NVIDIA and Sharon AI to Build 72MW AI Factory Infrastructure in Australia
A six-year collaboration deploying 40,000 Blackwell GPUs under a new revenue-sharing model to accelerate sovereign AI capacity.
NVIDIA has entered into a strategic six-year compute collaboration with Sharon AI to significantly expand artificial intelligence infrastructure across Australia. The partnership marks a critical step in the deployment of large-scale "AI factories" designed to support the next generation of agentic and sovereign AI.
Under the terms of the agreement, the two companies will deploy infrastructure based on NVIDIA's DSX (Data Center Services) design, adding 72 megawatts of new data center capacity to the region. This scaling effort will include the deployment of up to 40,000 Grace Blackwell GB300 GPUs, providing the raw compute power necessary for high-intensity AI training and inference workloads.
The Shift to Infrastructure Architecture
This expansion reflects a broader transition for NVIDIA as the company evolves from a primary chip supplier into a comprehensive infrastructure architect. Through the DSX platform, NVIDIA now provides partners with detailed reference designs and digital twins, allowing operators to build facilities optimized for the lowest possible cost per token.
This Australian initiative is part of a global pattern of rapid infrastructure scaling. NVIDIA is implementing similar large-scale projects in other territories, including a 360-megawatt campus being developed by Firmus Technologies in Batam, Indonesia. By standardizing the blueprint for AI factories, NVIDIA is ensuring that its hardware and software stack becomes the default foundation for national AI capabilities.
A New Economic Model for Compute
Beyond the hardware, the Sharon AI deal highlights a fundamental shift in NVIDIA's business strategy. The company is moving toward a revenue-sharing and credit-support model with cloud operators. By lowering the initial financial barriers to entry, NVIDIA can accelerate the global rollout of its latest hardware, such as the GB300, while creating recurring revenue streams.
This model is particularly significant for the industry as it addresses the compute gap facing many AI startups and regional operators. By sharing the financial risk and reward of the infrastructure, NVIDIA cements its ecosystem as the industry standard and ensures that its most advanced GPUs are deployed at scale faster than traditional procurement cycles would allow.
Future Outlook
As these AI factories come online, the focus will shift to how Australia leverages this 72MW capacity to develop sovereign AI models. While the hardware deployment is confirmed, the industry will be watching to see how the revenue-sharing model affects the pricing of compute for end-users and whether this blueprint will be replicated across other mid-sized markets to challenge the dominance of the largest hyperscalers.