AI Giants Set to Consume 945 TWh of Power by 2030 as US and China Lead the Charge

According to a report by the IEA, nuclear energy's share of the electricity supplied to data centers is expected to rise from 2013 to 2035, with significant growth anticipated in both the U.S. and China.

IEA’s comprehensive special report, Energy and AI According to projections, the energy requirements for AI are expected to double to approximately 945 terawatt-hours by 2030, slightly exceeding current annual electricity usage in Japan. The report indicates that AI will be responsible for escalating power consumption, primarily due to an anticipated fourfold expansion in AI-tuned data center operations by 2030.

An examination of the shares

Forecasted usage statistics for leading nations across the upcoming half-decade highlight considerable changes in the power supply sources for data centers.

Currently, nuclear power accounts for just 15% of the energy used by data centers in the United States. However, the emergence of small modular reactors (SMRs) might significantly boost this percentage. By 2035, nuclear and other low-emission sources could provide more than 55% of these facilities' power needs, thereby decreasing dependence on newly constructed natural gas plants.

Currently, China’s data centers predominantly use coal for about 70% of their energy needs, but this landscape is shifting. From 2024 to 2030, both coal and renewable sources are projected to contribute roughly an additional 90 terawatt-hours (TWh) each. After 2030, nuclear power might become more significant thanks to small modular reactors (SMRs). By 2035, renewables and nuclear together could make up 60% of the electricity supply, leading to a reduction in coal usage.

Europe is speeding up its shift towards cleaner energy sources with an aim to have renewables and nuclear provide 85% of the power needed for data centers by the year 2030.

Together, Japan and South Korea, which currently make up 5% of the world’s data center energy consumption, plan to boost their clean power usage (including renewables and nuclear) from 35% to almost 60% by the year 2030.

Across the remainder of the globe, such as in India and Southeast Asia, despite coal being predominant at present, renewable energy sources are increasingly making their mark, with clean energy expected to overtake coal in powering data centers by 2035.

The Tale of Nuclear Technology and Artificial Intelligence

Despite encountering specific data challenges, the nuclear sector continues to integrate artificial intelligence. Although AI is currently experiencing significant growth, it will not accelerate nuclear fusion or make small modular reactors (SMRs) operational before 2030—largely due to non-technical obstacles such as regulatory and logistics issues rather than technological limitations.

These constraints are primarily associated with regulatory approvals, extended construction periods, and difficulties in establishing new industrial supply chains.

Although AI appears to be highly beneficial for the nuclear sector, its influence is constrained by various regulatory, economic, and geopolitical elements that shape the nuclear industry.

A brief overview

The global requirement for AI technology Will necessitate as much energy as what is currently consumed by Japan. However, renewable energy sources might cover only about half of this requirement.

According to the report, by 2030, the electricity used for AI data processing in the US will surpass the total power consumed by manufacturing, steel production, chemical industries, and all other heavy-energy sectors combined.

By 2030, processing data primarily for artificial intelligence purposes is projected to require more electrical power in the U.S. than would be needed for producing steel, cement, chemicals, and all other heavy industries collectively, as stated in the report.

A single data center now uses as much power as 100,000 homes, yet some being built presently will demand up to 20 times more energy than that.

Berbagi :