Despite being the epitome of modern technology, artificial intelligence (AI) runs primarily on antiquated fossil fuels. As tech giants race to build data centers, the energy required to sustain this AI boom locks the U.S. into a dangerous dependence on natural gas—one fraught with dire consequences for the environment.
Most AIs are large language models (LLMs), software tooled to analyze linguistic patterns in billions of text sources and predict the next word based on context. The most capable LLMs, such as ChatGPT, Claude, and Gemini, generate text with impressive accuracy, but they come with a significant energy cost.
AI consumes energy primarily as electricity in two significant ways: training and inference. During training, billions of parameters—representing the various probabilities in word prediction—are adjusted using vast amounts of data. The high-performance computing chips tasked with this training require immense power. For example, the initial training of OpenAI’s GPT-4 used the equivalent energy of San Francisco’s total consumption over three days, just to keep the hardware running, including electricity, cooling, and maintenance.
Interestingly, around 80 to 90 percent of the total energy consumption stems from inference, the process where users interact with the AI. AI models rely on graphics or tensor processing units (TPUs) for fast parallel processing, enabling many computations to occur simultaneously. These chips consume significantly more power than central processing units typical in modern electronic devices. To put this into perspective, querying an LLM can cost ten times more energy than a standard Google search and emit 340 times more carbon dioxide.
Despite these alarming statistics, the International Energy Agency forecasts that by 2026, AI electricity will still account for just around 3% of global electricity demand. Moreover, it’s notable that AI-related chips have become over 100 times more energy efficient in under two decades. Thus, the primary issue with AI’s energy consumption isn’t the quantity but rather its source.
A significant challenge in sourcing energy sustainably is that AI data centers require “constant power, 24/7, 365 days a year,” according to Rahul Mewawalla, CEO of Mawson Infrastructure Group. This steady demand precludes many renewable energy sources, like solar and wind, which are subject to erratic weather conditions. Nuclear energy, often considered a reliable option, faces resistance due to the lengthy timeframes for plant construction, making it impractical for immediate energy needs.
In contrast, natural gas is abundantly available, and as of 2024, its prices reached historic lows in the U.S. As AI companies seek predictable and easily scalable energy sources, natural gas has emerged as the go-to option, often at the expense of environmental stewardship. A study by Harvard’s T.H. Chan School of Public Health revealed that energy used by AI data centers releases 48% more greenhouse gases than the average U.S. energy output. Furthermore, the extraction of natural gas itself contributes significantly to greenhouse gas emissions, particularly methane, which traps 28 times more heat than carbon dioxide.
The environmental implications are especially relevant in states like Virginia and Louisiana, where natural gas constitutes a significant portion of electricity generation. For instance, Meta is planning the construction of a massive data center in Louisiana designed to consume two gigawatts (GW) of power—enough to fuel approximately 1.6 million U.S. homes. To meet this demand, Meta is collaborating with Entergy, the region’s largest utility, to build three new natural gas plants totaling 2.3 GW in capacity. This development not only threatens the state’s goal of carbon neutrality by 2050 but also highlights a deeper systemic issue regarding reliance on fossil fuels.
In July 2025, Meta amplified its plans for the Louisiana data center, expanding power requirements from two to five gigawatts over an area comparable to Manhattan. While there has been some resistance to Entergy’s plans—from nonprofits filing lawsuits over concerns regarding environmental impacts—the local judiciary has dismissed these challenges, citing insufficient grounds for demanding transparency.
Similar trends are emerging in Southeastern states like Virginia, Georgia, and South Carolina, where utilities are projected to add more than 20 GW of power from natural gas, with data centers driving between 65% to 85% of this load growth. Moreover, Entergy has plans for a new 754 megawatt natural gas power plant in Mississippi, marking its first such endeavor in half a century, aimed primarily at serving two data centers for Amazon.
Concerns about overbuilding are voiced by organizations like the Institute for Energy Economics and Financial Analysis (IEEFA), which warns that Big Tech’s high energy growth projections could be inflated. The implications are dire; if natural gas remains the default energy option for data centers, the U.S. could see an increase of 278 million metric tons in carbon emissions annually by 2035—equivalent to the yearly emissions of Florida.
In light of these challenges, many tech companies are reevaluating their carbon neutrality commitments. Rather than addressing the immediate issues at hand, companies like Meta and Google have pledged to double global nuclear capacity by 2050. However, such timelines are too protracted to address current energy needs, leaving many “clean” energy promises perceived as vague or insufficient. In Louisiana, although Meta has expressed intentions to finance 1.5 GW of renewable power, there has been no follow-through in terms of project announcements or state permissions.
As skepticism rises, Senator Sheldon Whitehouse has pressed Meta’s CEO for clarity on the projected greenhouse gas emissions from its data center, emphasizing the importance of aligning such plans with stated carbon neutrality goals. However, by August 2025, there had been no acknowledgment from Meta regarding these requests.
Interestingly, while corporate climate pledges surged during 2022 and 2023—when half of the world’s largest companies committed to net-zero emissions—only about 4% are expected to meet these targets. Additionally, Microsoft, Google, and Amazon have all publicly stated that AI’s growth is complicating their efforts to fulfill sustainability objectives.
For instance, in 2024, Google’s parent company dropped its carbon neutrality claim, citing challenges posed by the increased intensity of AI compute requirements. This surge has led to a nearly 50% increase in emissions over the past five years, a stark contrast to Google’s prior emissions reductions.
Despite the challenges at hand, there is a glimmer of hope. A Duke University study suggests that data centers could mitigate their environmental impact by implementing “load flexibility,” which allows them to reduce energy usage during peak grid demand, thereby integrating into the existing grid without necessitating new power plants. Recently, Google announced such agreements in states like Indiana and Michigan, hoping to navigate these challenges more effectively.
With Big Tech holding a staggering value of $17.9 trillion, these companies possess the resources and influence necessary to invest in clean energy solutions. In a time when climate protections face potential rollbacks, corporate responsibility is more crucial than ever. Sustainable energy alternatives exist and can provide firm electricity sources when combined with adequate storage, allowing for innovative solutions to the challenges faced by AI data centers.
An example of a successful renewable project is the collaboration between the Emirati renewable energy company Masdar and Emirates Water and Electricity Company. They have initiated a 5.2 GW solar power project equipped with battery storage to deliver consistent 1 GW power around the clock, costing $6 billion and expected to be operational by 2027. This undertaking could potentially offer a more cost-effective solution compared to current natural gas power plants, showcasing the feasibility of renewable alternatives in meeting the surge in energy demands.
Ultimately, it is essential for lawmakers to demand transparency from Big Tech companies regarding their energy sourcing and environmental impacts. The burden of climate costs associated with data centers falls on the general public, making it critical for these firms to disclose annual carbon emissions and commit to meaningful environmental standards. If the information they provide does not meet sustainability benchmarks, it is imperative that representatives advocate for measures that prioritize public interests over corporate gains.