Soaring Energy Demands of Artificial Intelligence Challenge Climate Goals
San Francisco, Thursday, 22 January 2026.
While AI offers tools to reduce emissions, data centers now consume vast resources, with a single query using energy equivalent to streaming 16 seconds of HD video.
The Aggregate Impact of Micro-Consumption
While the energy cost of a single interaction may seem negligible, the cumulative effect of widespread adoption creates a massive environmental footprint. Ethan Mollick, a professor at the Wharton School of Business, notes that a typical prompt on GPT-4o consumes approximately 0.34 watt-hours of electricity [1]. When scaled across billions of daily interactions—ChatGPT alone processes 2.5 billion prompts every day [2]—the aggregate demand becomes staggering. This surge is reflected in national energy data: U.S. data centers consumed 176 terawatt-hours (TWh) of electricity in 2023, rising to 183 TWh in 2024 [3]. This represents a year-over-year increase of 3.977 percent, a trajectory that analysts project could see data centers devouring up to 12% of total U.S. electricity consumption by 2028 [1].
The Resource Reality Check
Beyond electricity, the physical infrastructure supporting artificial intelligence requires immense volumes of water for cooling, often in regions already facing scarcity. Research indicates that processing just 16 queries on a model like GPT-3 consumes a volume of water equivalent to a standard 500-milliliter bottle [2]. At the industrial level, hyperscale data centers can consume up to 19 million liters of freshwater daily, a figure comparable to the municipal usage of a town with a population between 10,000 and 50,000 [4]. This intensity is particularly concerning given that 40% of existing and proposed U.S. data centers are located in areas grappling with freshwater shortages [4].
A Divergence in Governance
As the environmental toll becomes clearer, regulatory approaches between major economies are diverging sharply. The European Union has moved toward transparency, requiring data center operators to report standardized Key Performance Indicators (KPIs) regarding energy, water, and heat reuse to a public database [1]. In contrast, the United States lacks a mandatory national database for these metrics [1]. Recent political moves have further widened this gap; a recent executive order from the Trump administration aims to override state-level regulations on data center construction, a move that has sparked local opposition in states like Arizona and Virginia [4]. While the EPA under the current administration has criticized certain permitting regulations as “red tape,” communities are increasingly organizing to block new developments due to resource concerns [4] [5].
The Efficiency Paradox
Despite these heavy costs, AI remains a potent ally in the fight against climate change, offering efficiencies that can offset emissions in other heavy industries. In the aviation sector, Alaska Airlines reported saving over 4.5 million liters of fuel in 2023 by using AI to analyze weather patterns and airspace restrictions [3]. Similarly, in agriculture—which accounts for nearly 70% of global freshwater usage—AI is driving significant conservation. The climate tech startup Kilimo has deployed AI-powered irrigation platforms in Chile’s Biobío region, allowing farms to reduce water use by up to 30% while generating additional profit through the sale of water credits [3].