OpenAI’s planned artificial intelligence data centre in Australia has become the focus of a growing debate over the environmental impact of AI infrastructure after developers decided to abandon plans to use recycled water for cooling, opting instead for a waterless cooling system that could increase electricity consumption.
The decision highlights the increasingly complex balancing act facing technology companies as they expand AI infrastructure while attempting to meet sustainability goals. As demand for generative AI continues to rise, data centres have become critical assets, but they are also drawing attention for their intensive use of energy and natural resources.
The proposed facility, known as the S7 data centre, is being developed in Western Sydney by Australian data centre operator NEXTDC, with OpenAI serving as a strategic partner. Once completed, the site is expected to become one of the largest AI-focused data centres in the world, designed to provide the immense computing power required to train and operate advanced artificial intelligence models.
Originally, the project was expected to incorporate recycled wastewater into its cooling system. The approach would have allowed the facility to reduce its dependence on potable water while maintaining efficient cooling for thousands of high-performance processors operating continuously.
However, developers have now confirmed that those plans have been abandoned after determining that the necessary infrastructure to supply recycled water would not be available in time for the project. Without the required pipelines and supporting utility network, the recycled water system could not be implemented as initially envisioned.
Instead, the facility will rely on an advanced air-cooled liquid cooling system that circulates coolant through servers before dissipating heat using large-scale air cooling equipment. While this approach significantly reduces water consumption, experts note that it generally requires more electricity than systems that make use of recycled water for heat removal.
The change illustrates one of the growing challenges facing AI infrastructure projects worldwide. Data centres require efficient cooling to prevent overheating of powerful graphics processing units (GPUs), which generate substantial amounts of heat during AI training and inference. As AI models become more sophisticated and computationally demanding, cooling systems have become one of the most important factors influencing both operating costs and environmental performance.
Water has traditionally been an effective cooling medium because of its ability to absorb and transfer heat efficiently. Many large data centres use evaporative cooling systems that consume significant quantities of water, particularly during warmer months. While these systems improve energy efficiency, they have increasingly attracted criticism in regions where water supplies are limited.
Australia presents a unique challenge because it is one of the driest inhabited continents in the world. Periodic droughts, population growth and climate variability have placed increasing pressure on water resources, making water-intensive industrial projects subject to greater public and regulatory scrutiny.
Against this backdrop, recycled water had been viewed as a practical compromise. Rather than using drinking-quality water, the facility would have relied on treated wastewater that was unsuitable for human consumption but appropriate for industrial cooling applications. Such systems have already been adopted in several large data centres around the world as part of broader sustainability initiatives.
The inability to implement the recycled water plan demonstrates how environmental ambitions can be constrained by local infrastructure. While the technology exists, developers require dedicated pipelines, treatment facilities and regulatory approvals before recycled water can be delivered reliably to large industrial sites. In Western Sydney, those supporting systems are not yet sufficiently developed to meet the project’s timeline.
The revised cooling strategy also reflects the growing importance of energy efficiency in modern data centres. Air-cooled liquid systems eliminate much of the water consumption associated with traditional cooling methods, but they often require more electricity to maintain the same operating temperatures. This creates a trade-off between conserving water and increasing energy demand.
As governments worldwide seek to reduce greenhouse gas emissions, technology companies are under growing pressure to ensure that increased electricity consumption is offset through renewable energy sources and more efficient infrastructure.
The OpenAI-linked project arrives at a time when AI companies are investing billions of dollars in new computing facilities to support rapidly expanding demand for generative AI applications. Since the launch of advanced large language models, demand for specialized AI chips has surged, prompting cloud providers and technology firms to accelerate construction of hyperscale data centres across multiple continents.
These facilities require vast quantities of electricity not only to power servers but also to cool them. Industry analysts estimate that AI workloads consume significantly more energy than conventional cloud computing tasks because of the intensive mathematical calculations involved in training and operating advanced neural networks.
As a result, environmental concerns surrounding AI infrastructure have expanded beyond electricity consumption to include water use, land development and carbon emissions. Governments are increasingly introducing regulations requiring developers to disclose expected resource consumption and demonstrate how projects will minimize their environmental impact.
Australia has been among the countries strengthening oversight of large-scale data centre developments. New planning requirements encourage greater transparency regarding energy efficiency, water management and long-term sustainability, particularly for projects expected to consume substantial public resources.
The S7 project is expected to play an important role in Australia’s ambitions to become a regional AI and cloud computing hub. By hosting advanced computing infrastructure domestically, the country hopes to attract international investment, support local AI innovation and strengthen digital infrastructure across the Asia-Pacific region.
For OpenAI, expanding global computing capacity has become a strategic priority as usage of its AI models continues to grow. Reliable access to large-scale data centres is essential for training increasingly capable models while delivering AI services to businesses and consumers worldwide.
Although the revised cooling approach marks a departure from the project’s original sustainability vision, developers maintain that the chosen system represents the most practical solution under existing infrastructure constraints. They have also indicated that recycled water could still become an option in future expansions if the necessary public infrastructure is eventually developed.
The situation underscores a broader reality facing the AI industry. Building the infrastructure needed to support next-generation artificial intelligence involves more than installing powerful processors. It also requires reliable electricity networks, water management systems, transportation infrastructure and regulatory frameworks capable of supporting facilities operating on an unprecedented scale.
As AI adoption accelerates around the world, decisions such as the one made for the Australian data centre are likely to become increasingly common. Technology companies, governments and utility providers will need to work together to develop infrastructure that supports innovation while minimizing environmental impacts. The success of future AI expansion may depend not only on advances in computing technology but also on how effectively societies manage the resources required to power the next generation of artificial intelligence.
