The Battle for Energy and Water in the Age of Artificial Intelligence
The operation of the data centers is based on an immutable physical principle: almost all electrical energy is converted into heat, which must be continuously dissipated to prevent processors and electronic circuits from overheating. Cooling is now one of the most significant cost factors. Despite industrial electricity contracts priced at 0.06 to 0.10 €/kWh, the total cost remains enormous for facilities with capacities of hundreds of MW.
At the same time, water consumption is becoming a critical issue. The Water Usage Effectiveness (WUE) figures reported by companies reflect ideal conditions. However, in practice, when temperatures rise or when the arrays are operating at maximum capacity, water consumption can reach 1.5 to 2 liters per kWh.
The surge in computing power is a decisive factor. From 5 to 10 kW per rack just a few years ago, today’s requirements range from 40 to 100 kW—or even more—per rack. Air cooling is no longer sufficient. The industry is turning to liquid cooling technologies, such as direct-to-chip or immersion in special fluids. These solutions reduce direct water consumption to nearly zero, but require high installation costs. Microsoft reports that it reduces water consumption by more than 125 million liters per data center annually, but the average for its network remains at 0.30 L/kWh.
The water-energy dilemma persists. Evaporative cooling systems reduce electricity consumption but require large amounts of water. “Dry” systems do not use water, but they increase electricity consumption. The UN warns that a one-sided focus on carbon reduction could exacerbate the water crisis.
Globally, the picture is even more alarming. By 2030, data centers powering AI will consume enough water to meet the basic needs of 1.3 billion people. The electricity consumption of artificial intelligence systems could reach 945,000 MWh, while the generation of this energy—particularly from thermal and nuclear power plants—also requires large amounts of water. Furthermore, the manufacture of microchips requires tens of thousands of cubic meters of ultra-pure water daily.
On a global scale, power in the age of artificial intelligence is determined by four factors: cheap and reliable electricity, sufficient water, advanced cooling technologies, and access to semiconductors. The U.S. has the advantage of scale but is vulnerable to water shortages and network congestion. China is expanding rapidly but is constrained by water shortages and semiconductor shortages. Europe faces high costs and regulatory barriers. India is emerging as a new power, with a massive population and strategic partnerships, but it faces extreme temperatures and water shortages. Russia has resources but lacks access to advanced technology.
The future of digital sovereignty will depend on the ability of nations to organize their space around the needs of computing power. This includes the siting of large clusters, water resource management, strengthening of electrical grids, and the implementation of closed-loop cooling technologies. The use of treated wastewater and the reuse of waste heat will be key tools.
Digital dominance will increasingly depend on hydrological dominance. In a world where artificial intelligence requires enormous natural resources, the management of water and energy is becoming the foundation of global power.
An older 1 MW data center can consume up to 25,500 cubic meters of water per year, an amount equivalent to the consumption of 100 European households. On a 300-MW scale, these needs become structural constraints for entire regions.
Kontra News/ Opinions, Sunday, June 29, 2026
