Major technology companies are constructing private natural gas power plants to supply electricity directly to AI data centers, bypassing public grids and building fossil fuel infrastructure that directly contradicts their public climate commitments, according to a Washington Post investigation.
The facilities, planned or under construction by Microsoft, Google, Amazon, and Meta, represent a fundamental shift in how Big Tech powers AI: rather than competing for limited grid capacity or waiting for renewable energy buildout, companies are securing dedicated fossil fuel generation to meet exploding AI energy demands on their own timelines.
The scale is staggering. AI model training and inference require exponentially more electricity than traditional computing, with a single large language model query consuming roughly ten times the energy of a standard Google search. As companies race to deploy AI across products and services, their energy projections have outpaced grid infrastructure and renewable capacity.
So they are building their own power plants—fueled by natural gas.
The contradiction is stark. Microsoft, Google, Amazon, and Meta have all made highly publicized commitments to carbon neutrality, net-zero emissions, or 100% renewable energy. Microsoft pledged to be carbon negative by 2030. Google committed to operating on carbon-free energy 24/7 by 2030. Amazon co-founded the Climate Pledge, targeting net-zero emissions by 2040.
Yet when AI's energy requirements exceeded what renewables could immediately supply, the companies chose fossil infrastructure over slowing AI deployment.
The plants are being built off-grid or with dedicated connections, sidestepping public utility oversight and community input. That design choice has sparked fierce opposition in several locations, particularly from environmental justice communities already burdened by industrial pollution.
In Ohio, a planned gas plant to power Meta data centers faces opposition from residents in nearby communities—disproportionately low-income and communities of color—who report existing air quality problems from industrial facilities. The new plant would add emissions in areas that already exceed federal air quality standards for particulate matter.
In Virginia, Amazon's plans for dedicated gas generation near Loudoun County data centers have drawn fire from local environmental groups highlighting that the facilities would lock in decades of fossil fuel use in a state attempting to transition to clean energy.
Georgia and Texas are seeing similar patterns: data centers sited in communities with limited political power to resist, powered by gas plants that bypass the renewable energy mandates increasingly applied to public utilities.
The environmental justice dimension is not incidental. Off-grid fossil infrastructure disproportionately impacts communities that lack resources to fight back, perpetuating a pattern where marginalized populations bear the health and environmental costs of technologies consumed primarily by wealthier users.
In climate policy, as across environmental challenges, urgency must meet solutions—science demands action, but despair achieves nothing. The AI power plant trend demonstrates that corporate climate commitments remain subordinate to product timelines when the two conflict.
Tech companies defend the plants as transitional infrastructure, arguing that AI's benefits justify temporary reliance on gas while renewable capacity scales up. They note that natural gas emits less carbon than coal, and some facilities incorporate carbon capture technology—though capture systems have historically underperformed promises and do nothing to address local air pollution.
But temporary is doing significant work in that framing. Power plants are 30-to-50-year infrastructure investments. Gas plants built today will operate into the 2050s and 2060s—well past the deadlines these same companies set for their net-zero commitments. The plants are not bridging to a renewable future; they are extending the fossil fuel present.
The regulatory gap is glaring. Public utilities face renewable portfolio standards, emissions caps, and community review processes. Private off-grid generation largely escapes those requirements, creating a two-tier system where publicly consumed electricity is regulated for climate impact while private corporate power is not.
Legislators in several states are now considering bills to close that loophole, subjecting off-grid industrial generation to the same climate standards as grid power. But the plants already under construction may be grandfathered in, locking in emissions for decades.
The AI power plant story reveals an uncomfortable truth: when frontier technology meets climate policy, technology timelines are winning. The companies building AI infrastructure are the same companies that positioned themselves as climate leaders, yet they are now constructing fossil fuel plants because renewables could not scale fast enough to match AI's energy appetite.
That choice has consequences that extend far beyond corporate carbon accounting. It entrenches environmental inequality, bypasses democratic input, and creates fossil infrastructure that will outlive the climate commitments it contradicts.
