Source: IMF
Artificial intelligence (AI) is an emerging source of productivity and economic growth that is also redefining employment and investment.
According to the scenarios contemplated in our last analysis, included in the April 2025 edition of the World Economics Perspectives (IO) of the IMF, AI can potentially increase the average rhythm of the annual world economic growth.
Now, AI also needs more and more electricity for data centers that make it possible. The pressure of that demand on electricity supply networks has notable implications for world demand for electricity.
According to the latest annual estimates of the Organization of Petroleum Exporting Countries (OPEC), the world data centers consumed more than 500 Electricity Teravatians-Hora in 2023. That total, which meant more than double the annual electricity consumption between 2015 and 2019, could triple until reaching 1,500 Teravatios-Hora in 2030, according to OPEC projections.
As the chart of the week shows, the electricity that only data centers consume, which is already located at the level of electrical consumption of Germany or France, would reach 2030 levels comparable to those of India, the third country in the world in electricity consumption. These levels would also far exceed those of consumption attributable to electric vehicles (VE) because, by the end of the decade, they would mean 1.5 times more electricity that sees them.
The place with the fastest growth in the energy consumption of data centers is the United States, where the highest concentration of centers in the world is given. According to the projection of McKinsey & Co. corresponding to a medium demand scenario, it is likely that the electrical need for the farms of the United States servers would multiply by more than three, exceeding the 600 Teravatios-Hora by 2030.
The boom in the construction of new warehouses to house data in the cloud and answer the questions aimed at AI makes the urgency faced by the authorities, which need effective energy strategies with which to ensure that an adequate supply is capable of responding to an increasing demand.
The growing energy demand of the technological sector will stimulate the total supply, which – if it responds well to the needs – will lead to a slight increase in electricity prices. On the other hand, if the answer is slower, there will be much more priced price increases that will harm consumers and companies, and possibly stop the growth of the AI sector itself.
If current energy policies are maintained, the increase in the demand for electricity due to AI could add about 1.7 gigatons of world emissions of greenhouse gases between 2025 and 2030, an amount similar to emissions related to the energy of all Italy for a period of five years.
The demand for computing and electricity by AI platforms is conditioned to wide uncertainty. The efficient open source AI models, such as Deepseek, reduce computing costs and electricity demand. However, lower costs make the use of AI grow, and reasoning models that consume more energy raise energy demand.
The net effect on electrical demand is still uncertain, which could delay energy investments and cause a price increase. The authorities and companies must collaborate so that AI reaches their entire potential and, at the same time, costs are minimized. Apply policies that encourage the use of multiple energy sources can improve the supply of electricity, help mitigate price increases and contain emissions.