The current trajectory suggests a deepening reliance on natural gas for a significant portion of the country's future power needs. Expect continued investment in pipeline infrastructure, particularly in regions with high data center concentration like Texas, Arizona, and New Mexico. The tension between rapid technology growth and environmental commitments will likely intensify, forcing policymakers and utility companies to confront difficult trade-offs. This dynamic could lead to increased scrutiny of data center energy consumption, potential regulatory shifts, and a renewed debate about the true environmental cost of the AI revolution.
Image: courtesy of Wired
The AI Power Surge: How Data Centers Are Forcing a Reckoning with US Energy Goals
The rapid expansion of artificial intelligence data centers across the United States is driving an unprecedented surge in electricity demand, primarily met by new natural gas power plants and infrastructure. By mid-2026, the gas power capacity dedicated to U.S. data centers nearly doubled from the end of 2025, reaching over 189 gigawatts. This swift shift towards fossil fuels to fuel the digital economy is creating a direct conflict with the nation’s climate targets, raising questions about future emissions, energy policy, and who ultimately bears the cost.
Outlook
Background
The scale of the U.S. data center build-out is staggering. Power demand for these facilities is projected to increase by 400 terawatt-hours by 2030, a figure that represents a substantial portion of the nation's total electricity consumption. To put this in perspective, data centers currently account for 3-4% of U.S. power demand but are expected to consume 11-12% by the end of the decade.
This growth is not abstract; it translates into concrete infrastructure projects. By mid-2026, gas-fired power capacity dedicated to U.S. data centers had already surged past 189 gigawatts (GW), an almost twofold increase from 97 GW recorded at the close of 2025. A gigawatt measures instantaneous power, equivalent to the output of a large power plant. This makes the U.S. the global leader in developing new gas power specifically for data centers.
Companies like Energy Transfer are actively expanding regional pipelines. One notable project is their 2.3 billion cubic feet per day (Bcf/d) Desert Southwest Expansion, designed to funnel Permian natural gas to demand hubs in states like Arizona and New Mexico. This indicates a broader-based growth in natural gas demand beyond just a few isolated areas.
The push extends beyond traditional grid connections. A recent review by the Environmental Integrity Project identified 74 proposed or planned gas-fired power plant projects designed to supply electricity directly to data centers, bypassing the standard grid connection process. These "behind the meter" plants, if all were built, could generate 143 GW of electricity and contribute an estimated 662 million tons of greenhouse gas emissions annually. Such projects suggest a concerted effort by data center operators to secure reliable power, even if it means constructing their own dedicated fossil fuel infrastructure.
Major tech players are central to this expansion. On August 7, 2026, it was reported that Entergy, a large utility, plans $12.9 billion in new gas plants specifically to power Meta's operations. This is a clear sign of the direct link between hyperscale data center demand and fossil fuel investment. This raises immediate concerns about who pays for this capacity. In the Southeast, for instance, ratepayers may ultimately bear the cost burden of these extensive new gas infrastructure projects.
Despite the ambitious plans, some projections appear overly optimistic. One report suggests that if all data centers announced for 2025 to 2030 were built, they would consume over 90% of all new computer chips produced globally. This is deemed "unrealistic," given that the U.S. currently uses less than 50% of the world's chips, with other nations also planning significant increases in their chip demand. This implies that some announced projects may be scaled back or face delays, though the underlying demand trend remains robust. An additional 60 GW of data center capacity is anticipated by 2035, with power use for data centers reaching up to 450 terawatt-hours (TWh) by then. A terawatt-hour measures energy consumption over time, like the amount of electricity a small country might use in a year.
See also
Precedents
The current surge in data center power demand, while unprecedented in its scale and speed, echoes historical patterns of industrial expansion driven by new technologies. From the early days of electrification to the internet boom of the late 20th century, periods of rapid economic and technological growth have consistently led to corresponding spikes in energy consumption. What makes this era distinct is the simultaneous global commitment, at least on paper, to decarbonization.
In the past, new energy demands were often met by the most readily available and cost-effective sources, which for decades meant fossil fuels. The rise of manufacturing in the industrial revolution, the widespread adoption of automobiles, and the post-war economic expansion all fueled massive build-outs of coal, oil, and gas infrastructure. Each wave brought with it debates about environmental impact, resource allocation, and economic benefits.
The current situation also mirrors the early days of cloud computing, where initial estimates of power consumption were often underestimated. As services scaled, so did the energy footprint, leading to a gradual recognition of data centers as major energy consumers. However, the AI revolution is accelerating this demand at a pace that is challenging existing energy supply models and grid capabilities, pushing utilities towards rapid deployment solutions, often involving natural gas. This is not simply a linear increase; it is an exponential jump driven by the intensive computational requirements of AI models.
The expansion of natural gas power for data centers represents a direct collision between the promises of the AI era and the pressing realities of climate change. On one side, AI offers transformative potential across industries, from healthcare to scientific research. On the other, its foundational infrastructure is becoming a significant new source of greenhouse gas emissions, potentially derailing efforts to transition to a carbon-free energy system.
For environmental advocates and policymakers, the implications are stark. The projected 662 million tons per year of greenhouse gas emissions from proposed "behind the meter" gas plants alone would represent a massive setback for U.S. climate goals. This is not just about meeting current targets, but about the long-term trajectory of the energy mix. If the fastest-growing sector of the economy becomes heavily reliant on fossil fuels, the path to decarbonization becomes significantly steeper and more costly.
Economically, this shift has profound consequences for utility companies, energy producers, and consumers. Midstream companies see a clear opportunity for pipeline expansion, while utilities face immense pressure to ensure reliable power for their largest customers, often at the expense of integrating slower-to-deploy renewable sources. Ratepayers, particularly in areas like the Southeast, could see their electricity bills rise to cover the capital expenditures for these new gas plants, even if they personally derive little direct benefit from the data centers themselves. This raises questions of equitable cost distribution and the broader public good.
Furthermore, the scale of demand from data centers could create new vulnerabilities in the energy supply chain, potentially leading to increased energy price volatility and concerns about grid stability. It forces a national conversation about energy priorities: how much growth in compute power can be sustained, and at what environmental and economic cost? The choices made in the next few years will shape not only the future of AI but also the future of the American energy landscape and its role in global climate action.
Scenarios
Analysis1. Accelerated Natural Gas Reliance and Missed Climate Targets: The current trend could solidify natural gas as the primary fuel for rapid data center expansion, especially given its relatively quick deployment compared to large-scale renewable projects. This would mean a significant increase in U.S. greenhouse gas emissions, making it considerably harder for the nation to meet its carbon reduction targets. Utility companies and energy providers, incentivized by the immediate need for reliable power and existing infrastructure, could continue to prioritize gas, potentially leading to further investment in pipelines and new gas-fired plants. This outcome would likely spark intensified protests from environmental groups and put political pressure on administrations to reconcile conflicting energy and climate policies.
2. Increased Scrutiny and Policy Interventions for Data Center Energy: As the environmental and economic costs become clearer, there could be a strong push for regulatory changes and policy interventions. This might include mandates for data centers to source a higher percentage of their power from renewable sources, incentives for developing and deploying energy-efficient AI hardware, or even carbon taxes on data center emissions. State and federal agencies could implement stricter permitting processes for new fossil fuel plants intended for data centers, or prioritize grid upgrades to integrate more intermittent renewable energy. This path would likely slow down some data center developments but could accelerate innovation in sustainable computing and renewable energy integration. The question of who pays for new infrastructure, particularly for "behind the meter" plants, would also likely become a central regulatory battle.
Timeline
2023
Data Center Power Consumption
U.S. data centers consume approximately 130 terawatt-hours of electricity, representing 3-4% of total U.S. power demand.
Late 2025
Gas Power Capacity Baseline
Gas power capacity for U.S. data centers stands at 97 gigawatts.
January 2026
Largest Gas Project Approval
Pacifico Energy’s GW Ranch in West Texas is approved, becoming the largest gas power project in the U.S.
Early 2026
New Gas Plant Approvals
Several other large-scale natural gas power projects are approved to meet growing demand.
Mid-2026
Gas Power Capacity Surge
Gas power capacity for U.S. data centers surges to over 189 gigawatts, nearly doubling from the end of 2025.
August 7, 2026
Utility Investment and Ratepayer Concerns
Reports highlight that data centers drive 183 GW of U.S. power demand and that Entergy plans $12.9 billion in gas plants for Meta, with potential cost burdens for Southeast ratepayers.
2025-2030 (Projected)
Chip Consumption Outlook
If all announced data centers are built, they would consume over 90% of new global chips, a projection deemed "unrealistic."
By 2030 (Projected)
Future Power Demand Projections
U.S. power consumption is expected to increase by 400 terawatt-hours due to data center growth. Data centers are projected to consume 11-12% of total U.S. power demand.
By 2035 (Projected)
Long-Term Capacity Growth
An additional 60 gigawatts of data center capacity is anticipated in the U.S., with power use reaching up to 450 terawatt-hours.
Frequently Asked Questions
A gigawatt (GW) measures instantaneous power capacity, like the maximum output of a large power plant. A terawatt-hour (TWh) measures energy consumption over time, similar to how much electricity a city uses in a year. One TWh is a trillion watt-hours, representing a massive amount of energy.
Discussion
Be the first to share your thoughts.