Amazon Texas gas plant could become top US CO2 source

The Amazon Texas gas plant is a 35-turbine, 7.65-gigawatt natural-gas facility in Pecos County, Texas, built to power a new AI data center campus that Amazon has confirmed it is funding. Regulatory permits for the project authorize as much as 33 million metric tons of CO2 a year, a ceiling that would make it the largest single source of climate pollution among US power plants. The confirmation, made in early August, sits in direct tension with Amazon's Climate Pledge and its net-zero-by-2040 commitment.

At the permitted maximum, the Amazon Texas gas plant would emit roughly double the annual CO2 of the current dirtiest US power station, the coal-fired James H. Miller Jr. plant in Alabama, which releases about 16 million tons a year. That comparison is the clearest measure of scale: one data center campus, supplied by a plant that does not connect to the regional grid, could out-pollute every other power station in the country. Amazon already operates more data centers worldwide than any other company, and this project is the largest single power commitment in its fleet.

What the permit authorizes

The Amazon Texas gas plant is built around 35 natural-gas turbines delivering up to 7.65 GW to a single Amazon Web Services campus in West Texas. Pacifico Energy is building the plant, and Amazon will buy its output through a power purchase agreement rather than owning the generation outright. That structure moves construction and operating risk to Pacifico while Amazon locks in dedicated capacity at a scale the interconnection system cannot currently offer.

The permit numbers are the core of the story: up to 33 million tons of CO2 annually, set by regulators as an authorized ceiling rather than a forecast of typical operation. The distinction matters because actual emissions depend on how hard the turbines run. But the ceiling is what was approved, and a facility sized for gigawatt-scale AI loads has a high operating floor. A gas plant commissioned now is typically expected to operate for two to three decades, which puts its emissions profile well past Amazon's 2040 net-zero deadline.

The purchase-agreement structure also raises an accounting question. Pacifico owns the generation, Amazon buys the power, and it is not yet clear how the plant's CO2 will be counted against Amazon's net-zero path. Depending on how the company reports the deal, the emissions could sit on either ledger, which will matter for how investors measure progress toward the 2040 target.

How the Amazon Texas gas plant compares

By the numbers in the permit filings, the Pecos County facility would leapfrog every existing US power plant in CO2 output. The Miller Jr. coal station in Alabama, currently the largest single emitter among US power plants at roughly 16 million tons a year, would be outdone by a margin of more than two to one. The project has also been described as the largest gas-fired power plant in the country, which sharpens the contrast: the newest technology in the US, powered by one of the oldest fuels, at record scale.

Why on-site gas beats the grid queue

The strategic logic behind the Amazon Texas gas plant is straightforward. Interconnection queues in most US electricity markets, including Texas, stretch for years, and transmission capacity for new gigawatt-scale loads is scarce. Building generation on site lets Amazon bypass that queue entirely. The plant is designed to serve the campus directly rather than feed the regional grid, sidestepping both interconnection timelines and the cost of new transmission lines.

In terms of process, the project moved quickly: the filings became public this month, Amazon confirmed its investment within days, and construction responsibility sits with Pacifico Energy. That pace is only possible off the grid, where the project avoids the interconnection studies and transmission upgrades that routinely delay grid-connected plants for years.

The trade-off is that the fastest path to power is also the least clean. Grid power in Texas increasingly comes from wind and solar, while a dedicated gas plant locks in fossil-fuel generation for the life of the asset. On-site solar and wind cannot guarantee firm supply for a campus drawing power around the clock, and storage at this scale is not yet economical. Gas delivers certainty on demand, which is precisely why the industry keeps choosing it.

The economics favor on-site gas today; the emissions ledger does not. That is the core tension of the AI power buildout: hyperscalers are choosing speed and certainty over the emissions profile their own climate targets imply, and Pecos County is the most extreme example of that choice so far.

The collision with Amazon's climate commitments

Amazon's Climate Pledge and its net-zero-by-2040 target treat carbon reduction as a company-wide obligation. The Amazon Texas gas plant alone is authorized to emit 33 million tons a year, which complicates that arithmetic before any other part of the business is counted. At ceiling, the plant would emit more CO2 than the current worst US power-plant emitter produces, roughly doubling the largest existing case.

The alternatives available to hyperscalers all carry costs. Grid power means multi-year queues. On-site renewables are intermittent and cannot carry a round-the-clock gigawatt load without storage that is not yet economical at this scale. Nuclear and advanced geothermal remain years from commercial deployment at this size. Gas is the only option that delivers now.

The result is the dirty-power economics at the center of the AI buildout: private revenue from AI compute, public cost in record-setting CO2 emissions. The climate cost of the Amazon Texas gas plant lands on a public ledger, while the power it generates flows to one company's customers. The 33-million-ton ceiling is the price of speed, and it is paid in the atmosphere rather than on Amazon's balance sheet.

The significance of the scale is that Amazon is the operator other hyperscalers measure themselves against. If the largest data center fleet in the world needs a plant of this size to power AI, the same template becomes the default answer for competitors facing identical queues. Pecos County is the purest case yet of the AI buildout's dirty-power economics, and its permit tonnage sets a benchmark that future projects will be compared against.

What to watch

Three questions will determine how much the project matters in practice. First, how close to the 33-million-ton ceiling the Amazon Texas gas plant actually operates, since utilization will follow AI demand and grid conditions. Second, whether a permit of this size draws regulatory or political attention that slows similar proposals elsewhere. Third, whether Amazon pairs the plant with additional renewable purchases, which would keep the emissions on its books while reducing the net carbon impact.

The verdict on the Amazon Texas gas plant does not question whether AI needs power. It questions how that power is sourced when clean supply cannot keep pace. On-site gas is the fastest option available, and the permit ceiling shows what that option costs in carbon: a record 33 million tons a year, for the operator of the world's largest data center fleet.

Why this matters

The Amazon Texas gas plant turns the AI buildout's power problem into a climate problem with a specific price tag: up to 33 million tons of CO2 a year from a single facility. For companies and investors in AI infrastructure, the distance between corporate net-zero pledges and the generation actually built to deliver AI is now visible in concrete tonnage, and it is a gap every hyperscaler will eventually have to answer for.

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.