Cooling the AI boom: how hydraulic modelling supports data centre water approvals
At a glance
The rapid development of data centres to support Artificial Intelligence (AI) is placing unprecedented short‑term pressure on urban water networks. Hydraulic modelling can test how these demands interact with existing systems and compare mitigation options before capital decisions are locked in. Our work on the data centre water supply shows that off‑peak supply supported by on‑site storage can meet high water demands at lower cost than network upgrades while maintaining customer pressures.
The water challenge behind data centre growth
AI-driven data centre growth is accelerating faster than traditional water planning cycles. In Sydney alone, data centres could use the equivalent of 25% of the city’s annual drinking water supply by 2035 for cooling. At a global scale, total water use across around 100 data centres is forecast to rise from 7.9 billion litres in 2020 to 18 billion litres by 2030, an increase of about 150%.
For water authorities, the issue is not just total demand but timing. Data centres are often delivered quickly to meet commercial timelines, while major water infrastructure projects take many years to plan and build. This mismatch creates supply gaps at the local network level. Without careful assessment, approving large new users such as data centres can risk pressure shortfalls for existing customers or trigger costly upgrades earlier than planned.
While lower water consumption cooling options are available, these can be more energy intensive and costly, targeted hydraulic modelling can provide a way through this challenge. By understanding how and when water is used, and how that demand interacts with network constraints, authorities can consider interim servicing strategies that support development now while protecting long‑term resilience.
A data centre water supply case study: short‑term servicing for a large data centre
We completed a short‑term water servicing assessment for a large data centre, focused on meeting demand in 2028. The development required a reliable supply of almost 3 megalitres per day (ML/d), primarily for cooling towers. Initial checks suggested the local water supply zone would struggle to accommodate this demand without intervention.
Using a calibrated hydraulic model, we assessed the existing network with and without the data centre demand. The modelling showed that a constant supply of almost 3 ML/d would adversely impact pressures at up to 30 customer points, breaching minimum requirements set by the water authority. This confirmed that mitigation would be needed for approval.
The assessment focused on two pressure mitigation options: targeted network pipe upgrades and an off‑peak supply strategy supported by on‑site storage.
Option 1: network pipe upgrades
This option involved upgrading over a kilometre of DN550 trunk mains in the local network. Modelling showed that these upgrades resolved the identified pressure issues and allowed the data centre to receive its required flow without affecting other customers. Reservoir and pump station assessments confirmed sufficient capacity under this scenario.
While technically effective, the option came with a higher capital cost and longer delivery timeframe compared to the other option assessed. It also provided little headroom for future increases in demand from the site without further works.
Option 2: off-peak supply with on-site storage
The second option applied a special, off-peak diurnal supply pattern. Under this approach, higher inflows are delivered to the site during periods of low network demand. On site storage tanks then buffer this supply, releasing water steadily throughout the day to meet operational needs.
Modelling demonstrated that this strategy could deliver over 4 ML/d without adverse impacts, maintaining customer pressures above pre-development levels. In effect, aligning supply with off peak conditions unlocked significantly more capacity from the existing network than initially estimated. This approach avoided the need for immediate trunk main upgrades while still meeting performance standards.
Informing future planning, not just one approval
Network upgrades addressed the low system pressure problem through capital investment in shared assets. The off‑peak strategy addressed it through operational change and site‑specific infrastructure.
The modelling showed that large data centre demands can be met in the short term without compromising network pressures (or reverting to higher cost/energy consumption technology), provided demand is managed in time as well as in volume. Off‑peak supply supported by on‑site storage emerged as a lower‑cost alternative that also provided flexibility for future demand growth at the site.
Just as importantly, the assessment supported the approval process. Clear evidence on pressure impacts and mitigation performance allowed the water authority to set informed conditions and align them with longer‑term network planning.
While this assessment focused on a single development, the insights extend further. As more data centres cluster in specific precincts, cumulative impacts will become a key consideration. Hydraulic modelling provides a framework for testing these scenarios and identifying when interim measures reach their limits.
For planners, this creates an opportunity to integrate data centre servicing strategies into broader water supply management. Off‑peak demand profiles, storage requirements and trigger points for upgrades can all be defined early, reducing uncertainty for both utilities and proponents.
Wider relevance of this work
The challenges observed in this case study are not unique. Across the developed world, data centre development is accelerating against a backdrop of constrained urban water networks and long lead times for new infrastructure. Similar short‑term servicing challenges are expected to arise, along with the need for pragmatic mitigation strategies that balance growth with community needs.
The same modelling‑led assessment approach can be applied in this context. By testing how off‑peak supply, storage and targeted upgrades perform under local conditions, authorities can make confident decisions that support economic development without undermining network performance.
As AI continues to scale, the water implications of data centres will remain firmly on the agenda. Hydraulic modelling offers a practical way to navigate the gap between rapid data centre development and slower water infrastructure delivery. For water authorities, it provides clarity. For data centre proponents, it opens up servicing pathways that are both workable and cost‑effective.