NVIDIA and Australian partners plan 2 GW AI factory buildout by 2027
NVIDIA has announced a broad collaboration with eight Australian data centre and cloud infrastructure operators to expand AI compute capacity across the country, with a stated target of up to two gigawatts of buildout by 2027. The initiative involves Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk, all of which will operate NVIDIA DSX-based AI factories while NVIDIA supplies the compute, networking, software and ecosystem support.
The scale of the commitment is notable. Sharon AI alone is deploying up to 68,000 NVIDIA GPUs connected via NVIDIA Quantum InfiniBand and Spectrum-X Ethernet networking. IREN is applying the DSX reference architecture to its 800-megawatt Bundey campus in South Australia. CDC, meanwhile, reports more than 550 megawatts of operational capacity across Australia and New Zealand, with a further 800 megawatts under construction, running on 100% renewable electricity with direct liquid-to-chip cooling.
Raj Mirpuri, vice president of global AI clouds and infrastructure ecosystem at NVIDIA, framed the initiative in economic terms: "AI factories turn energy into intelligence, the essential resource of the AI economy."
The DSX platform and sovereign compute push
NVIDIA DSX is a full-stack platform spanning facilities infrastructure, compute, networking, software and reference designs. The company positions it as compatible with the CUDA ecosystem and continuously upgradeable in software over the life of the physical infrastructure, which it argues extends asset durability and improves operator economics across generations of GPU hardware.
A significant driver cited in the release is data sovereignty. Several partners, including ResetData and Sharon AI, explicitly reference the ability to serve government and enterprise customers requiring Australian data residency, without transatlantic or cross-border data transfer. This framing aligns with a broader regional trend: cloud and AI buyers in the Asia-Pacific region, particularly those in healthcare, financial services and public sector, are under increasing pressure from local regulators and procurement rules to keep sensitive workloads onshore.
Australian organisations using the expanded capacity include healthcare AI startup Heidi, which is combining NVIDIA Nemotron 3 Ultra open models with clinical fine-tuning to build voice-driven diagnostic tools, and Atlassian, which is using accelerated NVIDIA infrastructure to power semantic search within its Rovo enterprise AI product.
Market context and competitive positioning
Australia is emerging as a meaningful node in the global AI infrastructure map, aided by relatively stable grid access, a skilled technology workforce and proximity to Southeast Asian markets. The announcement positions NVIDIA's Australian partner ecosystem as an alternative to hyperscaler-controlled capacity, a pitch that resonates with enterprise and government buyers concerned about vendor lock-in or cross-border data exposure.
The broader GPU cloud market in Asia-Pacific is contested. Hyperscalers including AWS, Microsoft Azure and Google Cloud have all announced Australian region expansions in recent years, and sovereign cloud programmes from domestic operators are competing for the same government workloads. The involvement of AirTrunk, acquired by Blackstone in 2024, signals that large-scale institutional capital is already committed to the region's data centre growth.
NVIDIA's approach of anchoring partner facilities to the DSX architecture also acts as a platform lock-in mechanism: operators who build out to DSX specifications are structurally incentivised to source successive GPU generations from NVIDIA rather than switch to AMD Instinct or Intel Gaudi alternatives.
Forward-looking, the two-gigawatt target by 2027 will require substantial grid and renewable energy commitments from state governments. Australia's national AI policy framework, still developing at the federal level, will likely shape how public-sector procurement flows toward locally operated AI infrastructure over the next 18 months.