US Data Centers Consume More Electricity Than Mid-Sized Countries, with Fossil Fuels Still Dominant

Study reveals US hyperscale data centers consume more electricity than mid-sized countries, with fossil fuels still dominant.
A new paper in *Nature Sustainability* finds that US hyperscale data centers collectively consume more electricity than many mid-sized countries, with fossil fuels remaining their largest power source—challenging the "clean and efficient" image of data centers. The article analyzes why fossil fuels dominate: renewable intermittency can't meet 24/7 stable demand, the grid remains fossil-heavy, and local sourcing is more economical. It also notes that despite continuous PUE optimization, surging server numbers offset efficiency gains, with total energy use and emissions still rising fast. The study calls on regulators to reassess environmental and disclosure requirements and on tech companies to make fundamental changes in energy procurement, site selection, and computing architecture, while exploring nuclear and long-duration storage solutions.
Data Centers Become Energy Giants
A newly published paper in the journal Nature Sustainability reveals a shocking fact: the total electricity consumption of US hyperscale data centers has surpassed that of many mid-sized countries, and fossil fuels remain the primary power source for these facilities. This finding challenges the conventional impression of data centers as "clean and efficient" and prompts a reexamination of the tech industry's carbon footprint.

The study focuses on hyperscale facilities—large data centers operated by tech giants that house hundreds of thousands of servers. These facilities are often regarded as models of efficiency, but the research shows that their overall energy consumption has reached national scales, and the resulting greenhouse gas emissions cannot be ignored.
Hyperscale data centers typically refer to facilities operated by a single organization, with server counts in the hundreds of thousands and floor areas exceeding tens of thousands of square meters. The exact threshold varies slightly across different statistical bodies, but the core characteristic is the operational model difference brought about by economies of scale. Such facilities are mostly built by tech giants like Amazon, Microsoft, Google, and Meta, employing highly automated operations and customized hardware, and are therefore often seen as energy efficiency benchmarks. However, precisely because of their enormous individual scale, their total energy consumption—even with high unit efficiency—can accumulate to levels comparable to a country's electricity usage. Understanding this concept helps grasp why researchers directly compare them with national-level energy and emissions data.
Why Fossil Fuels Remain Dominant
Although many tech companies have pledged to use renewable energy, the paper points out that fossil fuels are still the largest source of power for US hyperscale data centers. There are multiple reasons behind this: the intermittency of renewable energy makes it difficult to meet the 24/7 stable power supply demands of data centers; fossil fuels still account for a large share of the existing grid structure; and some data centers are located in regions rich in fossil fuel resources, where drawing power locally is more economical.
The high energy consumption of data centers not only brings environmental pressure but also exacerbates the burden on local power grids. With the explosive growth in demand for cloud computing and artificial intelligence, the electricity demand of data centers is expected to continue rising. How to balance computing expansion with low-carbon goals has become a challenge the industry must face.
The intermittency of renewable energy refers to the fact that output from sources like solar and wind fluctuates with day-night cycles, weather, and seasons, and cannot provide stable output on demand like fossil fuel generators. Data centers have extremely high requirements for power supply reliability, typically needing availability of 99.999% or higher, as any outage could cause service interruption and data loss. Therefore, in the absence of large-scale long-duration energy storage or stable low-carbon baseload power sources, grid operators often still need to rely on natural gas, coal, and other fossil fuel generators as backup. This is also one of the deeper reasons why many tech companies, despite purchasing renewable energy certificates, still have an actual electricity mix dominated by fossil fuels.
Efficiency Gains Fail to Offset Scale Expansion
Over the past decade, data center energy efficiency has indeed improved significantly, with power usage effectiveness (PUE) continuously optimized. But the paper suggests that the benefits of efficiency gains are being offset by the surge in server numbers. The overall energy consumption of hyperscale data centers continues to rise rapidly, and relying solely on energy efficiency improvements is no longer sufficient to curb the growth of carbon emissions.
This conclusion poses a challenge to the tech industry's sustainability narrative. Companies need to go beyond energy efficiency optimization and make more fundamental changes in energy procurement, site selection, and computing architecture.
Power Usage Effectiveness (PUE) is the core metric for measuring data center energy efficiency, defined as the ratio of total data center power consumption to IT equipment power consumption. The ideal value is 1.0, meaning all electricity is used for computing equipment. The industry average PUE has dropped from around 2.0 in earlier years to the 1.1–1.5 range, with hyperscale facilities potentially reaching below 1.1. However, PUE only reflects the efficiency of auxiliary facilities (such as cooling and power distribution) and does not measure the relationship between computing output and energy consumption of the IT equipment itself. When server numbers multiply, total energy consumption will still rise even as PUE continues to optimize—this is precisely the mechanism by which efficiency gains are offset by scale expansion.
Implications for Industry and Policy
The study provides important reference for policymakers and industry participants. On one hand, regulators may need to reassess environmental assessment standards and energy disclosure requirements for data centers; on the other hand, tech companies should increase investment in clean energy and explore stable low-carbon power solutions such as nuclear energy and long-duration energy storage.
For readers concerned about AI and cloud computing development, this research reminds us that behind the relentless surge in computing power lies a very real energy bill. While pursuing technological breakthroughs, how to make data centers truly "green" will be one of the most challenging issues in the coming years.
Related articles

WorkBuddy for Absolute Beginners: A Hands-On AI Agent Tutorial
Based on a Bilibili tutorial series, this guide explains WorkBuddy AI agent's core positioning, four-layer architecture, and real-world performance, comparing its execution capabilities with chat tools like Doubao, and analyzing its three advantages: zero configuration, free quota, and ecosystem integration—ideal for beginners exploring AI agent automation for office work.

GLM 5.5 Arrives: Has Claude Cracked the 167-Year-Old Riemann Hypothesis?
Claude and a mathematician claim to have proved the 167-year-old Riemann Hypothesis, Zhipu's GLM 5.5 is imminent, Google's Loop agent surfaces, and Thinking Machines open-sources Inkling 2 Mini.

AI Coding Made Building Cheap, but Not Maintenance
AI coding agents make internal tools cheap to build but not to maintain. This article examines the hidden liabilities of tool sprawl and proposes three governance principles: inventory, ownership, and reuse first.