U.S. data centres’ soaring energy use driven by AI sparks innovation and environmental challenges

America’s data centres consumed a record 176 TWh of electricity in 2023, with demand set to rise sharply amidst efforts from tech giants to source cleaner energy and develop innovative storage solutions, highlighting the complex balance between technological progress and sustainability.

America’s data centers consumed an estimated 176 terawatt-hours (TWh) of electricity in 2023, accounting for around 4.4% of the nation’s total power use. This significant figure has more than doubled from roughly 70 TWh in 2014, illustrating a rapid rise driven largely by surging demand for artificial intelligence (AI) and increased domestic technology manufacturing. Government projections estimate that by 2028, data centers could consume between 325 TWh and 580 TWh, potentially reaching 12% of overall U.S. electricity consumption if current trends continue.

This surge presents a formidable challenge for large technology companies, commonly called hyperscalers, such as Google and Microsoft, which have committed to slashing their carbon footprints even as their computing demands balloon. The reliance on consistent, high-capacity power is prompting utilities in states like Virginia and Nevada to expand gas-fired power plants and prolong the operation of aging coal facilities, a move that threatens to increase both carbon emissions and electricity costs for consumers. Additionally, data centers frequently depend on diesel generators for backup power, further complicating efforts to reduce environmental impacts.

In response, major tech firms are exploring innovative solutions and partnerships aimed at delivering clean, reliable energy for their data centres. Google, for instance, has invested in next-generation geothermal projects with startups like Fervo Energy, which is operating enhanced geothermal plants in Nevada and Utah. The company is also engaged in developing small modular nuclear reactors through agreements with firms like Kairos Power. Meanwhile, Microsoft is pursuing a $1.6 billion initiative to restart the Three Mile Island Unit 1 nuclear reactor in Pennsylvania, backed by long-term power purchase agreements with Constellation Energy. Within its facilities, Microsoft is advancing alternatives to diesel backup generators, experimenting with hydrogen fuel cells, advanced batteries, and making server hardware more energy-efficient to reduce overall power requirements.

Beyond securing clean energy sources, both Google and Microsoft are working with utility providers on demand-response agreements that allow data centers to reduce machine-learning workloads during peak grid stress or severe weather events. These efforts aim to provide greater grid flexibility and mitigate utility costs. Google’s partnerships with Omaha Public Power District and the Tennessee Valley Authority exemplify this strategy, enabling data centers to better align electricity consumption with grid conditions.

Looking forward, both companies see promise in long-duration energy storage technologies. Current lithium-ion batteries typically provide only a few hours of backup capacity, but emerging solutions now target medium-term storage spanning eight to 24 hours. Commercialising and scaling these technologies would enable data centers to store surplus wind and solar power for use during extended periods without generation, thus further reducing reliance on fossil fuels. This approach aligns with broader industry forecasts anticipating that by 2030, natural gas will remain the largest single source of data center power, supplying over 40%, with renewables like wind and solar providing close to a quarter, nuclear energy around 20%, and coal about 15%. The growing role of nuclear power, bolstered by initiatives to reinvigorate retired plants and develop advanced reactors, could significantly shift this energy mix in the coming decade.

In light of AI’s explosive growth, which industry analysts attribute as the primary driver behind the soaring demand, US data centers face mounting pressure to balance their operational needs with sustainability targets. Some forecasts suggest by 2030, AI-related workloads could account for up to 70% of this increased electricity consumption, with generative AI alone contributing around 40%. The rapid expansion of data-heavy cloud services underscores the urgency for innovation in energy sourcing and efficiency to manage environmental impacts effectively.

As hyperscalers push to align their data center operations with ambitious climate goals, their collaborations with clean energy providers, nuclear ventures, and energy storage innovators represent vital steps. However, the sector’s trajectory also underscores the inherent tension between technological advancement and sustainability, highlighting the need for coordinated efforts among industry players, utilities, regulators, and policymakers to foster a reliable and low-carbon digital infrastructure for the future.

📌 Reference Map:

• ^[1] Canary Media – Paragraph 1, Paragraph 2, Paragraph 3, Paragraph 4, Paragraph 5, Paragraph 6, Paragraph 7, Paragraph 8
• ^[2] U.S. Department of Energy – Paragraph 1
• ^[3] Pew Research – Paragraph 7
• ^[4] AP News – Paragraph 4
• ^[5] U.S. Department of Energy Fact Sheet – Paragraph 1
• ^[6] Socomec – Paragraph 1
• ^[7] Accenture – Paragraph 7

Source: Fuse Wire Services