# Amazon Is the Buyer Behind a 7.65-Gigawatt Texas Gas Plant Permitted for 33 Million Tons of Carbon Dioxide

The Pecos County project would deliver more than five gigawatts directly to an Amazon AI campus, with first power targeted for the first quarter of 2027.

- Published: 2026-08-09T07:02:50.434Z
- Canonical: https://polylog.news/ai/2026-08-09/amazon-is-the-buyer-behind-a-7-65-gigawatt-texas-gas-plant-p
- Publisher: Polylog (AI desk)
- Section: markets
- Sources: [Polylog editors](https://polylog.news), [Distilled](https://www.distilled.earth/p/scoop-amazon-is-behind-one-of-the-largest-planned-gas-power-plants-in-the-us), [TechXplore](https://techxplore.com/news/2026-08-amazon-massive-private-gas-centers.html)

Reporting this week [identified Amazon](https://www.distilled.earth/p/scoop-amazon-is-behind-one-of-the-largest-planned-gas-power-plants-in-the-us) as the offtaker, the buyer that has contracted for the plant's power, behind a 7.65-gigawatt natural gas plant that Pacifico Energy is developing in Pecos County, West Texas, alongside a company data center campus on the GW Ranch site. The Texas Commission on Environmental Quality has issued an air permit authorizing up to 33 million tons of carbon dioxide a year, and [independent coverage](https://techxplore.com/news/2026-08-amazon-massive-private-gas-centers.html) describes the site as among the largest privately built gas projects planned in the United States.

The engineering configuration matters for anyone modeling AI serving costs. The plant is specified with 35 gas turbines, with more than five gigawatts delivered to the campus, supported by 1.8 gigawatts of battery storage and 750 megawatts of alternating-current solar capacity. That mix is a response to a specific constraint: training and inference clusters need firm, dispatchable power at a scale and speed that grid interconnection queues in most of the country cannot supply, and behind-the-meter generation (power produced on-site, outside the utility grid) bypasses the queue.

Some early social coverage misplaced the site. A widely shared Telegram post [described the plant as located near Pittsburgh](https://t.me/aipost/7780) while correctly citing the 7.65-gigawatt figure and the 35-turbine count. The permitted site is in West Texas. Satellite imagery from late July shows land clearing underway, and three data center building permits were filed in early August, with first power targeted for the first quarter of 2027.

The broader pattern is that hyperscalers, the largest cloud computing operators, have stopped treating power as a procurement problem and started treating it as a construction problem. That solves the schedule constraint and creates a new exposure, because a campus tied to one private generation asset carries the operational and regulatory risk of that asset directly.

## What this means

Behind-the-meter gas generation lets Amazon convert a grid-interconnection delay into a capital expenditure, which favors operators with balance sheets large enough to build power plants and disadvantages smaller AI cloud providers who must queue for grid capacity. It also adds firm, contracted demand for gas turbines and pipeline capacity in the Permian region at a time when turbine order books are already extended, which supports pricing for equipment makers and gas producers serving that corridor. The offsetting risk is regulatory: a permit authorizing 33 million tons of annual carbon dioxide from one site is a visible target for litigation and for state and federal rulemaking, and a successful challenge would force Amazon back into the same interconnection queue it is trying to avoid.

## What to watch

- Whether environmental groups or Texas regulators challenge the air permit, since a delay would push the 2027 power date and the campus schedule with it.
- Whether other hyperscalers disclose similar private generation deals, which would confirm that self-built power is becoming standard practice rather than one company's workaround.
- Whether gas turbine lead times lengthen further, because that is the physical limit on how fast anyone can copy this approach.