The project for the massive 612 MW S7 data center in Australia, planned by OpenAI, has encountered infrastructural difficulties. Developers have abandoned the use of recycled wastewater for server cooling due to the inability to obtain permits to construct the necessary pipeline.
What Happened
OpenAI's partner, NEXTDC, is forced to abandon the use of recirculated water to cool the S7 infrastructure. Instead of the eco-friendly solution, the company is switching to more energy-intensive cooling methods, which could lead to an increased load on Sydney's power grid.
Context
The problem is linked to regulatory and infrastructural barriers: obtaining permits to build a pipeline to deliver water proved impossible. This forces operators to choose less efficient methods, which negatively impacts the PUE (Power Usage Effectiveness) metric and increases the project's carbon footprint.
Why It Matters for the Industry
This case demonstrates that scaling AI infrastructure is limited not only by chip shortages but also by the physical availability of resources—water and electricity. Bureaucratic obstacles can negate the technological advantages of modern resource management methods, increasing the Total Cost of Ownership (TCO) and operational expenses for providers.
Why It Matters for Users
The development of LLM-based services directly depends on the local sustainability of regional energy and water systems. Physical infrastructure constraints could lead to rising cloud computing costs and a slowdown in the deployment of new capacities in regions with strict resource regulation.
Sources
Author
Look at AI, Editorial Team