AI’s New Power Deal: Data Centers Built on Fossil Fuel Megaprojects
The environmental impact of AI data center energy is emerging as vast fossil-fuel power projects dedicated to server farms reshape how technology companies secure electricity, drive emissions, and consume water, revealing a hidden cost of AI infrastructure that standard discussions of innovation and productivity mostly ignore. This is not a story about clever algorithms; it is a story about gigawatts of gas-fired generation built for machines, not homes. In Reeves County, a rural area with about 4,000 households, Chevron is developing enough gas-fired electricity to power more than 2 million homes, yet that energy will instead feed Microsoft’s data centers under Project Kilby. Meanwhile, Amazon is constructing the GW Ranch natural-gas plant beside a large data center in nearby Pecos County. These projects expose a simple truth: AI power consumption is no longer an abstract metric—it is reshaping the physical and ecological landscape.
Project Kilby: AI Hyperscalers Turn Big Oil into a Power Provider
Chevron’s Project Kilby marks a turning point where Big Oil recasts itself as a core supplier of AI data center energy rather than a climate liability. The 20-year deal with Microsoft covers 2.67 gigawatts of natural gas-fired power, with future room for solar and batteries, all built on Chevron’s land and gas resources in West Texas. In effect, this is a bespoke fossil-fuel grid for one hyperscaler, designed to sit behind the meter and avoid stressing public infrastructure while Texas policy demands that new data centers "bring your own power" so consumer bills are shielded. Chevron now openly pitches this as “a platform for growth” for more massive AI hyperscaler deals in gas-rich regions stretching from West Texas to the Rockies and Midwest. When the grid “catches up,” Kilby can connect, but there is no urgency—because the priority is AI energy security, not decarbonization.
GW Ranch: A Single AI Facility with Coal-Scale Emissions
If Project Kilby hints at AI’s climate cost, Amazon’s GW Ranch natural-gas plant makes it hard to ignore. The facility is planned with 35 natural-gas turbines and a combined capacity of about 7.65 gigawatts, much of it expected to run an adjacent AI-focused data center rather than homes or businesses on the wider grid. According to Cleanview, the project holds a Texas permit allowing up to 33 million tons of carbon dioxide emissions—higher than the limit for the largest coal-fired power plant in the United States. It may not reach that ceiling, but the allowance shows how extreme data center emissions could be for a single site. Amazon frames this expansion as compatible with its net-zero pledge, yet acknowledges “the world looks different now,” even as its emissions have risen with data center and AI growth. The contradiction is stark: climate pledges on one side, permitted coal-scale pollution for AI workloads on the other.
Beyond Carbon: Water Use and the Overlooked AI Environmental Burden
The AI environmental impact is not just about smokestacks and CO2—it is also about water in places that can least spare it. Data centers use substantial amounts of water for cooling, and Amazon’s operations around the world withdrew approximately 2.5 billion gallons of water in 2025. That is a startling figure when viewed alongside gas-fired plants in arid West Texas, where new power and server complexes are rising in sparsely populated counties. These withdrawals and thermal discharges rarely feature in upbeat narratives about AI innovation, yet they shape local ecosystems and compete with agriculture and communities in dry regions. Even Chevron admits it is "cognizant of the negative environmental implications of the AI boom and the surge of new gas-fired power plants," but awareness is not the same as restraint. The pursuit of low-latency AI workloads is now driving water-intensive industrial footprints in landscapes already under stress.
Unconventional Alliances, Uncomfortable Trade-offs
The most troubling part of this trend is how energy security for AI workloads is rewriting corporate alliances while sidestepping public debate. Tech giants that once prized renewable branding are turning to Big Oil precisely because natural gas plants can scale "quickly and reliably" for AI and cloud demand. Chevron, previously peripheral to AI deals, now calls itself an AI leader among oil majors, stitching together land, turbines, gas, and long-term contracts for hyperscalers. Amazon, for its part, is willing to back a dedicated gas facility with a permit that dwarfs coal-plant emission limits to keep its AI infrastructure supplied. These unconventional partnerships show that the industry’s real priority is uninterrupted compute, even if that means locking in decades of fossil generation. The conclusion is blunt: unless regulation and public pressure catch up, the future of AI will be built on privately negotiated power islands—high-emission, water-hungry, and largely invisible in the glossy story AI tells about itself.
