Amazon confirmed this week that it is exploring plans to connect its private power system to the Texas grid, marking a shift for a campus once designed to operate entirely off it. The 8,000-acre site in Pecos County was conceived as a behind-the-meter build, a setup where generation and computing share a single wire and never touch the wider grid. That design now appears to be evolving, and the shift carries implications for how Texas manages the flood of gigawatt-scale AI campuses arriving across ERCOT territory. An Amazon spokesperson told Data Center Knowledge the company is “actively exploring” a transition to front-of-the-meter grid service. The same spokesperson said Amazon sees an opportunity for its energy infrastructure to eventually “add supply to benefit other customers.” Amazon has not disclosed a timeline for that transition, nor has it confirmed whether it has filed a large-load interconnection request with ERCOT.
A Behind-the-Meter Build With Front-of-the-Meter Ambitions
Behind-the-meter power generation lets a company build its own plant purely to serve its own facility, bypassing the queues, fees, and oversight that come with standard grid interconnection. Front-of-the-meter service works differently, tying generation directly into the grid so that surplus electricity can flow to homes, businesses, and other data centers nearby. Amazon initially structured its campus as the former: a closed loop between the Pacifico gas plant and the servers it feeds. Company statements now indicate that the closed loop can transition to grid-connected service as interconnection timelines allow. Amazon said the campus draws power from new on-site generation that will not raise electricity costs for Texas families and can eventually connect to the grid as interconnection timelines allow. That framing positions Amazon as a private-power developer whose campus can eventually connect to Texas’s broader power system.
The Texas Commission on Environmental Quality’s preliminary determination describes a natural gas-fired plant with a nominal output of 5,000 megawatts and 35 simple-cycle turbines. Pacifico Energy developed the plant under the GW Ranch name, and it sits about 17 miles north of Fort Stockton on greenfield land in Pecos County. Under the current design, the plant will provide electricity directly to an on-site AI data center and cannot initially sell or receive power from the local utility distribution system. The turbines will direct their output to the AI data center next door rather than the regional grid. However, the scale of the buildout, with 35 turbines feeding a single campus, dwarfs many municipal power systems across Texas. The arrangement illustrates just how far hyperscale AI developers are willing to go to secure power outside the constraints of traditional utility timelines.
Emissions Estimates Draw National Scrutiny
National outlets including The New York Times, The Verge, and Futurism have zeroed in on TCEQ documents that estimate allowable annual emissions of roughly 33 million tons of carbon dioxide equivalent. That figure has traveled widely, with many reports presenting it as a forecast of what the plant will actually emit once it begins operations. The number instead reflects a regulatory ceiling that assumes each turbine runs for 8,760 hours per year, covering every hour in a calendar year. Real-world operations rarely, if ever, reach that theoretical maximum because plants cycle down for maintenance, demand fluctuations, and grid conditions. The gap between a permitted ceiling and an actual emissions profile matters enormously for how the public interprets a project of this size. Meanwhile, the headline number continues to anchor coverage of the campus, regardless of how the plant ultimately operates.
Pacifico Energy describes GW Ranch as a far larger undertaking than the 5,000-megawatt plant alone, with plans for up to 7.65 gigawatts of gas generation across the site. The company’s public materials also point to 1.8 gigawatts of battery storage and as much as 750 megawatts of solar capacity across the project. Pacifico has targeted first power for the first quarter of 2027 and expects to bring 1 gigawatt online by 2028. The company projects more than 5 gigawatts of total capacity online by 2031, a timeline that would make GW Ranch one of the largest privately developed power complexes in Texas. What remains unclear is how the 7.65-gigawatt figure relates to the 5,000-megawatt, 35-turbine configuration that TCEQ describes in its preliminary determination. Pacifico’s public materials do not explain that relationship, leaving analysts to piece together the buildout sequence from permit filings rather than company disclosures.
Water Strategy Signals a Different Approach
Amazon said the campus will use custom cooling technology that minimizes water use wherever possible, a detail the company has emphasized in response to water-scarcity concerns common to West Texas. Later phases will rely on water that does not suit drinking or irrigation, with Amazon identifying non-potable sources such as brackish groundwater as part of its water strategy. Brackish groundwater sits too deep and carries too much salinity for most agricultural or municipal use, which is precisely why developers increasingly view it as a viable alternative. Amazon also said it is exploring produced water, a byproduct of oil and gas extraction, as another non-potable source for future cooling needs. Pecos County sits atop the Permian Basin, one of the most active oil and gas regions in the country, so local oil and gas operations produce substantial volumes of water.
Amazon said the campus is expected to create thousands of new jobs, a figure the company has cited in its public statements. The company has not disclosed its planned investment in the data center itself, leaving the project’s total cost a matter of speculation. Given the scale of the plant and the campus it feeds, industry estimates for comparable AI infrastructure projects run into the tens of billions of dollars. Amazon’s silence on the dollar figure stands in contrast to its relative openness about water sourcing and emissions framing. That selective disclosure pattern is not unique to Amazon; most hyperscalers guard capital expenditure details tied to individual sites. Nonetheless, the jobs claim gives local officials in Pecos County a tangible number to point to as the project moves through permitting and construction.
A Broader Pattern Across Hyperscale America
Amazon’s Texas campus is not an isolated experiment; it fits a pattern playing out across the country as hyperscalers race to secure power ahead of grid timelines that can stretch past five years. Meta and Microsoft have pursued similar behind-the-meter arrangements, pairing dedicated natural-gas generation directly with new data center campuses. Texas has emerged as a preferred landing spot for this model, while ERCOT has developed its own large-load interconnection framework as demand from AI and other large customers grows. That model comes with trade-offs, since projects operating outside the traditional grid connection process can raise separate questions about their eventual impact on Texas’s power system. Amazon’s potential pivot toward front-of-the-meter service suggests the company sees limits to how long a fully isolated power system can remain isolated. As AI compute demand accelerates, the line between private power plant and public utility asset is becoming harder to hold in place.
Amazon’s planned transition to front-of-the-meter service leaves open how its generation assets and data-center load would be treated inside ERCOT’s interconnection process. It also leaves open what transmission work would be required to physically link a plant built for a single customer into a shared grid. Texas regulators have spent much of the past two years wrestling with a wave of large-load requests from AI and crypto operations, many of which strain existing planning models. A campus of this size shifting even partially onto the public grid would test ERCOT’s capacity forecasts in ways few existing projects have. Therefore, the outcome of Amazon’s exploration could shape how future hyperscale campuses are permitted, financed, and connected across the state. For now, the company’s own words, “actively exploring” and adding supply “to benefit other customers,” suggest a strategic option is being kept open rather than a decision already made.
The Bottom Line
The GW Ranch project, and Amazon’s data center sitting beside it, represents a test case for how private power and public infrastructure can converge as AI demand reshapes the electricity map. Pacifico’s multi-gigawatt buildout, layered with gas, batteries, and solar, signals a level of ambition that goes well beyond serving a single customer. The unresolved questions, on interconnection status, transmission cost, and total investment, will matter as much to Pecos County residents as the emissions figures that first drew national attention. What began as a story about one company’s private power plant is turning into a broader conversation about who gets to build, own, and eventually share the infrastructure powering the AI era. Amazon’s next disclosures, whether through ERCOT filings or updated public statements, will determine how quickly that conversation moves from exploration to execution.
