SharonAI’s AI compute deal with NVIDIA refocuses attention on its valuation, investors eye new Australian capacity

SharonAI Holdings’ latest push into AI infrastructure is drawing fresh attention from investors, after the company described a six-year collaboration with NVIDIA centered on adding new AI compute capacity in Australia. The plan, as outlined in recent market coverage, targets 72MW of additional data center capacity and is built around NVIDIA’s Grace Blackwell GB300 GPUs, a next-generation graphics processing unit design aimed at accelerating large-scale AI workloads.

The collaboration framing matters because it ties SharonAI’s capital deployment and future revenue opportunities to a specific compute stack, rather than a general move into “AI data centers.” By stating the GPU platform up front, the company is indicating that its infrastructure roadmap is meant to align with NVIDIA’s AI software and hardware ecosystem, which many enterprise and cloud customers rely on to run modern machine learning models.

NVIDIA’s role in the announcement also helps explain why the deal is being treated as more than a branding exercise. In practice, AI compute partnerships are often evaluated by markets on the strength and durability of customer demand and the quality of the compute platform being deployed, including whether it matches current performance and software needs for AI training and inference.

Even so, details that investors typically want are not fully spelled out in the market coverage that has circulated so far. The reporting highlights the headline elements of the agreement and the capacity figure, but it does not provide, in the available excerpt, specifics such as total project cost, expected ramp schedule for the 72MW buildout, named customers beyond the partnership structure, or the expected timeline for revenue recognition from the collaboration.

The new attention has also spilled into valuation discussions. The same coverage points to a “fresh look” at SharonAI’s trading valuation, citing an approximately 11.3-times price-to-book (P/B) level. Price-to-book is a market metric that compares a company’s market capitalization to its accounting book value, and it is often used as a rough check on whether investors are paying a premium or discount relative to the company’s net assets.

For SharonAI, a higher P/B typically implies investors expect the market value of its assets, contracts, or future cash flows to exceed what is currently captured in book value. That view can be supported if the company’s infrastructure is expected to generate scalable, recurring earnings, particularly if capacity additions translate into contracted demand for AI compute services.

The episode sits within a broader Technology and data center theme that has dominated equity markets over the last year, as firms weigh how quickly AI-related power, cooling, and GPU capacity can be scaled. Markets have increasingly focused on practical constraints, including site readiness, power availability, and whether planned capacity actually comes online and is utilized at attractive margins.

What remains unclear is how quickly the 72MW is expected to come online, what proportion of the capacity will be contracted under the NVIDIA-linked arrangement, and whether the collaboration includes specific financial terms. Investors will likely look for more disclosure, such as project milestones, utilization assumptions, and any concrete customer commitments tied to the compute buildout, before drawing conclusions about durability of earnings power from the deal.