In a decisive move to secure the Southern Hemisphere’s place in the global artificial intelligence arms race, Australian cloud provider Sharon AI has finalized a landmark six-year infrastructure agreement with Nvidia. The deal centers on the deployment of Nvidia’s DSX AI Factory architecture, a sophisticated blueprint designed to transform traditional data centers into high-performance processing hubs specifically for generative AI. By integrating 72 megawatts of new capacity, Sharon AI is positioning itself as a critical intermediary for regional startups and research institutions starved for compute power.
At the heart of this expansion is the Grace Blackwell GB300 GPU, Nvidia’s latest flagship silicon. Sharon AI plans to scale its fleet to an staggering 40,000 Blackwell units, with the total number of deployed GPUs expected to exceed 55,000 by mid-2027. This isn't merely a hardware acquisition; it represents a fundamental shift toward the 'AI Factory' model, where the infrastructure is tailored for the massive parallel processing requirements of large language models rather than general-purpose cloud storage.
The commercial appetite for such localized compute is already evident. Of Sharon AI’s projected 132MW total capacity, over 100MW has already been secured by end-user contracts before the infrastructure is even fully operational. This high utilization rate underscores a global trend where enterprise clients are bypassing traditional hyperscalers in favor of specialized AI cloud providers who can offer immediate access to the latest generation of Nvidia hardware.
For Australia, the deal serves as a significant boost to its domestic tech ecosystem, which has often struggled to compete with the sheer scale of Silicon Valley or Beijing. By hosting one of the world's largest deployments of Blackwell-class GPUs, the country is essentially building a 'sovereign AI' capability. This infrastructure will allow Australian researchers and businesses to train sensitive models locally, mitigating data residency concerns while benefiting from the latency advantages of domestic hosting.