Sharon AI and Nvidia sign $4.88B GPU deal for 72MW AI factory in Australia

This is one of the largest single compute commitments we’ve seen outside the US hyperscalers, and it materially reshapes the GPU landscape in the Southern Hemisphere. A 72MW AI factory built around up to 40,000 Grace Blackwell GB300 GPUs effectively turns Australia into a regional hub for frontier‑scale training and inference. The revenue‑sharing model also marks a shift in how Nvidia participates in downstream cloud economics, moving from pure hardware vendor toward platform partner with recurring exposure to AI workload growth.

Strategically, Sharon AI is using Nvidia’s DSX AI factory template and financing structure to vault from niche AI cloud provider into a sovereign‑scale infrastructure player. That matters because compute access has been the main bottleneck for emerging model labs, startups, and even universities outside the US–China duopoly. If Sharon executes, a lot more APAC experimentation can happen without routing everything through US hyperscalers. For Nvidia, this deal both locks in Blackwell demand and signals that it’s willing to co‑underwrite data center build‑outs when a partner can aggregate sufficient regional demand.

For the broader race to AGI, this is another proof point that capital and GPUs are rapidly globalizing. High‑end clusters are no longer confined to Silicon Valley and a handful of Chinese cloud giants; they’re popping up wherever power, regulation, and capital line up. That diffusion increases the number of serious AGI contenders and accelerates the pace at which non‑US ecosystems can run large‑scale experiments.