As AI advances, a new study warns that its energy consumption could soon surpass even Bitcoin mining, raising urgent sustainability concerns.

According to research by Alex de Vries-Gao, a PhD candidate at Vrije Universiteit Amsterdam’s Institute for Environmental Studies, AI may account for nearly half of global data centre electricity usage by the end of 2025 – up from an estimated 20% today.

De Vries-Gao’s analysis highlights a critical challenge: major tech firms rarely disclose AI’s exact energy demands, making accurate assessments difficult.

Drawing parallels to cryptocurrency mining, he warns that the tech industry’s bigger is better” mindset is driving unsustainable energy use, with his projections suggesting that AI hardware could soon consume as much electricity as entire nations, potentially rivalling the UK’s total power demand by 2025.

The surge in AI adoption has triggered a wave of new data centre construction, particularly in the U.S., where electricity demand is expected to rise 25% by 2030 due to AI, traditional data centres, and Bitcoin mining.

However, the environmental impact varies widely. For example, an AI query processed in West Virginia – where coal dominates the grid – generates nearly double the emissions of one in California, which relies more on renewables.

While companies like Google and Microsoft report overall emissions, few break down AI’s specific contribution. De Vries-Gao’s triangulation method – using chip production data, corporate reports, and analyst estimates – reveals AI’s staggering energy footprint.

Despite efficiency improvements, experts fear a Jevons paradox effect: as AI grows more powerful, its energy use may skyrocket regardless.

The least of our problems?

As artificial intelligence advances at breakneck speed, its voracious energy demands threaten to overshadow even Bitcoin mining’s notorious consumption. While data centres now account for over 1.1% of global electricity use, the AI revolution is pushing this figure higher – with potentially dire consequences for climate goals and power grids worldwide.

The numbers paint a worrying picture. Research suggests AI could consume half of all data centre power by 2025, up from 20% today.

At this rate, the electricity needed to run AI hardware may soon match the entire consumption of countries like the United Kingdom.

The environmental cost is already visible: Google’s emissions have surged 48% in five years, with Microsoft similarly struggling to meet sustainability targets as AI workloads multiply.

Yet there are glimmers of progress. The carbon intensity of data centre operations is falling as tech giants shift toward renewables—hyperscalers now source 91% of their power from clean energy.

Efficiency gains are real too, with average power usage effectiveness scores improving from 1.44 in 2019 to 1.38 today. But these advances risk being overwhelmed by AI’s exponential growth.

The International Energy Agency predicts global data centre electricity demand could double by 2026, reaching levels comparable to Japan’s annual consumption.

The challenges multiply when considering AI’s broader footprint. Soaring demand for specialised hardware drives mining of rare minerals like lithium and cobalt.

Recommended reading

• Experts Call for Urgent Action to Help UK Meet Net Zero Targets
• Are High Costs Impeding the Public From Adopting Low Carbon Tech?
• How Many Scots Support Clean Energy Infrastructure Being Built?

Water use for cooling has jumped nearly 10% as server densities increase. And the geographical lottery of energy sources means the same AI query can generate twice the emissions in coal-dependent regions versus renewable-powered ones.

Industry leaders face tough choices. While AI optimists like Bill Gates argue the technology will ultimately help solve climate challenges, current trends suggest a painful interim period.

The IMF estimates AI’s cumulative emissions through 2030 could reach 1.7 gigatons – equivalent to Italy’s five-year output – even as it boosts global GDP. With 40% of AI data centres projected to hit power capacity limits by 2027, the infrastructure strain is becoming impossible to ignore.

What emerges is a paradox of progress: the same innovation promising to revolutionise energy efficiency is itself becoming a major emissions source – it’s a problem that has to be reconciled.

Don’t Miss Scotland’s Biggest Tech Event!

Join us at DIGIT Expo West on 5th June at the SEC Glasgow. Get leading industry insights across AI, Cyber Security, and Data, and grow your network at Scotland’s largest gathering of tech leaders – with 1500+ attendees, 50+ speakers, and 50+ exhibitors.

Register your FREE place now at www.digitexpowest.com