Rystad Puts Data Center Water Demand on Track to Nearly Triple by 2030
Rystad Energy's three cooling scenarios show a 256 billion liter spread between best and worst case by 2030, forcing water and power planning into unfamiliar territory.
Rystad Energy estimated on Friday (2026-08-28) that global data center water consumption could rise to 644 billion liters per year by 2030 if operators make no changes to cooling technology, nearly tripling the 222 billion liters consumed in 2025. The projection is driven by accelerating AI adoption, which has pushed data center infrastructure buildout to a scale that water and power systems were not designed to absorb.7
Rystad framed the 644 billion liter figure as the worst case. Active mitigation — deploying more water-efficient cooling systems — could bring the 2030 total to 388 billion liters in an aggressive conservation scenario, or 543 billion liters under a more moderate efficiency push. Even the lowest figure represents a 75% increase over 2025 consumption.7
The 256 billion liter gap between best and worst case is not a planning footnote. Water utilities, grid operators and hydrogen producers in markets where hyperscale development is concentrated face materially different infrastructure requirements depending on which path operators take. Cooling technology decisions being made now, as facilities are designed and financed, set that trajectory.7
The electricity load behind that cooling demand is already visible at the state level. Commercial electricity sales in Virginia rose by nearly 30 million megawatt-hours between 2019 and 2025, with the EIA attributing a significant portion of that increase to the state's dense concentration of data centers, according to Forbes reporting published August 23 (2026-08-23). Virginia remains the clearest existing example of what a high-density data center market does to regional power consumption.5,6
The United States carries that dynamic at national scale. US data centers consumed nearly 40% of global data center electricity in 2025, more than any other country including China, Canary Media reported on July 3 (2026-07-03). That share means American water and power systems bear a disproportionate portion of the cooling burden even as the buildout expands to new geographies.3
On-site hydrogen fuel cells have gained traction among data center operators as a way to reduce dependence on congested power grids, though the technology addresses grid exposure rather than cooling water use directly. Rystad projects fuel cell market revenues will rise roughly tenfold, from around $2.8 billion in 2025 to approximately $30 billion by 2030, driven largely by AI data center demand.2
Texas shows how state-level planning is struggling to keep pace with the buildout. The draft 2027 Texas state water plan forecasts the power sector will account for 3.8% of state water demand by 2030, up from 3.5% in 2023, E&E News reported in May (2026-05-20). The plan did not account for the full projected surge in data center water consumption, leaving an acknowledged gap between official projections and the trajectory implied by Rystad's estimates.1
Competition for renewable electricity introduces further pressure. Data centers and energy-intensive manufacturers are drawing on the same pool of clean power, with electricity potentially meeting 34% of industry's final energy consumption by 2035, Asian Power reported on August 21 (2026-08-21). Operators who adopt more electricity-intensive cooling alternatives to reduce water use will add to that competition.4
None of the three Rystad scenarios returns global data center water consumption to 2025 levels. Texas's incomplete water accounting illustrates how quickly official projections fall behind the physical buildout, and the 256 billion liter spread in Rystad's range is now the number water regulators, hydrogen suppliers and facility developers in high-concentration markets are working against. Whether operators invest in aggressive cooling efficiency or default to the path of least resistance will determine the scale of that gap.1,7