In the old data center building of the Northwest Business Park in Western Sydney, the lights of the renovation construction were on all night, and the plan to install 8200 Nvidia Blackwell Ultra GPUs has made this long dormant existing building the most eye-catching new landmark in the Australian AI computing field.

This newly established long-term cooperation agreement directly pushes the delivery node of large-scale AI computing power clusters in Australia to the fourth quarter of 2026.

The largest local AI computing power cluster is about to land

The long-term cooperation between GreenSquare and Sharon AI is not just an ordinary data center leasing order, but a landmark deep binding in the Australian AI infrastructure field. As a cloud partner of NVIDIA, Sharon AI has won the first batch of core computing space in the first phase of the SYD1 data center park in Western Sydney. It plans to deploy approximately 8200 NVIDIA Blackwell Ultra GPUs and build a leading AI computing cluster in Australia's publicly announced projects based on the NVIDIA HGX B300 platform. The initial operation time is set for the fourth quarter of 2026.

The overall planned capacity of SYD1 park exceeds 100 megawatts, and the total scale of GreenSquare's under construction and planned data center projects in New South Wales and Victoria has exceeded 1 gigawatt. This cooperation agreement is a key milestone in GreenSquare's commercialization process and also marks a new stage in the construction of Australia's local AI infrastructure, completely bidding farewell to the scattered state of small-scale pilot projects in the past, and officially entering a new phase of centralized delivery of 10000 card level computing power clusters.

Transforming and reshaping the low-carbon path of computing power infrastructure through the use of old resources

Unlike the high carbon development model of traditional data centers that often involve leveling and rebuilding, the SYD1 Phase I project has taken a completely different path: adaptive upgrading and transformation of existing data center buildings. The project retains 92% of the structural carbon of the original building, which is equivalent to locking the large amount of carbon emissions generated by construction in the past few years directly into the walls and foundation. According to independent sustainability assessment data, the project has achieved a 76% reduction in carbon emissions per square meter of front-end structure, and a 93% reduction in carbon emissions per kilowatt of front-end structure based on IT capacity. The implicit carbon footprint of buildings has been compressed to an extremely low level in the industry. Keywords: computing power infrastructure, AI computing power track

The 80 MW plan for the second phase of the project has received official support from the New South Wales Government Investment Delivery Agency, and special mechanisms such as cross departmental coordination, power grid support, and regulatory process optimization have been implemented. The upgraded building is fully compatible with high-density computing deployment requirements, balancing low water consumption and high resilience architecture, and can continuously match the computing deployment requirements of AI customers in future iterations. It provides a replicable low-carbon and high resilience infrastructure construction sample for Australia's domestic sovereign AI computing system.Editor/Cheng Liting