A single ChatGPT query consumes 10 times more electricity than a Google search. Training one large language model burns through as much energy as 130 American homes use in a year. These numbers, documented by Stanford researchers and the International Energy Agency, explain why Bernie Sanders and Alexandria Ocasio-Cortez just introduced legislation to freeze construction of new AI datacenters — facilities that The Guardian US reports are proliferating across the country without federal oversight.
The proposed moratorium represents the first congressional attempt to confront a stark reality: artificial intelligence infrastructure is devouring America's energy supply faster than regulators can respond. Microsoft's electricity consumption has doubled since 2020. Google's carbon emissions have surged 48% in five years, despite corporate pledges to reach net zero. Amazon Web Services now operates datacenters that each require the electrical capacity of a small city.
"We're watching tech companies build the digital equivalent of coal plants while claiming to save the world," Sanders said in announcing the bill Wednesday morning. The Vermont senator's assessment reflects a documented pattern: technology corporations are reviving dormant fossil fuel plants, securing exclusive access to nuclear reactors, and in some cases building their own power generation facilities — all to feed datacenters that didn't exist five years ago.
The energy crisis these facilities create isn't abstract. In Ireland, datacenters now consume 21% of the nation's electricity, forcing the government to impose connection moratoriums in Dublin. In Northern Virginia — home to 70% of global internet traffic — Dominion Energy has warned that datacenter growth could trigger rolling blackouts by 2028. Singapore banned new datacenter construction entirely in 2019, only partially lifting restrictions this year after companies agreed to stringent efficiency standards.
What makes the Sanders-AOC proposal particularly significant is its framing: not as anti-technology legislation, but as a pause to create what the lawmakers call "strong federal safeguards for AI." The bill would establish a temporary ban on new datacenter construction while Congress develops comprehensive regulations addressing energy consumption, water usage (datacenters require millions of gallons daily for cooling), and community impact standards.
Tech industry representatives have already mobilized against the proposal. The Information Technology Industry Council, whose members include Apple, Amazon, and Meta, released a statement warning that "artificial restrictions on infrastructure development will cede American leadership in AI to China." This argument — that any regulation equals capitulation to geopolitical rivals — has become the industry's primary defense against oversight attempts.
But the geopolitical framing obscures a more immediate crisis. Communities from Ohio to Oregon are discovering that datacenter projects approved by local officials often come with hidden costs: strained electrical grids, depleted aquifers, and property tax agreements that shift infrastructure burdens to residential ratepayers. In Chandler, Arizona, datacenters consume 66% of the city's water supply in a region experiencing its worst drought amid irreversible climate shifts in 1,200 years.
The federal regulatory vacuum around AI infrastructure has created what energy policy experts call a "gold rush dynamic." Tech companies race to secure power purchase agreements and water rights before competitors, while utilities scramble to meet demand by any means available — including reactivating coal plants that were scheduled for decommissioning. The Crane Clean Energy Center in Pennsylvania, a nuclear plant that was set to close, will instead sell its entire output to Microsoft under a 20-year agreement.
Labor unions have emerged as unexpected allies in the push for datacenter regulation. The Communications Workers of America notes that modern datacenters employ fewer than 50 people while consuming resources equivalent to manufacturing facilities that once provided thousands of jobs. "These facilities extract community resources without providing community benefits," the union stated in supporting the Sanders-AOC bill.
The proposed moratorium would last 18 months, during which Congress would be required to develop standards addressing several key areas: mandatory renewable energy requirements for new facilities, water conservation standards in drought-affected regions, community benefit agreements ensuring local hiring and infrastructure investment, and transparency requirements for energy consumption reporting. The bill would also establish a federal AI Infrastructure Commission modeled on the Nuclear Regulatory Commission, with authority to approve or deny datacenter projects based on environmental and community impact.
Opposition from tech companies centers on claims that pausing construction would harm innovation. Yet the companies making these arguments have simultaneously argued in other contexts that AI development is moving too fast for safety. OpenAI CEO Sam Altman has called for AI regulation while his company's infrastructure demands help drive the very crisis Sanders and AOC seek to address.
The bill arrives as private equity firms pour billions into datacenter development, viewing the facilities as recession-proof investments. Blackstone alone has committed $100 billion to datacenter projects over the next five years. These financial dynamics create additional urgency: once built, datacenters become stranded assets that communities must continue powering regardless of environmental cost.
State-level attempts to regulate datacenter development have met fierce industry resistance. When Oregon proposed efficiency standards for new facilities, tech lobbyists threatened to relocate projects to states with fewer regulations. This dynamic — what economists call a "race to the bottom" — demonstrates why federal standards are necessary, advocates argue.
The international context provides models for what effective regulation might accomplish. The European Union's Energy Efficiency Directive requires datacenters to report energy consumption and implement waste heat recovery systems. Denmark mandates that new datacenters connect to district heating networks, turning waste heat into residential heating. These requirements haven't stopped datacenter construction; they've simply forced companies to build more efficient facilities.
Environmental justice communities have particularly felt the impact of unregulated datacenter growth. In Chicago's West Side, a proposed Amazon datacenter would sit adjacent to a predominantly Black neighborhood already burdened by industrial pollution. Residents learned about the project only after permits were approved — a pattern repeated in low-income communities across the country where datacenters cluster near existing electrical infrastructure.
The water consumption issue deserves particular scrutiny. A single datacenter can use 5 million gallons daily for cooling — water that increasingly scarce Western communities can't spare. Yet water usage reporting remains voluntary, and many facilities negotiate discounted rates that encourage consumption rather than conservation. The Sanders-AOC bill would mandate water usage disclosure and require new facilities to use air cooling or closed-loop systems in water-stressed regions.
Perhaps most significantly, the moratorium would force a conversation about AI's actual utility versus its resource consumption. While tech companies promote AI as essential to human progress, much of current datacenter capacity supports applications like targeted advertising, cryptocurrency mining, and large language models that help students cheat on homework. The question isn't whether these applications should exist, but whether they justify consuming resources that communities need for basic services.
The bill faces steep odds in a Congress where tech lobbying expenditures reached $70 million last year. But its introduction shifts the terms of debate. Instead of accepting unlimited datacenter growth as inevitable, lawmakers are finally asking whether the communities bearing the cost should have a say in the process. As federal climate funding continues flowing and climate change intensifies resource scarcity and energy demands, that question will only become more urgent.
