AI, energy, and climate change are closely interconnected from several perspectives

According to a study by the World Economic Forum (WEF), the survival of data centers is inseparably linked to the climate crisis: extreme heat, drought, and other climate risks could increase the cumulative annual operating costs of data centers—currently around $3.3 trillion—by 2055.

The choice of where to locate data centers has become more complex, as it requires not only connectivity and space, but also energy availability and resilience to physical risks (extreme or chronic climate events). Beyond energy consumption—the latest chips can consume up to ten times more power than previous generations—it is also necessary to consider cooling requirements for servers, which demand large quantities of water.

It is estimated that in the United States, which accounts for over half of global data center capacity, the AI boom will increase water demand from 150 to 275 billion liters by 2028 (compared with about 60 billion liters in 2022), a potential fivefold increase in six years.

Another main reason for high freshwater consumption is the limited reuse of cooling water: some evaporates, and the remainder becomes wastewater often contaminated with dust, chemicals, and minerals, which reduces cooling efficiency if recirculated. As a result, data centers are often unable to fully reuse wastewater.

Major industry players are already facing pushback from local communities in regions that are already suffering from water scarcity and do not want to passively accept the drainage of this precious resource by data centers.

According to the International Energy Agency (IEA), there is also considerable uncertainty regarding data center energy consumption, both current and future. While the technology sector evolves rapidly and a data center can be operational in two to three years, energy systems require longer lead times to plan and build infrastructure, as well as substantial upfront investment.

The United States, China, and Europe will remain the regions with the highest electricity demand for data centers in the coming years, but other regions, such as Southeast Asia, are also experiencing strong growth, generating increasing electricity demand.

Regarding the energy mix linked to AI, coal—accounting for about 30%—is the main electricity source, with significant regional variations (led by China). Renewable energy currently supplies roughly 27% of electricity consumed by data centers globally; natural gas is the third largest source, followed by nuclear. The IEA also forecasts that CO2 emissions from electricity generation for data centers will peak by 2030 before beginning a slight decline by 2035. Despite rapid growth, data centers remain a relatively small part of the overall energy system, representing less than 1% of global CO2 emissions.

On the other hand, applying AI to the management and maintenance of power plants could generate potential cost savings of up to $110 billion per year by 2035, through reduced fuel use and operational costs.

AI also enables greater integration of renewable electricity into the grid by improving transmission capacity on existing lines. Finally, it enhances the accuracy of weather forecasting and the ability to analyze climate change, which is essential to optimize operation, planning, and resilience of energy systems.

AI thus presents a double-edged role with respect to energy and the environment: careful evaluation of risks, opportunities, and impacts by policymakers and companies will be crucial to ensure that the ultimate balance for human and planetary health is, ideally, positive.