Texas is turning the electricity needs of artificial intelligence into a new market for natural gas generation. The state now has 122 gigawatts of gas-fired capacity in development, a 51% increase from the beginning of 2026, according to research. Nearly two-thirds of that pipeline, or about 77 gigawatts, targets data centers directly, making Texas the clearest expression of a broader American bet on gas-backed computing infrastructure. The U.S. now has 189 gigawatts of gas-fired capacity in development aimed directly at powering data centers, nearly twice the level recorded six months earlier. The numbers point to a growing development strategy in which developers plan new gas-fired generation specifically to meet data-center electricity demand. That distinction matters because the organization separately identifies gas plants that will directly power data centers, including projects designed to provide electricity specifically for those facilities.

The Texas pipeline is particularly striking because developers are pursuing projects at a scale rarely associated with a single industrial customer. The 12 largest gas power projects planned in the state are intended to serve data centers, according to Global Energy Monitor’s latest tracking. GW Ranch Energy Center near Fort Stockton illustrates the scale, with its permit allowing roughly 33 million tons of annual greenhouse gas emissions and Amazon acquiring the project earlier this month. Fermi America‘s Project Matador in the Texas Panhandle proposes 90 combined-cycle gas turbines, 16 cooling towers and six emergency generator engines, with projected greenhouse gas emissions of about 24 million tons a year. Nexus Data Center Hubbard in Southeast Texas represents another large installation, with permitting documents projecting approximately 22 million tons of annual greenhouse gas emissions. 

Gas Is Becoming an AI Infrastructure Tool

The more consequential development sits underneath those headline projects: an increasing share of proposed gas generation is being planned specifically to supply data centers. That approach can reduce dependence on grid interconnection schedules, which have become a major constraint as utilities process enormous new loads. A dedicated gas plant can provide a data center with an onsite source of dispatchable electricity rather than relying entirely on power supplied through the public grid particularly when AI clusters require sustained high utilization rather than occasional industrial demand. It can also allow developers to pair generation directly with a data center rather than depending entirely on the timing and availability of a conventional grid connection. For hyperscale campuses, that changes power from a utility procurement problem into an infrastructure design decision.

That proposition becomes more relevant as developers confront lengthy grid-connection processes and growing competition for available power capacity. Texas has encouraged developers to provide their own generation as policymakers confront rising electricity demand and concerns about the burden large data centers can place on the ERCOT system. Gov. Greg Abbott has pushed developers toward self-supply as a way to limit pressure on the broader grid. His position reflects a basic political and commercial calculation: if AI companies want enormous quantities of electricity on accelerated schedules, they may need to bring some of that electricity with them. The strategy can allow developers to supply a data center independently of the public grid while taking on the operating requirements associated with onsite generation. Yet the apparent simplicity of building gas generation hides a difficult question about how many proposed campuses will actually materialize. 

Turbine Shortages Are Changing Project Designs

Natural gas has an advantage that solar and wind cannot easily replicate for a rapidly expanding computing campus: dispatchable generation can follow a continuous load without depending on weather conditions. The problem is that the gas industry cannot manufacture large numbers of conventional combined-cycle turbines overnight. Global Energy Monitor says turbine production has become a bottleneck, with manufacturer backlogs stretching into 2030. Developers therefore have started examining smaller gas engines and simple-cycle turbines that can reach the market faster, even though they generally produce more emissions for each unit of electricity generated. For a data center operator facing long equipment lead times, faster-deploying generation can provide an alternative to waiting for larger turbine systems, although the available technologies can carry higher emissions per unit of electricity.

The technology mix already shows how quickly that tradeoff is spreading. Engines and simple-cycle gas turbines now account for nearly half of the generating technologies proposed for gas projects tied to data centers across the United States. Kilby Power Plant in West Texas, for example, proposes two combined-cycle gas turbines alongside 15 simple-cycle turbines and projects roughly 12 million tons of annual greenhouse gas emissions. Its permit application describes the simple-cycle machines as critical for uninterrupted behind-the-meter operation that can respond to the changing electricity demand of onsite data center customers. That configuration resembles an electrical insurance policy as much as a conventional power station, with fast-response equipment providing another layer of control around a highly valuable computing load. Meanwhile, the proliferation of engines introduces another issue for developers: an architecture optimized for rapid deployment may create a more complicated long-term operating fleet. 

Gas Could Become AI’s Bridge—and Its Burden

The immediate attraction of gas is straightforward: it offers dispatchable electricity at a time when data-center developers are confronting rising power demand, long grid-connection queues and turbine supply constraints. Texas developers are therefore building projects around two observable market conditions: the availability of natural-gas generation and expectations for continued growth in data-center electricity demand. Jenny Martos, project manager for the Global Oil and Gas Plant Tracker, captured that uncertainty directly: “This frenzy of gas proposals is running headlong into the hurdles of an already tight gas market.” “It is nearly impossible nowadays to guess what is a pie in the sky proposal, and what has a real chance of getting built.” 

For data center developers, the lesson is not simply that gas is returning to Texas. The deeper shift is that power availability has become a defining component of compute strategy, capable of determining where campuses get built, how quickly they can operate and what infrastructure they must own. The largest projects now resemble integrated energy systems designed around computing loads rather than conventional data centers that happen to require electricity. That architecture may help operators bypass some of the industry’s most painful grid constraints, but it transfers those constraints into fuel procurement, turbine supply, emissions exposure and long-term asset utilization. Texas has effectively become a laboratory for that tradeoff, with a large pipeline of proposed gas infrastructure designed to supply electricity directly to data centers as developers seek alternatives to conventional grid connections.