last year, which it presents as evidence that AI-related growth is already affecting the company’s climate targets.
The sources do not provide detailed evidence on how other businesses are responding to this specific Amazon project. But they do support a narrower point: major technology companies are actively pursuing large, dedicated power arrangements to meet AI-related computing demand, while the resulting fuel choice is attracting political and environmental scrutiny.
What this means for Amazon’s climate commitments
Both reports connect the Texas project to Amazon’s Climate Pledge, but they describe that pledge differently. TechCrunch says Amazon pledged to eliminate its carbon emissions by 2040, while The Verge says Jeff Bezos pledged to make the company carbon neutral by 2040.
The sources therefore present a clear tension. On one side is Amazon’s continuing climate commitment. On the other is a reported infrastructure plan that would depend on a large gas-fired power source for a new data centre. Amazon spokeswoman Margaret Callahan, quoted in both reports via The New York Times, said: “The world looks different now than when we co-founded the climate pledge,” and also said, “Our commitment hasn’t changed.”
The articles do not resolve that tension, and neither can this piece. What they establish is that Amazon is arguing it can continue backing its climate goals while also supporting new on-site gas generation for a major data centre project.
Why it matters
This development matters because it shows how AI and cloud expansion can depend on energy projects that look less like abstract digital infrastructure and more like heavy industry. In these reports, the key constraint is not server demand alone, but how a company secures enough electricity to run a very large computing site.
For IT leaders and businesses, the strongest conclusion supported by these sources is that power sourcing is becoming a central part of AI infrastructure strategy, and that those choices can collide with climate commitments. The sources do not show broad enterprise reaction to this specific project, but they do show major technology companies pursuing dedicated power projects to support AI growth, even as those plans attract political and environmental pressure.
What remains open is how much this Texas plant would actually emit if built and operated, whether its initial separation from the grid changes over time, and how far similar power arrangements spread across the industry. A likely implication is that future AI infrastructure growth may be shaped not just by chips, servers, and land, but by what kind of power companies can secure and what regulators are willing to permit.