The Hidden Environmental Cost Behind America’s Expanding AI Infrastructure

A new scientific assessment examines electricity consumption and carbon emissions across hundreds of hyperscale data centres in the United States.

The rapid expansion of artificial intelligence infrastructure is creating new challenges for electricity systems, according to research published in the journal Nature Sustainability on November 16, 2026. The study examines the electricity consumption and associated carbon emissions of hyperscale data centres operating across the United States.

Researchers compiled facility-level information covering 403 hyperscale data centres between May 2024 and April 2025. Their assessment provides a more detailed picture of the energy requirements associated with some of the largest computing facilities supporting cloud services and artificial intelligence.

Under the study’s central scenario, these facilities consumed approximately 82 terawatt-hours of electricity during the assessment period. This represented roughly 1.8 per cent of total US electricity consumption.

The researchers also estimated that the associated carbon dioxide emissions reached approximately 25.7 million metric tonnes under their central attribution method.

The findings offer an important perspective on the infrastructure required to operate modern digital services. Artificial intelligence models, cloud computing platforms and large-scale data processing systems depend on facilities containing thousands of servers, specialised networking equipment and extensive cooling infrastructure.

Unlike traditional office buildings, hyperscale data centres can require substantial amounts of electricity continuously. Computing equipment produces heat while operating, making cooling systems an important part of their overall energy requirements.

As demand for AI services grows, technology companies are investing in larger facilities and expanding existing computing capacity. These developments have raised questions about where the additional electricity will come from and how the expansion will affect national and regional power networks.

The Nature study estimates that approximately 54 per cent of the electricity generation attributed to the examined data centres came from fossil-fuel sources. This finding demonstrates why the environmental impact of digital infrastructure cannot be assessed simply by looking at the efficiency of individual servers.

Electricity generation remains a major factor in determining the carbon footprint of a computing facility. A data centre connected to a region with a relatively carbon-intensive electricity supply can have a very different environmental impact from an otherwise comparable facility operating in an area with more renewable generation.

The research also offers a framework for examining individual facilities and the electricity systems connected to them. Instead of relying exclusively on broad national estimates, the authors assembled information about hundreds of facilities and linked their electricity requirements to regional power generation.

There are, however, important limitations. The estimates rely on facility-load scenarios and an attribution method that connects electricity consumption with the generation mix within relevant balancing authorities. They should therefore not be interpreted as precise measurements of every individual data centre’s directly contracted electricity supply.

The findings are particularly relevant as governments and electricity providers attempt to plan for future demand. New data centres can require additional transmission capacity, local power infrastructure and long-term electricity agreements.

This creates potential challenges for communities where major facilities are being developed. Questions about electricity costs, water requirements, land use and the distribution of infrastructure benefits are increasingly becoming part of discussions about large technology projects.

Technology companies have several options for reducing their environmental impact. These include improving server efficiency, using advanced cooling systems, locating facilities near cleaner electricity sources and investing in additional renewable energy generation.

Battery storage and improved grid management could also help balance electricity demand as computing facilities expand. However, the effectiveness of these approaches depends on how they are implemented and whether clean electricity is genuinely available when facilities need it.

Artificial intelligence is expected to play an important role in scientific research, manufacturing, healthcare and business operations. Its infrastructure requirements therefore have implications extending well beyond the technology industry.

The new study highlights the importance of including energy and emissions considerations in future AI development. Digital innovation may create significant economic and scientific opportunities, but its physical infrastructure must also be planned responsibly.

As data centre construction accelerates, transparent energy reporting and more detailed environmental assessments will become increasingly important for policymakers, investors and local communities.

The challenge will be to support the expansion of computing infrastructure while ensuring that the electricity needed to operate it increasingly comes from reliable, lower-carbon sources.

Exit mobile version