The AI era gives fracking an encore, a surprising twist to an industry that was blamed, even in its early 2010s boom, by climate activists for the poisoning of water tables, the man-made earthquakes and for fossil fuels’ stubborn persistence.
AI firms are building massive data centres near major gas-production areas, often generating power directly from fossil fuels. This trend has been overshadowed in recent headlines on the intersection of AI, healthcare and climate change. But it could reshape the communities that host the facilities and raise difficult questions.
Let’s take the latest example. The Wall Street Journal reported this week about AI coding assistant Poolside’s plans to build a data center on 500 acres of land in West Texas, approximately 300 miles west from Dallas. This is an area roughly two-thirds of the size of Central Park. The facility will generate power by using natural gas from the Permian basin, the nation’s largest oil and gas field. Hydraulic fracturing, which is common, is the only way to go.
Horizon, the project, will produce 2 gigawatts in computing power. This is equivalent to the Hoover Dam’s entire electric power, except that instead of harnessing Colorado River, it burns fracked gas. Poolside is building the facility in conjunction with CoreWeave. This cloud computing company rents out Nvidia AI chip access and has access to over 40,000 of them. The Journal describes it as an “energy Wild West”which is apt. Poolside is not the only one. Nearly all major AI players have similar strategies. OpenAI CEO Sam Altman visited his company’s flagship Stargate Data Center in Abilene Texas, around 200 miles away from the Permian basin, and was honest, saying “We’re burning gasoline to run this data centre.”
According to the Associated Press, the complex requires approximately 900 megawatts across eight buildings, including a new gas fired power plant that uses turbines similar to the ones used to power warships. The companies claim that the plant is only used for backup power and the majority of electricity comes from the local grid. This grid, it should be noted, is a mixture of natural gas, wind and solar farms located in West Texas.
However, the people who live near these projects don’t seem to be comforted. Arlene Mendler, who lives across the street is not happy. She told the AP that she wished someone had asked for her opinion before bulldozers removed a large tract of mesquite to make way for what is being built atop.October 27-29, 2025
“It has completely changed the way we were living,” Mendler told the AP. She moved to the area in search of “peace, tranquility, and tranquility” 33 years ago. Now, construction is the background soundtrack, and bright lights have spoiled the nighttime views.
And then there’s the issue of water. Locals in West Texas, a region prone to drought, are worried about the impact of new data centers on water supply. Altman visited a city whose reservoirs were about half full, and whose residents were watering their lawns twice a week. Oracle claims that each of the eight buildings only needs 12,000 gallons of water per year, after a million-gallon initial fill for closed-loop systems. Shaolei Ren is a professor at the University of California Riverside who studies AI’s impact on the environment. She told AP this was misleading. These systems require a higher amount of electricity, which in turn means that the power plants that generate this electricity will use more water indirectly.
Meta has adopted a similar approach. The company plans to build in Richland Parish, Louisiana’s poorest region, a $10 billion datacenter the size of 1,700 soccer fields. It will need two gigawatts for computation. Entergy, a utility company, will spend $3.2 Billion to build three large natural gas power plants with 2.3 Gigawatts capacity to feed the facility using gas extracted from the nearby Haynesville Shale. Residents of Louisiana, like those in Abilene are not thrilled to be surrounded by bulldozers round the clock.
Meta is also building in Texas but elsewhere in the state. This week, the company announced the construction of a ahref=””https://www.reuters.com/business/meta-commits-15-billion-ai-data-center-texas-2025-10-15/” “rel=””noreferrer noopener nofollow””target=””_blank” “>$1.5 Billion data center near the New Mexico-Texas border. One gigawatt is expected to be online by 2028. Meta says that the facility in El Paso will be powered by 100% clean and renewable electricity, despite the fact that it is not near the Permian basin. Meta gets a point.)
Elon Musk’s xAI has fracking links. Their Memphis facility has been the subject of considerable controversy this year. Memphis Light, Gas and Water, which sells electricity to xAI, but will eventually own substations xAI builds, purchases natural gas at the spot market and pipes the gas to Memphis through two companies: Texas Gas Transmission Corp. and Trunkline Gas Company.
Texas Gas Transmission, a bidirectional gas pipeline, transports natural gas from Gulf Coast areas and several major hydraulically fracturing shale deposits through Arkansas, Mississippi Kentucky and Tennessee. Trunkline Gas Company is the other Memphis supplier that carries natural gases from fracked sources.
Ask AI companies why they are taking this path. They’ll tell that it’s not only about electricity, but also about beating China.
Chris Lehane made this argument last week. Lehane, an experienced political operative, who joined OpenAI in 2024 as vice president of global Affairs, laid out his case during a TechCrunch onstage interview.
Lehane stated, “We believe in the not too distant future, at least for the U.S. and in fact around the globe, we will need to generate in the vicinity of a gigawatt per week.” He cited China’s massive energy buildout, which included 450 gigawatts of power and 33 nuclear plants built in the past year. Lehane referred back to geopolitics when TechCrunch asked him about Stargate’s choice to build in economically challenged regions like Abilene or Lordstown, Ohio where more gas-powered plant are planned. “If we [as a country] get this right, we have the opportunity to reindustrialize countries, bring back manufacturing, and also transition our energies so that we can do the modernization we need to take place.” The Trump administration certainly is on board. The July 2025 Executive Order accelerates the construction of gas-powered AI data centres by streamlining environmental permits and offering financial incentives. Federal lands are also opened for projects that use natural gas, nuclear, or coal power. Renewables, however, are not included in this order.
Most AI users are unaware of the carbon footprint associated with their new toys and tools. They’re more concerned with the capabilities of OpenAI’s hyperrealistic Sora 2 – a product that generates videos that require exponentially more power than a simple bot – than where the electricity is coming from.
These companies are counting on it. They’ve positioned the natural gas as a pragmatic, inevitable solution to AI’s exploding energy demands. The speed and scale of the fossil fuel buildout deserve more attention than they are getting.
This will not be pretty if this is a ‘bubble. The AI sector is a circular firing squad with many dependencies. OpenAI needs Microsoft, Nvidia, Broadcom, Oracle and data center operators. All of them are buying and selling each other, creating a self-reinforcing cycle. The Financial Times reported this week that if the foundation fails, there will be a lot more expensive infrastructure, both digital and gas-burning, left behind. The Financial Times wrote that OpenAI’s ability to meet its obligations alone is “increasingly of concern for the wider economic”.
A key question that has been largely ignored in the conversation is if all this new capacity is actually necessary. According to a Duke University study, utilities use only 53% or their capacity on average throughout the year. MIT Technology Review published a report earlier this yearthat showed there was significant room for new demand to be met without building new power plants.
According to Duke researchers, if data centers cut their electricity consumption in half for a few hours at peak demand times each year, utilities could accommodate an additional 76 gigawatts. This would be enough to absorb the 65 gigawatts that data centers are expected to require by 2029.
This kind of flexibility will allow companies to launch AI-data centers faster. It could also give utilities more time to develop cleaner alternative technologies, allowing them to avoid the rush to build new natural gas infrastructure.
However, this would mean losing out to an autocratic government, according to Lehane and others in the industry. Instead, the natural-gas building spree is likely to saddle the regions with more fossil fuel plants and leave residents with high electricity bills to pay for today’s investments.
Meta has, for example, guaranteed that it will cover Entergy’s costs for the Louisiana generation for a period of 15 years. Poolside’s lease agreement with CoreWeave is for 15 years. What happens to customers after these contracts expire remains a question.
The situation may change in the future. Private money is being invested in small modular reactors and on solar installations, with the hope that these cleaner energy options will become more central sources of energy for these data centres. Fusion startups such as Helion and Commonwealth Fusion Systems raised similar funding from AI pioneers, including Nvidia. This optimism is not confined to the private investment world. The excitement has spilled over into public markets, where several “non-revenue-generating” energy companies that have managed to go public have truly anticipatory market caps, based on the expectation that they will one day fuel these data centers.
In between — which could take decades — the biggest concern is that those who will be left with the bill, both financially and environmentally speaking, did not ask for this.
