An Amazon data center could have the worst polluting power plant in the country
To power its new West Texas data center, Amazon is investing in the construction of a new power plant that could be one of the largest single producers of greenhouse gases in the US, according to the New York Times. The new gas-burning plant in Pecos County, Texas has received significant investment from Amazon and, at least initially, would not be connected to the state's power grid.
The arithmetic of the AI boom has finally been put on paper, and the numbers are stark. Amazon is backing a natural-gas plant in Pecos County, Texas, designed to deliver 7.65 gigawatts almost exclusively to one data center. By the New York Times' accounting, that facility could become the single largest greenhouse-gas emitter in the country. The scale is not an accident; it is the logical endpoint of a race where compute capacity has become the only currency that matters.
What makes this particular project notable is not just its size, but its architecture. The plant is initially being built off the state grid, with its 35 turbines wired directly to the data center. That is a deliberate move. It sidesteps the delays, interconnection queues, and regulatory friction that have slowed grid-tied power projects across the country. For a company that treats time as the most expensive input, a private power source is not a luxury; it is a competitive necessity.
The environmental math is equally deliberate. A dedicated gas plant means the data center's carbon footprint is not hidden inside a utility's broader portfolio. It is visible, measurable, and attributable to a single customer. That transparency cuts both ways. It gives critics a precise target, but it also signals that Amazon has already priced in the reputational cost. The company is betting that the value of guaranteed compute outweighs the scrutiny.
This is the quiet reality of the AI infrastructure buildout. Hyperscalers have moved past the stage of buying renewable credits and negotiating with utilities. They are now building their own generation, choosing the fastest path to power regardless of the emissions profile. The market signal is unambiguous: when compute demand is urgent enough, the grid becomes an obstacle rather than a partner.
For the rest of the industry, this project is a benchmark. It establishes what a serious player is willing to spend, what it is willing to build, and what it is willing to emit to secure capacity. The debate over AI's energy appetite is no longer theoretical. It has a name, a location, and a turbine count.