OpenAI Data Centre Drops Water Recycling Plan
  • The proposed 612-megawatt S7 data centre will use waterless chip cooling after recycled water infrastructure proved unavailable.
  • The alternative cooling system may increase electricity demand as Sydney’s transmission network faces limited capacity for large new data centres.
  • New South Wales will require the project to meet a power usage effectiveness limit of 1.3 and disclose expected water consumption.

Developers of a 612-megawatt data centre planned for OpenAI have abandoned a proposal to cool the facility using recycled water. The project will instead rely on a waterless system that experts expect to consume more electricity.

The S7 facility, being developed by Australian data centre operator NEXTDC, could become one of the world’s largest data centres. Planning documents filed in April showed the company had considered using treated sewage from Sydney’s water network.

That approach would have reduced demand for drinking water and limited the amount of electricity needed for cooling. However, NEXTDC ended discussions with potential suppliers because the required pipeline infrastructure lacked planning approval.

“Recycled water … has real merit and it can support lower PUE and better overall sustainability and NEXTDC considers it wherever that infrastructure exists,” CEO Craig Scroggie said in an email.

CEO Craig Scroggie

PUE, or power usage effectiveness, measures how much electricity a data centre uses beyond the computing equipment itself. Cooling systems are a major contributor to that additional energy demand.

“Potential sources were canvassed in early planning. In Western Sydney, the enabling infrastructure … is not yet in place,” Scroggie added.

Water constraints shift pressure to the grid

NEXTDC plans to circulate liquid coolant around computer chips to capture heat. Fans will then remove the heat from the facility.

The system requires no water during normal operation. However, analysts said it could use more electricity than a cooling system supplied with recycled water.

“If we take the current situation at face value it will be more energy intensive, so NEXTDC will have to address these concerns,” said Sarvaesh Mohan, an analyst at data centre consultant DCByte.

The change exposes a difficult trade-off for governments seeking to attract AI investment while controlling its environmental impact. Australia faces particular pressure because water scarcity remains a persistent national risk.

NEXTDC said the S7 facility would not use drinking water for cooling. Yet the move away from recycled water could deepen concerns about electricity supply in Western Sydney.

Transgrid, the transmission operator serving Sydney, has warned that the network has limited capacity to accommodate large data centre connections without significant upgrades.

The company said developers would need to help finance any expansion required to connect new facilities.

In May, Transgrid Executive General Manager Jason Krstanoski told a New South Wales parliamentary inquiry that “existing regulatory arrangements were not designed for this level of large, clustered load growth”.

Transgrid Executive General Manager Jason Krstanoski

RELATED ARTICLE: Uber Launches AI Assistant Powered by OpenAI’s GPT-4o to Accelerate Electric Vehicle Transition

Infrastructure development trails AI demand

The cooling decision also reflects the difference between the rapid growth of AI infrastructure and the slower development of utilities.

Mohan said the lack of a recycled water connection illustrated “the wider gap between the current AI infrastructure boom and the slower deployment cycles of local utilities”.

Recycled water provider coNEXA was identified in planning documents as a possible supplier. Its chief executive, Kurt Dahl, declined to comment on the S7 decision.

However, he said connecting data centres to recycled water systems should be treated “with urgency, as data centres are currently making long-term ⁠choices between energy and water for their preferred approach for cooling”.

Economists expect AI-related construction spending in Asia to exceed $1 trillion by 2030. Governments are competing for investment, computing capacity and associated economic activity.

At the same time, communities are challenging data centre projects over electricity demand, water consumption, land use and pressure on local infrastructure.

Regulation tightens around resource use

The Australian government has announced plans to introduce energy and water limits for data centres as part of broader AI safeguards.

New South Wales has already begun imposing more detailed requirements. A review of 35 data centre applications found authorities started requesting water consumption estimates in late 2025.

Officials have also introduced fixed PUE limits for some developments. The OpenAI and NEXTDC project is one of six facilities subject to the new requirements.

S7 must disclose its expected water consumption and comply with a PUE limit of 1.3. That figure sits near the more efficient end of global data centre performance, although NEXTDC has not published a target for the project.

The company previously reported using its waterless cooling system for 50 megawatts of computing capacity. S7 would operate at more than 12 times that scale.

Independent Senator David Pocock said federal rules must address broader public interests, including mandatory requirements for “local sovereign compute power access, water usage, renewable energy as well as appropriate planning controls to ensure they aren’t using land better used for housing or creating adverse living conditions for existing communities”.

Senator David Pocock

For investors and technology executives, S7 shows that securing land and capital is no longer enough. Grid access, water infrastructure and planning approvals are becoming central constraints on AI expansion.

Australia’s policy response may influence other markets confronting the same tension. Governments want sovereign computing capacity, but the infrastructure supporting it must develop without transferring excessive costs or environmental risks to communities.

The original article can be found here:

Categories: International, News

Share