New high‑water‑demand industrial and digital investments are arriving in Hungary; the planned AI campus beside Paks is a clear illustration. The project’s announced centerpiece is an AI data center with up to 96 megawatts of power draw, accompanied by a 530‑hectare agrivoltaic park and battery storage — the initiative is currently at the memorandum‑of‑intent stage (ParTec and 3D Lézertechnika signed in July 2025).
In the same region, an unusually low Danube level in summer 2026 caused operational difficulties at the Paks Nuclear Power Plant: shutdowns began at the end of July, by early August three of four blocks were offline, and on August 10 the 2nd block returned to nominal output. According to MVM, the operating blocks required roughly 100 cubic meters per second of cooling water, while safe cooling for all four shut blocks would need at most 2.5 cubic meters per second in total. This episode showed that even a long‑standing, riverside major facility can be constrained by extreme hydrological conditions.
The real question: infrastructure responsibility and legacy value
For a data center, water demand tied to computing and power is only the starting point. The strategic issue is whether to equip each site with dedicated wells and pipes, or to invest in regional, multi‑user water systems that can serve industrial parks, farms, towns and landscape water needs. A regional system with trunk lines and storage can shift water from wetter periods into dry spells and substitute some surface water for groundwater withdrawals.
How much water might a data center actually use?
Site water use at data centers is typically measured by WUE (Water Usage Effectiveness), liters per kWh of IT energy. Technology makes a large difference: Microsoft’s average WUE in its own centers fell from 2.3 l/kWh in early generations to 0.27 l/kWh by 2025. A quick example: a 100 MW IT capacity (close to the announced 96 MW) at continuous load would consume about 5,520 m3/day at 2.3 l/kWh, but only about 648 m3/day at 0.27 l/kWh. The first corresponds to roughly the daily water use of a 50,000‑person town, the second to a town of about 6,000.
Thus, megawatt figures alone are insufficient: planners must know the cooling technology, annual and summer peak water demands, the water source, reuse rates and drought‑operation modes.
Financing, market signals and the missing investment
The OECD’s 2026 report on Hungarian water management notes that the average cost‑recovery ratio for water utilities' operating and maintenance costs fell from 99% in 2009 to 79% in 2018. The report also estimates an annual investment gap of roughly €600 million needed to meet EU water objectives. Long‑frozen household tariffs have tightened utilities’ investment room.
Regional trunk pipelines, reservoirs, pump stations and recycled water systems are capital‑intensive and long‑lived; single irrigation associations, small municipalities or nascent industrial parks typically cannot shoulder such costs alone. That is where a creditworthy, long‑term contracting ‘‘anchor customer’’ becomes valuable — a large data center can play that role.
International example: Loudoun County, Virginia
Loudoun County is a global data‑center hub. Loudoun Water began building a separate pipeline network in 2010 to deliver treated wastewater for irrigation and industrial cooling to meet rising data‑center demand. In 2025 the system supplied over 750 million gallons (nearly 2.8 million m3) of recycled water, offsetting an equivalent volume of potable water. Loudoun is also developing a raw‑water storage in a disused quarry with about 3.8 million m3 capacity to support treatment when Potomac flows are low. This integrated model — dependable industrial demand, multi‑user network, recycled water and storage — is instructive for Hungary.
Financial risks and mitigation
A major risk is that a data center may close, downsize, or adopt much less water‑intensive cooling in twenty years, turning a single‑customer network into a stranded asset. Mitigation measures include:
- modular, phased construction tied to additional user contracts;
- anchor customer minimum capacity fees and guarantees;
- open trunk infrastructure to allow new industrial, agricultural and municipal offtakers;
- long‑term agreements that handle price indexation, drought restrictions, energy price shifts and early‑exit compensation.
Bankability comes from predictable, contracted long‑term revenue — the same principle used in projects such as the King Abdulaziz International Airport desalination project, where a 20‑year take‑or‑pay contract enabled private financing.
Heat as a second product
Data centers generate substantial waste heat. In a Finnish project with Fortum and Microsoft, once fully built, data‑center waste heat could meet about 40% of the annual district heating demand for roughly 250,000 customers; heat recovery rollout began in 2027. The Paks campus developers already mentioned agricultural heat use. Realising heat value requires customers, pipelines, heat pumps, seasonal demand profiles and long‑term heat purchase contracts.
Ownership, regulation and public interest
Ownership form alone does not determine a water system’s long‑term condition. Allocation rules, ecological limits and drought‑priority decisions require strong public oversight. A possible model is a regional infrastructure company where state and municipalities ensure public interest control while institutional or private capital helps finance and operate the asset. Fees must cover operating, maintenance and replacement costs over the long run; state oversight preserves water‑resource protection and equitable access.
Conclusion: the regional water system is the real stake
The Paks announcement and the 2026 Danube episode create a rare opportunity. Large industrial and digital investments bring new water demand: a risk, but also a chance to launch regional water and heat infrastructure that Hungary needs regardless of data‑center siting. Surface water can be delivered to inland regions, storage can buy time, some groundwater extraction can be substituted, and waste heat can serve towns or greenhouses. The Paks data center could be among the first large anchor customers.
However, the water plan must be as integral to project preparation as the business and energy plans: sources in normal and drought years, storage strategies, connection terms, and long‑term financial guarantees must be defined. Without those answers, the result will be a company‑sized technical by‑product rather than enduring regional infrastructure. The true value of pipes and reservoirs becomes evident when they outlast the investment that built them — hence, for the next major industrial project, the water plan must sit on the table alongside the business plan and the energy plan.
