IMF: Power demand for AI could equal that of the entire India
JAKARTA – Artificial intelligence (AI) is rapidly becoming a key driver of global productivity and investment growth.
However, this technological surge presents a significant challenge to the availability of electricity worldwide, according to the latest analysis by the International Monetary Fund (IMF) published in its April 2025 edition of the World Economic Outlook on the organisation’s official website (IMF.org).
The IMF acknowledges AI’s vast potential to accelerate global economic growth annually. However, its enormous electricity requirements could become a major obstacle without effective energy policy intervention.
According to the Organization of the Petroleum Exporting Countries (OPEC), global electricity consumption from data centres reached 500 terawatt-hours (TWh) in 2023—more than double the average level seen during 2015–2019.
This figure is projected to triple to 1,500 TWh globally by 2030, equalling the total electricity consumption of India, the world’s third-largest power consumer.
By the end of this decade, electricity consumption by data centres is expected to be 1.5 times greater than the total electricity used by electric vehicles (EVs) globally.
The United States, home to the largest number of data centres in the world, is experiencing the fastest growth in AI-related electricity consumption.
According to McKinsey & Co. projections cited by the IMF in its release on Friday (23/5), electricity consumption for AI servers in the US is expected to exceed 600 TWh by 2030.
This surge is fuelling a wave of large-scale construction of new data warehouses to store cloud information and process AI queries, making it imperative for policymakers to ensure a stable power supply.
If electricity supply fails to keep pace with this surging demand, prices could spike significantly.
The IMF warns that without new energy policies, AI could contribute up to 1.7 gigatonnes of additional global carbon emissions between 2025 and 2030—equivalent to Italy’s energy-related emissions over five years.
This scenario underscores the critical importance of collaboration between governments and the private sector. Policies that promote energy diversification could help stabilise supply and prices while curbing emissions.
Although energy-efficient models like DeepSeek have emerged, their impact may be negated by soaring demand for complex, reasoning-intensive AI models that consume far more electricity.
This uncertainty could delay energy investment and worsen potential crises in electricity supply and pricing over the coming years. (EF/ZH)