Water Is the Real Grid: The Unpriced Input That Quietly Governs the World's Power System

Electricity systems are planned around fuel, capital and megawatts, yet a single physical input constrains all of them at once. Water sits inside generation, cooling, digital infrastructure and the new fuels of the transition, and it is priced almost nowhere. This is the structural case for reading the global power system as a water system.

Every serious conversation about the energy transition is conducted in the language of electrons: gigawatts of capacity, terawatt-hours of demand, the cost of capital behind a project, the fuel mix behind a grid. That language is not wrong, but it is incomplete. Beneath the electron sits a physical input that no market prices consistently and that every part of the system depends on at the same time. Water generates electricity directly, cools the plants that generate the rest, sustains the data centres now consuming a fast-growing share of demand, and is embedded in the hydrogen and critical minerals meant to carry decarbonisation. When water becomes scarce, hot or contested, each of these functions degrades simultaneously. The proposition of this analysis is deliberately structural rather than seasonal: the global power system is, in a precise physical sense, a water system, and it is being managed as though water were free and unlimited.

Channel one: generation still runs on rivers

Hydropower remains the single largest source of low-carbon electricity in the world. The International Hydropower Association reports that hydro supplied 14.3 percent of global power, with generation rising roughly 10 percent in 2024 to 4,578 terawatt-hours as output recovered from a drought-driven trough; earlier IEA framing placed hydro nearer 17 percent of global electricity in 2020. Pumped-storage hydropower now exceeds 200 gigawatts of capacity, functioning as the world's largest fleet of grid-scale storage, a water battery that absorbs surplus generation and releases it on demand. A system this dependent on flowing water is a system exposed to the hydrological cycle, and 2023 demonstrated the exposure in full. The IEA's Electricity 2024 report documented a synchronous global hydro drought in which Canada, China, India, Türkiye, the United States and others experienced simultaneous shortfalls, pushing the global hydro capacity factor below 40 percent, its weakest level in three decades. The immediate consequence was not a gap in supply but a return to coal and gas to fill it, converting a water shortage directly into higher emissions.

Channel two: cooling turns heat into a water constraint

The larger and less visible dependency lies in thermal and nuclear generation, where water is the medium that carries away waste heat. In the United States, the USGS Circular 1441 assessment found that thermoelectric power accounts for roughly 41 percent of national water withdrawals, the single largest category. That water is not merely consumed; its availability and temperature determine how much a plant can produce. When rivers run low or warm, discharge limits and reduced thermal efficiency force plants to cut output precisely when demand for cooling peaks. The academic literature has quantified this derating with unusual clarity. Van Vliet and colleagues, writing in Nature Climate Change in 2012, projected summer generating-capacity reductions of 6.3 to 19 percent in Europe and 4.4 to 16 percent in the United States over 2031 to 2060. A later 2016 assessment, modelling more than twenty thousand plants worldwide, found that over 60 percent of the world's power plants could face reduced output between 2040 and 2069. The technology-specific water factors compiled by Macknick and colleagues in 2012 remain the standard basis for these calculations, translating each generating technology into cubic metres per megawatt-hour.

Channel three: the digital load arrives thirsty

A new and rapidly scaling demand centre has entered the same water balance. The IEA's Energy and AI analysis estimates that data-centre electricity consumption stood at around 415 terawatt-hours in 2024 and could reach roughly 945 terawatt-hours by 2030, a trajectory driven substantially by artificial intelligence. That electricity carries an embedded water cost through the power plants that supply it, and a direct one through on-site cooling. The Lawrence Berkeley National Laboratory's 2024 data-centre report quantified both: direct cooling water for United States data centres of approximately 66 billion litres in 2023, alongside indirect water use of around 800 billion litres tied to the electricity they consumed. At the level of individual workloads, the figures are equally concrete. The analysis by Li and colleagues estimated that training a model of GPT-3's scale consumed on the order of 700,000 litres of freshwater, and projected that global AI could withdraw 4.2 to 6.6 billion cubic metres of water by 2027. The significance is not the absolute volume, which remains modest against agriculture, but the location: this demand is being sited for latency, land and power, rarely for water availability, and increasingly in regions already under stress.

Channel four: the new fuels inherit an old scarcity

The fuels and materials intended to decarbonise the system carry their own water intensity, which is frequently omitted from transition planning. Green hydrogen is produced by electrolysis, and the work of Beswick and colleagues establishes that each kilogram of hydrogen requires around 9 litres of purified water as a direct input, with substantially more consumed in the purification stage that feeds it. Against an electrolyser fleet the IEA records growing from 1.4 gigawatts installed in 2023 toward roughly 520 gigawatts of announced capacity by 2030, the aggregate water requirement becomes a siting constraint rather than a rounding error. The mineral side is comparable. Lithium extraction in the Atacama, assessed by Marinova and colleagues, carries a water footprint of around 442 cubic-metre-equivalents per tonne, drawn from one of the most water-stressed basins on the planet, and the IEA's critical-minerals work notes that more than half of global lithium and copper production is located in areas of high water stress. The transition, in other words, does not escape the water constraint; in several respects it concentrates it, placing the highest water demands in precisely the places least able to meet them.

The geopolitical layer: shared rivers and mega-dams

These four channels operate within a scarcity that is already acute and increasingly bounded by borders. Mekonnen and Hoekstra estimate that roughly 4 billion people face severe water scarcity for at least one month a year, and the World Resources Institute's 2023 analysis identifies 25 countries living under extremely high water stress. Where the resource crosses a frontier, energy infrastructure becomes an instrument of state. The 2024 UN World Water Development Report notes that about 40 percent of the world's population lives in transboundary basins, yet only around one in five countries has a cooperation agreement covering shared waters. The single global framework, the 1997 UN Watercourses Convention, rests on two principles in permanent tension: equitable and reasonable use, and the obligation not to cause significant harm. A dam built upstream to generate power is, from downstream, a decision about water security taken by another state. The Grand Ethiopian Renaissance Dam on the Nile, the cascade of dams on the Mekong, the Indus system and the Euphrates-Tigris are the defining cases, and the scholarship of Zeitoun and Warner explains why outcomes on shared rivers are shaped less by law than by asymmetries of power. The most recent peer-reviewed work, by AghaKouchak and colleagues, projects elevated conflict risk in roughly 40 percent of transboundary basins over 2041 to 2050. Türkiye, Central Asia and Pakistan appear here not as the boundary of the story but as instances of a global pattern.

The reverse current, and why this matters for emerging markets

The dependency runs in both directions. Sanders and Webber found that roughly 12.6 percent of United States primary energy is embedded in the movement, treatment and heating of water, so that water constrains energy and energy constrains water in a single closed loop. Desalination illustrates the trade at its sharpest: the global fleet assessed by Jones and colleagues comprises 15,906 facilities producing around 95 million cubic metres per day, with energy the principal barrier to scaling it. For emerging and fast-growing power markets the implication is direct and unforgiving. These are the systems adding hydropower, thermal capacity and, increasingly, data-centre and industrial load fastest, and they are disproportionately located in water-stressed and transboundary settings where the margin for error is thinnest. A capacity plan that models fuel and finance but treats water as an unpriced constant is not a conservative plan; it is an incomplete one. The strategic conclusion of this analysis, and the premise of the water-energy column it opens, is that water risk belongs inside the core of energy planning rather than in an environmental appendix. The real grid has always been the water beneath it.

Sources cited in-text: IHA World Hydropower Outlook 2025; IEA Electricity 2024, Energy and AI, Global Hydrogen Review, and critical-minerals work; USGS Circular 1441; van Vliet et al. (2012, 2016); Macknick et al. (2012); LBNL 2024 Data Center Energy Report; Li et al. (2023); Beswick et al. (2021); Marinova et al. (2024); Mekonnen & Hoekstra (2016); WRI (2023); UN World Water Development Report 2024; UN Watercourses Convention (1997); Zeitoun & Warner (2006); AghaKouchak et al. (2026); Sanders & Webber (2012); Jones et al. (2019). Figures are cited from institutional sources; copyrighted charts and tables are not reproduced. Analysis by UzEnergyNews. This Weekly Insight opens the UzEnergyNews water-energy column. Figures reflect public sources as of early July 2026.