Nvidia Supports 2GW AI Data Center Buildout in Australia

A 2GW signal for Australia’s AI infrastructure race

Nvidia has put Australia more clearly on the global AI infrastructure map, announcing that it is working with a group of Australian cloud, data center and infrastructure partners on projects that could support up to 2 gigawatts of AI-related capacity by 2027 [1]. The company’s framing is important: Nvidia is not describing a conventional single data center project, but a distributed buildout of “AI factories” designed to host multiple generations of NVIDIA DSX infrastructure across an ecosystem of local operators [1].

The named participants are Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk, a mix of AI cloud, neocloud, connectivity and hyperscale infrastructure players [1]. Under the model set out by Nvidia, those participating providers will operate the AI factories, while Nvidia supplies the DSX platform, accelerated computing, networking, software, reference designs and broader ecosystem support [1]. ChannelNews reported the same structure, noting that the partners are expected to provide the land, power, facilities and operations required to host Nvidia’s infrastructure [2].

That distinction matters. The announcement is a major endorsement of Australia as a site for AI compute, but it is also a partnership architecture rather than a disclosure of a single capital budget, construction schedule or fully commissioned power block. ChannelNews noted that the companies have not detailed the total investment required or exactly how much of the proposed 2GW will be operational by the end of 2027 [2]. Reuters, in a short dispatch carried by AOL, similarly reported that the projects “could support” up to 2GW by 2027, without attaching financial terms to the arrangement [3].

What Nvidia is actually backing

Nvidia’s announcement says the work is focused on expanding “land, power and shell capacity” capable of hosting DSX AI factories [1]. In data center terms, that phrase points to the hard constraints that increasingly define AI infrastructure: not just access to chips, but powered sites, cooling-ready buildings, network connectivity, and the operational expertise needed to run dense accelerated-computing clusters.

The DSX platform is being positioned as the common technical layer. Nvidia describes DSX as a full-stack AI factory platform spanning facilities infrastructure, compute, networking, software and reference designs [1]. It is intended to be compatible with the CUDA ecosystem and improved through software over the life of the infrastructure, which Nvidia says can help operators serve different workloads and customers while improving performance, utilization and asset life [1].

In practical terms, that means Nvidia is offering a blueprint for AI data centers that partners can repeat across different sites. The partners bring real estate, power development, cooling systems, connectivity, hosting operations and customer channels. Nvidia brings the compute architecture and software stack that make the facilities useful for training, inference and AI application deployment.

This is why the announcement uses the phrase “AI factories.” Nvidia’s Raj Mirpuri described AI factories as infrastructure that turns energy into intelligence, and argued that DSX can make those assets more productive, durable and investable [1]. The language is promotional, but it captures the strategic shift: AI data centers are being treated not merely as server warehouses, but as industrial production sites where electricity, chips and software are combined to produce model training, inference and AI services.

The partner map: from power campuses to cloud access

Each named partner appears to fill a different role in the emerging Australian AI stack. Firmus is expanding Project Southgate, its Australian AI infrastructure initiative, with Nvidia-powered AI factories designed to support hyperscale and AI-native demand across Australia and Asia-Pacific [1]. In a separate 8 September update, Firmus said its broader portfolio now spans seven AI factories across Australia, Singapore, Indonesia and Malaysia, with two operational sites and five under development targeting ready-for-service over the next 24 months [4]. That regional context helps explain why Nvidia’s Australia announcement is not only about domestic demand, but also about Australia’s possible role in Asia-Pacific compute supply.

Sharon AI is described by Nvidia as deploying up to 68,000 Nvidia GPUs connected through Quantum InfiniBand and Spectrum-X Ethernet networking, using the DSX platform to support startups, enterprises and research organizations [1]. ChannelNews also highlighted the 68,000-GPU figure as one of the most concrete capacity indicators in the announcement [2].

IREN’s role is tied to large-scale development capability. Nvidia says IREN is combining DSX architecture with its expertise in power, land, data centers, GPU deployment and infrastructure operations [1]. The company is applying that blueprint across its development portfolio, including an 800MW Bundey campus in South Australia [1]. For a 2GW national target, a single 800MW development is not incidental; it would represent a large share of the potential capacity if built and energized on schedule.

Megaport’s contribution is different. Nvidia says the Australian-founded connectivity company is expanding access to Nvidia accelerated computing through its Latitude.sh subsidiary, while using Megaport’s global software-defined network to bring AI processing closer to users and reduce latency [1]. That highlights an often-overlooked point: large AI capacity is valuable only if customers can reach it with reliable, low-latency connectivity and integrate it into existing cloud and enterprise workflows.

ResetData is targeting enterprises, government and research users that require high-performance compute and Australian data residency [1]. That sovereign-compute angle is central to the Australian narrative. If universities, public-sector agencies, healthcare companies and regulated industries want to use advanced AI while keeping sensitive data onshore, domestic AI factories become more than export infrastructure; they become national capability.

CDC, NEXTDC and AirTrunk represent the heavy physical-infrastructure side. Nvidia says CDC has AI-ready data center infrastructure across Australia and New Zealand, including direct liquid-to-chip cooling for high-density workloads [1]. CDC’s chief executive Greg Boorer said the company operates more than 550MW of capacity across Australia and New Zealand, with a further 800MW under construction, and that its facilities use 100% renewable electricity and zero-water advanced cooling technologies certified for Nvidia accelerated computing [1]. NEXTDC is described as developing liquid-cooled, high-density infrastructure, including DSX-based AI factories interconnected with major cloud platforms [1]. AirTrunk is expanding AI-ready powered shells with direct-to-chip liquid cooling for high-density AI workloads [1].

Why 2GW is a big number

Two gigawatts is a striking figure in the data center sector because it refers to power capacity at industrial scale. AAP, in a report carried by Brunei’s Brudirect, said Nvidia’s buildout could reach 2GW by 2027 and described that amount as enough to power around 1.5 million homes [5]. Such comparisons are imperfect because homes and AI data centers have very different load profiles, but the analogy underlines the scale of grid demand that AI infrastructure now implies.

For Australia, the opportunity is clear. Nvidia argues that expanded domestic capacity will allow local innovators to develop models, applications and agents while supporting additional power-generation projects for regional and global AI compute demand [1]. The company also points to Australia’s sustainable energy base, digital infrastructure, technology talent and reputation as a trusted economic partner as reasons the country can host AI infrastructure for startups, universities, researchers, enterprises, developers and AI-native companies [1].

The challenge is equally clear: power, cooling and timing. A 2GW target by 2027 requires not only chips and racks, but grid connections, substations, transmission capacity, construction delivery, permitting and energy procurement. Nvidia’s release emphasizes DSX, liquid cooling and renewable power claims from partners, but the announcement does not by itself resolve the practical bottlenecks that determine when megawatts become operating AI compute.

From data centers to local AI applications

Nvidia is also tying the buildout to software and model access. The company says expanded infrastructure will give Australian users greater access to accelerated computing and Nvidia Nemotron open models for locally relevant models, applications and agents [1]. It named Heidi and Atlassian as Australian organizations using Nemotron models, datasets and tools for healthcare AI and enterprise search, respectively [1].

That matters because the political and economic case for AI data centers cannot rest only on exporting compute cycles. The larger argument is that local infrastructure should help Australian companies and researchers build domestic AI products, preserve data residency, and support sectors such as healthcare, education, government, finance, mining, agriculture and climate science. If the 2GW buildout mostly serves offshore hyperscalers, the local benefits will be narrower. If it lowers access barriers for Australian startups, universities and public agencies, the infrastructure could become a more durable national asset.

Firmus’s 8 September announcement points in that direction by saying it intends to establish an Australian AI Access Program for eligible researchers and organizations applying AI to areas of national and regional importance, including science, education, agriculture, energy and resource efficiency, and climate resilience [4]. That proposal is not the same as a completed access scheme, but it shows how partners are trying to link large-scale AI factories with public-interest use cases.

What remains unanswered

The headline number is up to 2GW by 2027, but the current record leaves several open questions. First, there is no disclosed total investment figure across the full partner group [2]. Second, Nvidia and the partners have not specified how much of the 2GW is contracted, financed, under construction, grid-connected or already operational [2]. Third, the capacity appears to be spread across multiple companies and sites, which means delivery risk will vary by project.

There is also a terminology issue. “Capacity” can refer to different things: land and shell readiness, power availability, data center critical IT load, contracted customer capacity, or deployed GPU capacity. Nvidia’s language points to land, power and shell capacity designed to host DSX infrastructure, while partner disclosures include a mixture of megawatts, GPUs and AI factory portfolios [1]. For investors, policymakers and customers, the next proof points will be signed power agreements, construction milestones, equipment deployments, customer contracts and commissioning dates.

Still, the direction is unmistakable. Nvidia is helping organize an Australian AI infrastructure ecosystem around its DSX architecture, with eight partners expected to operate the facilities and deliver the physical footprint [1]. If the group can convert the announcement into energized, liquid-cooled, networked and customer-ready capacity, Australia could move from being a consumer of imported AI compute to a regional producer of it.

For now, the story is best understood as a large-scale infrastructure commitment with a firm strategic signal and incomplete execution detail. Nvidia has put its platform and ecosystem behind a 2GW Australian AI data center buildout; the next test is whether the partners can turn that ambition into operating AI factories by 2027.

Developments

1. NVIDIA Supports 2GW AI Data Center Expansion in AustraliaChannelNews.com.au · Sep 10, 2026, 1:50 AM UTC · 9/10

Sources from the last 72 hours

1. [1]NVIDIA Expands AI Infrastructure Capacity in Partnership With Australia’s Data Center EcosystemSep 9, 2026, 12:00 AM UTC
2. [2]Nvidia Backs Massive 2GW Australian AI Data Centre BuildoutSep 10, 2026, 12:00 AM UTC
3. [3]Nvidia teams up with Australian partners to build AI factory capacitySep 10, 2026, 12:29 AM UTC
4. [4]Firmus surpasses 900 MW contracted capacity, adds OpenAI as anchor customer and expands into MalaysiaSep 8, 2026, 12:00 AM UTC
5. [5]AI powerhouse NVIDIA partners with eight Aussie firmsSep 10, 2026, 12:00 AM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.