Global Energy Sector Redesigns Cooling Infrastructure as Rising Water Stress Threatens 31 Percent of Global GDP

Wood Mackenzie reports 31% of global GDP faces water stress by 2050. Learn how AI and power plants are pivoting to liquid and dry cooling to survive.

By: AXL Media

Published: Apr 2, 2026, 5:35 AM EDT

Source: The information in this article was sourced from Engineering News

The Growing Conflict Between Energy Production and Water Scarcity

The global energy landscape is facing a structural crisis as declining water availability threatens the cooling infrastructure essential for 80 percent of world power generation. Research from Wood Mackenzie indicates that by 2050, 31 percent of global GDP will be subject to high water stress, a significant increase from 24 percent in 2010. This environmental pressure is particularly acute for thermoelectric, nuclear, and hydroelectric facilities, all of which rely heavily on consistent water access to maintain operations. In Europe, recent episodes of high river temperatures and diminished flows have already forced output cuts and reactor curtailments, highlighting a systemic vulnerability that traditional infrastructure can no longer ignore.

Technical Shifts in Power Plant Cooling Systems

To mitigate these risks, the power sector is increasingly pivoting toward hybrid cooling and dry systems, as well as more strategic catchment site selection. Traditional once-through cooling methods, while low in consumption, require massive withdrawals of 132.5 cubic meters per megawatt-hour. In contrast, the current industry standard of wet recirculating towers reduces withdrawals to 4.6 cubic meters but triples actual consumption through evaporation. Jom Madan, a principal analyst at Wood Mackenzie, notes that while dry cooling can eliminate water use entirely, it imposes a seven percent efficiency penalty and adds significant capital costs of roughly $160 per kilowatt.

AI Computing Demands Drive Liquid Cooling Adoption

The rapid expansion of artificial intelligence is creating a parallel demand spike for cooling that air-based systems simply cannot sustain. Modern AI training nodes now exceed 120 to 200 kilowatts per rack, far surpassing the 20-kilowatt limit of traditional fan-based air cooling. This thermal reality is pushing data centers toward liquid cooling solutions, such as single-phase direct-to-chip systems that circulate warm water against processors. These advanced methods can handle loads of up to 150 kilowatts per rack, while specialized two-phase refrigerant systems are capable of exceeding 250 kilowatts.