NVIDIA is testing a new recurring-revenue model in Australia, deploying 72MW of GPU infrastructure through a usage-linked partnership instead of relying solely on upfront hardware sales.

What happened: NVIDIA has entered a partnership with SharonAI to deploy 72MW of GPU infrastructure in Australia. Rather than selling GPUs outright, the arrangement uses a recurring revenue and usage-linked model, meaning NVIDIA shares in the cloud revenue that SharonAI generates from the capacity over time.

Why it matters: This arrangement diversifies how NVIDIA can extract cash flows from the same hardware and broadens its reach into a geography with historically limited AI compute access. It also shows management experimenting with capital-light ways to seed AI infrastructure for startups and research institutions that may not afford large GPU clusters on their own. The model shifts NVIDIA closer to the economics of infrastructure operators without owning the data centers itself.

What to watch: The partnership relies on a six-year revenue-sharing structure with a single partner in an emerging AI region, which introduces execution and counterparty risk—if uptake or utilization is slower than expected, it could challenge a straight-line growth narrative. The focus on capital-efficient access for startups and sovereign AI capacity in Australia has been only lightly reflected in NVIDIA's existing data center narrative, which centers mainly on large hyperscalers (major cloud providers).