Arvica Cloud wants to open a liquid-cooled modular data center in Swanbank, Queensland, in 2028. The Australian AI cloud provider has not yet chosen a site or secured a grid connection. That gap sits at the center of every claim the company has made so far.

The facility will be compact and built for next-generation GPU hardware with liquid cooling, as SMBTech reported. Arvica plans to deploy coolant distribution units (CDUs), the devices that circulate coolant between facility plumbing and server cold plates, alongside supplementary air cooling. The target is an annual power usage effectiveness (PUE) of 1.3 or lower. PUE measures total facility energy divided by IT energy; 1.3 would place the site near the better end of current hyperscale practice.

What hardware is the design built around?

Construction will follow a modular approach. Each module will support racks drawing between 150 and 250kW. Arvica says the design draws on the requirements of Nvidia's Blackwell and Rubin systems. That is a design assumption, not a deployed result. No rack hardware exists on site yet, and no vendor contract has been announced.

The initial deployment targets a 1.5MVA grid connection and between 600 and 800kW of IT capacity. A second stage could raise IT capacity to 2.2MW. Arvica did not state a timeline for the second stage, a cost estimate, or a named power distributor for the connection. For comparison, 2.2MW of IT capacity would serve roughly a dozen large GPU clusters, a fraction of what hyperscale campuses deliver.

What does the company claim about the market?

"Competition in AI infrastructure is shifting from GPU counts toward energy efficiency and upgradeability," the company stated. "We want to use more compact, more efficient facilities to support the deployment of next-generation computing platforms in Australia."

Both statements are claims about direction, not verifiable metrics. Arvica did not publish projected energy efficiency figures beyond the 1.3 PUE target, and it did not define what upgradeability means in engineering terms.

Who is Arvica Cloud?

Public information about the company is sparse. Arvica offers GPU cloud compute capacity in Australia and APAC, including Nvidia H200, B200, and B300 units. It has not disclosed revenue, headcount, funding, or existing data center assets. The Swanbank project would be, as far as available records show, its first purpose-built facility.

The 2028 operations date is aspirational. Site selection, grid connection approval, and construction all remain ahead. Queensland grid connection requests for multi-megawatt loads routinely take years, and Arvica has not stated where its application stands.

Until a site and a signed grid agreement appear, the 1.3 PUE target, the 250kW rack specification, and the 2028 date remain commitments on paper. Each number is plausible for a modern liquid-cooled build. None is yet backed by steel, coolant, or a utility contract.