×

Sydney’s New AI Powerhouse: How GreenSquare Turned a 30-Year-Old IBM Data Centre Into a High-Density AI Campus

Sydney's New AI Powerhouse: How GreenSquare Turned a 30-Year-Old IBM Data Centre Into a High-Density AI Campus

Sydney’s New AI Powerhouse: How GreenSquare Turned a 30-Year-Old IBM Data Centre Into a High-Density AI Campus

Australia’s race to become a serious player in AI infrastructure just took a tangible step forward. GreenSquare DC has brought the first phase of its SYD1 data centre campus in Norwest Business Park, Western Sydney, into operation, fully leased to Nasdaq-listed Australian neocloud Sharon AI. The tenant is expected to deploy roughly 8,200 NVIDIA Blackwell Ultra GPUs at the site, in what is described as one of Australia’s most significant announced NVIDIA HGX B300 deployments.

A 30-year-old building reborn for the AI era

A 30-year-old building reborn for the AI era

Rebuilding old infrastructure for new workloads is an engineering puzzle. Legacy data centres were designed for a different era of computing, with lighter power loads and simpler cooling. GreenSquare kept the building’s existing structural slab but moved the data halls down to the ground floor, a change that let the facility carry far heavier, denser compute without extra reinforcement. The result: a 275 percent increase in IT capacity inside the original footprint. Chief technology officer Joshua Martens put it this way: “We saw an opportunity to give a 30-year-old data center a completely new purpose. By rethinking the building’s configuration, introducing prefabricated systems design, and using the latest technology to reduce carbon, we’ve created a facility that can support some of today’s most demanding compute requirements while using resources far more efficiently.” The modular approach helped deliver the turnaround in eight months, and it is a model other operators may copy: refurbish rather than rebuild.

Water, power and the sustainability question

Data centres are increasingly under scrutiny for their energy and water appetite, and Australia is no exception. GreenSquare has tried to get ahead of that conversation with the cooling design. The original cooling towers were replaced with a closed-loop system that operates without evaporative cooling for roughly 96 percent of the year. Rainwater harvesting is expected to offset about 10 percent of the facility’s annual cooling water requirements, cutting reliance on potable water.

These numbers matter because they directly address one of the loudest criticisms of the AI infrastructure boom. Australian regulators and communities are asking hard questions about whether the benefits of data centres justify the resource demands. The draft expectations for data centres published by the federal government, along with an AI infrastructure consultation that closed on October 9, signal that the era of building first and answering questions later is over.

What 8,200 Blackwell Ultra GPUs actually mean

Sharon AI is an Australian neocloud that positions itself as a provider of sovereign AI infrastructure. It went public earlier this year and currently runs capacity through Equinix in Sydney and NextDC in Melbourne. The company has said it has around 212 megawatts of capacity online or in development, with some 64,000 NVIDIA GPUs slated for deployment by mid-2027. It has also acquired a former paper mill in New South Wales for conversion and has a large project planned in Texas in the United States.

The SYD1 deal anchors Sharon AI’s Australian footprint with serious hardware. The NVIDIA HGX B300 platform, built on Blackwell Ultra GPUs, sits near the top of NVIDIA’s current lineup for AI training and inference. Eight thousand of them in one Western Sydney campus makes SYD1 one of the densest concentrations of AI compute in the country. Initial operations at the site are targeted for the fourth quarter of 2026.

For GreenSquare, landing a long-term anchor tenant of this size is a commercial milestone. Chief executive Charles Penny called it validation of the company’s platform: “Securing this customer commitment is an important milestone for GreenSquare. It validates our platform and the demand for AI-ready infrastructure that can be delivered with speed, density, and operational discipline. Our focus now is on execution, bringing next-generation capacity online safely, reliably, and on schedule.”

Stage two and the NSW connection

This is only the first stage. Planning for a second stage with roughly 80 megawatts of capacity is already underway, and it has the direct backing of the New South Wales government’s Investment Delivery Authority, which is working with developers to streamline regulatory approvals.

The SYD1 story sits inside a much bigger national picture. NVIDIA announced plans to help more than double Australia’s current AI data centre capacity through partnerships with firms including Firmus, CDC, NEXTDC and AirTrunk, targeting up to two gigawatts of AI-related capacity by 2027. Australia’s existing computing capacity stands at about 1.6 gigawatts, according to research cited by industry lobby group Data Centres Australia.

Why this matters beyond Sydney

The GreenSquare project captures several trends at once: the rise of specialist AI neoclouds competing with hyperscalers, the growing acceptance of adaptive reuse as a faster and lower-carbon path to new capacity, and Australia’s deliberate positioning as a regional hub for AI compute, with its land, power access and proximity to Asia-Pacific markets.

There are risks, too. The sector’s recent turbulence, including the withdrawal of Firmus Technologies’ planned ASX listing earlier this month amid investor concerns about debt and valuation, is a reminder that capital for these projects is not unlimited.

But SYD1 Stage 1 is now real: 15 megawatts online, a long-term tenant signed, and thousands of cutting-edge GPUs arriving. For an industry that trades in announcements, that counts for something.

Frequently Asked Questions

What is the GreenSquare SYD1 data centre?

SYD1 is GreenSquare DC’s data centre campus in Norwest Business Park, Western Sydney. Its first stage, 15 megawatts of AI-ready capacity, is now operational inside a refurbished 1995 IBM data centre at 3 Brookhollow Avenue, fully leased to Australian AI neocloud Sharon AI.

How many GPUs will Sharon AI deploy at SYD1?

Sharon AI is expected to deploy approximately 8,200 NVIDIA Blackwell Ultra GPUs (on the HGX B300 platform) at the site, with operations targeted for Q4 2026. The company reports around 64,000 NVIDIA GPUs due for deployment across its portfolio by mid-2027.

Why refurbish an old data centre instead of building a new one?

GreenSquare says adaptive reuse was faster and carried significantly lower embodied carbon than greenfield construction. The eight-month refurbishment increased IT capacity by 275 percent within the original footprint, while new cooling systems sharply reduced water use. The approach shows how legacy sites can be repurposed for high-density AI workloads.

What comes next for the SYD1 campus?

A second stage of around 80 megawatts is already being planned, supported by the New South Wales Investment Delivery Authority to accelerate approvals. GreenSquare is also developing a broader pipeline of more than a gigawatt across New South Wales and Victoria, including a large-scale Melbourne development.

Post Comment

You May Have Missed