AI-driven data centres to double electricity use by 2030, says IEA

Artificial intelligence surges energy demand as tech giants turn to nuclear solutions.

AI-driven data centres to double electricity use by 2030, the IEA warns, highlighting rising energy challenges amid the global push for carbon neutrality. Photo by Ronny Hartmann/AFP

By Anna Fadiah and Hayu Andini

AI-driven data centres to double electricity use by 2030, according to a groundbreaking report by the International Energy Agency (IEA) released on Thursday. As artificial intelligence applications rapidly expand, so too does the strain on global energy supplies—a trend the IEA warns could significantly challenge the world’s climate goals and energy security.

While AI has the potential to make energy systems smarter and more efficient, its demand for computing power, particularly through generative AI models, is already triggering a sharp rise in electricity consumption. According to the IEA, data centres worldwide consumed around 460 terawatt hours (TWh) of electricity in 2024, equivalent to nearly 1.5 percent of global demand. That figure is now projected to soar to more than 1,000 TWh by 2030, should current trends continue unchecked.

This startling growth raises urgent questions about how to balance innovation with environmental responsibility. And it’s no longer a theoretical debate—it’s already reshaping how the world’s largest tech companies power their operations.

The AI boom and its energy cost

Artificial intelligence systems—especially generative AI models like ChatGPT and Google Gemini—depend heavily on massive datasets and real-time processing. These require colossal computational resources, and that translates into unprecedented energy demand. The new IEA report outlines how AI training and deployment workflows use significantly more energy than traditional digital services.

Each time a user sends a prompt to a generative AI platform, the system responds by running through layers of complex neural networks hosted on power-hungry GPU clusters within data centres. Multiply that by millions of users interacting with these models daily, and the total energy cost becomes staggering.

According to the report, AI-driven data centres to double electricity use by 2030 is not just a projection—it’s an unfolding reality. Data centre electricity demand has already been rising at a rate of 12 percent per year for the last half-decade. Now, with AI development accelerating, that trajectory is steepening.

Where the power is going—and who’s using it

Currently, the United States, Europe, and China account for approximately 85 percent of global data centre electricity consumption. These regions are home to the largest clusters of tech innovation, the most significant investments in AI, and the densest networks of hyperscale data centres.

In response, major tech companies are aggressively pursuing alternative power solutions to offset their mounting energy bills and reduce their carbon footprint.

Google, for example, inked a deal in 2023 to source electricity from small modular nuclear reactors (SMRs) in the U.S. This bold move signalled the company’s intent to decouple its AI ambitions from fossil fuel-based power grids. The reactors are expected to provide steady, low-carbon energy for Google's expanding network of AI servers.

Similarly, Microsoft has committed to drawing power from nuclear reactors located at Three Mile Island, Pennsylvania—the infamous site of a nuclear meltdown in 1979. While the location is historically controversial, Microsoft’s move reflects a growing belief that nuclear energy may be essential to sustainably support the rise of AI.

Amazon, too, entered into a nuclear energy agreement last year, reinforcing the trend of big tech aligning itself with long-term, stable sources of low-emission energy. All three companies—Google, Microsoft, and Amazon—recognise that AI’s energy appetite cannot be fed by renewables alone, especially during peak demand periods or in regions with inconsistent solar or wind generation.

Energy security and carbon concerns

The IEA’s warning that AI-driven data centres will double electricity use by 2030 comes at a critical moment in the global energy conversation. As countries strive to reach net-zero carbon emissions by mid-century, ballooning energy demands from the tech sector threaten to outpace decarbonisation efforts.

The IEA report doesn’t downplay the benefits AI can bring to the energy world. Smart grids, predictive maintenance, and real-time monitoring are just a few of the ways AI could enhance energy efficiency and lower emissions in other sectors. However, these gains could be nullified if the foundational infrastructure supporting AI itself becomes a runaway source of CO2.

If left unchecked, the energy demand from AI and data centres could undermine hard-won progress made in renewable adoption, energy conservation, and emissions reduction. That’s why the IEA is calling on governments and corporations alike to implement clear energy-use standards for data centres and invest in innovative cooling systems, higher-efficiency hardware, and renewable integration strategies.

A fork in the road: Opportunity vs. overconsumption

IEA Executive Director Fatih Birol noted that “AI offers tremendous opportunity to accelerate energy transitions, but only if its own energy footprint is responsibly managed.” This dual nature of AI—as both a problem and a solution—underscores the challenge ahead.

On one hand, advanced AI models are being used to fine-tune power grids, optimise battery storage, and predict weather patterns for solar and wind deployment. On the other, those same models are driving up energy demand at a rate faster than many utilities can supply sustainably.

If trends continue as they are, the world may face a paradoxical future: one where smarter systems coexist with more emissions, where gains in efficiency are undercut by rising consumption, and where the tech sector’s pursuit of innovation runs up against planetary boundaries.

Policy recommendations and next steps

The IEA's report concludes with several key policy recommendations. It urges governments to:

Develop regulatory frameworks to improve data centre energy transparency.
Mandate energy efficiency benchmarks for AI computing infrastructure.
Encourage public-private partnerships to build low-carbon energy supply chains.
Invest in R&D for next-generation cooling, chips, and quantum processing.

Without such interventions, the path forward may become increasingly unsustainable.

AI’s energy future must be intentional

The forecast that AI-driven data centres will double electricity use by 2030 is both a warning and a call to action. As the world embraces artificial intelligence, the systems that support it must evolve too—intelligently, sustainably, and urgently.

If energy policy and technological ambition can work hand-in-hand, AI’s immense potential doesn’t have to come at the planet’s expense. But the clock is ticking.

The future of AI isn’t just about algorithms and datasets—it’s also about watts, emissions, and the choices we make today.