Nvidia Partners with 8 Australian Data Centres to Build 2GW AI Infrastructure by 2027

Nvidia Targets Massive AI Infrastructure Expansion in Australia

Nvidia has signed partnerships with eight Australian data centre and cloud companies to build what the chipmaker calls AI factory capacity, targeting up to 1

1

2 gigawatts of AI capacity by 2027. The partners include Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NextDC and AirTrunk, marking a significant push to expand AI infrastructure capacity across the nation2

2

. Australia's current computing capacity stands at 3

3

1.6GW, meaning this buildout could more than double the country's AI compute capabilities.

The expansion addresses surging AI compute demand from technology giants including Amazon, Meta and Microsoft4

4

. However, Nvidia acknowledged that its chips will use more energy in Australia next year than all existing data centres currently require, highlighting the significant energy strain this expansion will create.

Understanding the AI Factory Model and Nvidia DSX Platform

An AI factory, in Nvidia's definition, represents a full-stack arrangement rather than just physical buildings. Built on the Nvidia DSX platform, these facilities integrate compute infrastructure, networking, software and reference designs around the CUDA ecosystem1

1

. The DSX platform is compatible with the CUDA ecosystem and enhanced through software over the infrastructure's lifetime, improving performance and extending asset life.

Raj Mirpuri, vice president of global AI clouds and infrastructure ecosystem at Nvidia, explained that "AI factories turn energy into intelligence—the essential resource of the AI economy"2

2

. The platform aims to make AI factories more productive, fungible and durable, creating what Nvidia positions as a new investable asset class.

Sharon AI is deploying up to 68,000 NVIDIA GPUs connected with NVIDIA Quantum InfiniBand and NVIDIA Spectrum-X Ethernet networking using the DSX platform2

2

. This infrastructure will support startups, enterprises and research organizations seeking high-performance accelerated computing.

The Sovereign AI Pitch and Its Limitations

Nvidia frames this expansion as enabling sovereign AI for Australia, emphasizing data residency rules, locally trained regional AI models, and industries preferring not to send data offshore1

1

. The expanded capacity will give Australian innovators greater access to accelerated computing and NVIDIA Nemotron open models to develop locally relevant applications and agents.

James Manning, cofounder and CEO of Sharon AI, stated that "Australia has the energy, connectivity and ambition to lead in AI"2

2

. Yet the sovereignty claim carries an inherent contradiction: while the racks sit on Australian soil, the accelerators, interconnect, software stack and reference architecture remain entirely American1

1

.

This arrangement represents a meaningful improvement over shipping data to Virginia, but falls short of true independence. The distinction matters increasingly in Canberra, particularly as Washington pressures Australia over platform regulation1

1

.

Scale of Partner Commitments and Infrastructure

IREN's Bundey campus in South Australia operates as an 800MW site1

1

. CDC operates more than 550MW across Australia and New Zealand with a further 800MW under construction. Firmus is expanding its Project Southgate initiative, developing NVIDIA-powered AI factories designed to support hyperscale and AI-native demand across Australia and Asia-Pacific2

2

.

Oliver Curtis, cofounder and co-CEO of Firmus, described Project Southgate as "a tangible commitment to building Australia's AI future"2

2

. The initiative aims to give Australian businesses, researchers and innovators access to world-class computing capacity domestically while positioning Australia to meet growing demand across Asia-Pacific.

Notably, six of the eight partners are landlords or connectivity providers rather than model builders1

1

. This reveals the market reality: capacity is being committed well ahead of demand that would justify it, on the assumption demand will materialize.

The Power Question and Missing Details

2 gigawatts represents roughly the output of two large nuclear reactors, committed in a country whose grid operator has warned about firm capacity in eastern states for several years1

1

. Nvidia does not build power stations, and the announcement makes no claim about where the electricity comes from—the standard gap in these arrangements.

The announcement from Nvidia5

5

provided no financial terms for the partnership. Critical details remain undisclosed: no timetable per site, no capital commitment figure, no specific customers named, and no word about power generation1

1

. These four factors determine whether 2 gigawatts represents a concrete plan or merely a headline.

This approach contrasts sharply with Europe's AI gigafactory competition, which runs through state aid, consortium bids and site selection, attempting to answer ownership and power supply questions before construction begins1

1

. Europe's energy regulators actively debate whether data centres should bring their own clean supply. The Australian model, assembled from eight private operators without public subsidy, moves faster but defers these fundamental questions.

Watch for announcements on power generation partnerships, specific customer commitments, and how Australia's grid infrastructure adapts to this unprecedented AI computing capacity expansion.