Nvidia has signed eight Australian data centre and cloud companies to build what it calls AI factory capacity, in an arrangement the chipmaker says could support up to two gigawatts of AI compute in Australia by 2027.
The partners named in the announcement are Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NextDC and AirTrunk. IREN’s Bundey campus in South Australia is an 800MW site.
CDC says it operates more than 550MW across Australia and New Zealand with a further 800MW under construction. No capital figure is given for the programme as a whole.
An AI factory, in Nvidia’s usage, is a full-stack arrangement rather than a building: facilities, compute, networking, software and reference designs built around its DSX platform and the CUDA ecosystem.
That definition is doing commercial work. It converts a set of independent Australian landlords into deployments of one company’s architecture, which is the point of the exercise from Santa Clara.
Two gigawatts is the number to sit with. It is roughly the output of two large nuclear reactors, committed in a country whose grid operator has spent several years warning about firm capacity in the eastern states.
Nvidia does not build power stations, and the announcement makes no claim about where the electricity comes from, which is the standard shape of these arrangements and the standard gap in them.
The Australian buildout has been running for a while, and TNW has covered its financing. AirTrunk is at the centre of the American money going into Australian data centres, and NextDC has been issuing convertible notes to fund exactly this class of capital expenditure.
What is new is not that the capacity is being built. It is that Nvidia is now naming the ecosystem as a single programme with a gigawatt target attached.
Sovereign AI is the pitch, and it is a real one for Australia: data residency rules, locally trained models, industries that would rather not send their data offshore.
It is also a slightly strange use of the word. Sovereignty here means the racks are on Australian soil and the accelerators, the interconnect, the software stack and the reference architecture are all American.
That is a meaningful improvement on shipping the data to Virginia. It is not independence, and the distinction matters more in Canberra this month than it did last year, with Washington already leaning on Australia over its platform regulation.
Two gigawatts in one mid-sized economy by 2027, assembled from eight private operators without a public subsidy programme attached, is a different mechanism from the EU’s AI gigafactory competition, which is running through state aid, consortium bids and site selection. One approach is faster.
The other is trying to answer questions about ownership and power supply before the concrete is poured, and Europe’s energy regulators are actively arguing about whether data centres should have to bring their own clean supply.
The quotations from the partners are what you would expect: Oliver Curtis of Firmus on building Australia’s AI future, James Manning of Sharon AI on the country having the energy, connectivity and ambition to lead.
Daniel Roberts of IREN made the more useful observation that building AI infrastructure at scale requires integrating every layer, which is precisely the argument for buying the whole stack from one supplier and precisely the concern.
There is one more thing worth noting about the list. Six of the eight partners are landlords or connectivity providers rather than model builders, and the only customer-side voice in the announcement is a healthcare startup describing what it can do with an Nvidia model.
That is an accurate picture of the market: the capacity is being committed well in advance of the demand that would justify it, on the assumption the demand arrives. It usually has so far.
What the announcement does not contain is a timetable per site, a capital commitment, a customer, or a word about generation. Those are the four things that determine whether two gigawatts is a plan or a headline, and none of them is disclosed.
Get the TNW newsletter
Get the most important tech news in your inbox each week.
Facts Only
* Nvidia signed agreements with eight Australian data centre and cloud companies.
* The goal is to build AI factory capacity, potentially supporting two gigawatts of AI compute by 2027.
* Partners include Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NextDC, and AirTrunk.
* IREN’s Bundey campus in South Australia is an 800MW site.
* CDC operates over 550MW across Australia and New Zealand, with 800MW under construction.
* An AI factory is defined as a full-stack arrangement including facilities, compute, networking, software, and reference designs based on Nvidia’s DSX platform and CUDA ecosystem.
* The capacity target of two gigawatts is comparable to the output of two large nuclear reactors.
* Existing buildout financing has been covered by TNW, and NextDC has issued convertible notes for capital expenditure.
Executive Summary
Full Take
Sentinel — Human
This text demonstrates a high degree of human analytical engagement, weaving factual reporting with deep structural critique and comparative reasoning.
