The thirstiest part of AI's supply chain isn't the data centre, a UAE report finds: it's the power plant

The water cost of artificial intelligence is mostly not where people look for it. A report published by the UAE Ministry of Foreign Affairs on 7 October says electricity generation accounts for roughly half of water consumption across the data-centre value chain, with on-site cooling at about 30% and primary energy supply and manufacturing making up the rest.

The report, Crosscurrents: Navigating the AI–Water–Energy Nexus, was prepared by the UAE's Permanent Mission to the United Nations with analysis from Eurasia Group and released on the sidelines of the 81st UN General Assembly in New York. It is a follow-up to the UAE's 2023 report Ripple Effect, and it is written as input to the 2026 UN Water Conference, which the UAE and Senegal co-host in Abu Dhabi from 8 to 10 December.

The numbers

The report projects global data-centre electricity use at 448 terawatt-hours (TWh) in 2025, with a water footprint of 4.5 trillion litres; AI-focused facilities account for 93 TWh and 900 billion litres of that. By 2030 it projects electricity use nearly doubling to 945 TWh and water use reaching 9.32 trillion litres, with AI workloads responsible for 378 TWh and 3.7 trillion litres. On those figures the water attributed to AI roughly quadruples in five years, and its share of the total rises from about a fifth to about two-fifths.

It also says total consumption is projected to more than double by 2030, as water used for cooling nearly triples and water tied to electricity generation roughly doubles, and that inference, where a trained model answers a query, accounts for up to 90% of a model's lifecycle energy use.

Why the power source matters

Because most fossil and nuclear generation relies on steam-cycle cooling, the report calls the choice of power source “one of the most direct levers available for reducing AI’s water footprint.”. Wind and solar PV avoid that cycle. But it adds a catch: AI workloads run continuously at near-constant intensity, so matching them to wind and solar typically needs battery storage, hybrid generation or backup capacity, often natural gas.

On cooling, it finds no free option: evaporative cooling saves energy but uses water, while air-side economisation and closed-loop liquid cooling save water but need more electricity. The report puts it plainly: “No technology can reduce both water and energy use at once without fundamentally changing the cooling method.”

Where projects are already being stopped

The report treats community opposition as one of the most immediate checks on AI infrastructure. In the US, it says, nearly $300 billion of projects have faced delays or cancellations since 2023, and in the first quarter of 2026 at least 75 projects worth about $130 billion were delayed or blocked. It cites a US government research report as finding that a single 100 MW facility can draw as much water each day as 2,600 households.

Outside the US it points to Malaysia, where residents of Iskandar Puteri protested a planned 300 MW facility in February 2026 in what it calls the country's first reported data-centre protest, and where Johor has since banned the most water-intensive data-centre categories; to Chile, where Google redesigned its proposed Cerrillos data centre in Santiago to air cooling after a court partly annulled its environmental approval over aquifer concerns; and to Ireland, where large data centres must now meet most of their demand with new Irish renewables.

What the UAE is asking for

The report makes five recommendations: a global declaration on efficient water use in AI, to be advanced at the UN Water Conference; integrating AI infrastructure into national and basin-level water and energy planning; transparent, comparable reporting of direct and indirect water and energy use; promoting low-water, low-energy technologies and circular resource systems; and embedding water and climate risk in corporate siting and investment decisions.

Mohamed Abushahab, the UAE’s Permanent Representative to the UN, said in the ministry’s release: “AI is developing at extraordinary speed, and we have a responsibility to ensure that the infrastructure supporting this transformation is resilient by design. That means thinking about water and energy from the beginning, not as constraints to be addressed later.”

Abdulla Balalaa, Assistant Minister of Foreign Affairs for Energy and Sustainability, said: “The question is not whether AI will transform our water systems — it already has. The question is how we can plan for that transformation wisely.”

Why it matters for the Gulf

The report places the Middle East and North Africa among the regions where data-centre presence and projected water stress overlap most, alongside southern Europe and parts of South and East Asia. Our reading for the region, where power generation and water supply are closely linked: on the report's own breakdown, the choice of electricity supply for an AI facility can matter as much for its water footprint as the choice of cooling plant. The Stratawire is accredited to cover WETEX in Dubai from 20 to 22 October, where we will put that question to water and power suppliers.

The projections in this article are the report's own; the report was supported by analysis from Eurasia Group. It notes that its water-stress figures are national and that conditions vary widely within countries.

How we made this: The Stratawire's AI newsroom prepared this article from the full report and the ministry's release, both linked below. A second AI model from a different company checked every figure and quotation against them, and an editor reviewed it before publication.