DXN lands AI data centre beachhead with $8.8m US contract

DXN has put a meaningful marker down in the overheated, and highly capital-hungry, artificial intelligence infrastructure market, signing a binding $8.8 million contract with a US-based neo-cloud operator for a 1.36 megawatt AI high performance computing modular data centre. For a company with an ASX-quoted market capitalisation of about $6.28 million, the contract is not pocket change - it is larger than the company’s entire equity value.

The deal covers design, engineering, manufacturing, commissioning and delivery of a turnkey modular data centre, with the customer described as a US-listed global neo-cloud operator. The customer has not been named, but DXN says the disclosure is sufficient for investors to assess the materiality of the deal.

The proof-of-concept that could matter most

The initial contract is structured as a pilot proof-of-concept deployment. That phrase can sound modest, but the investor interest lies in what may follow. DXN says that subject to successful delivery, the customer has indicated an intention to progress to a much larger campus-scale AI compute program, representing a potential revenue opportunity of more than US$200 million over the next one to two years for the existing site.

That is not bankable revenue yet, and investors should treat it accordingly. But it does provide a glimpse of the size of the prize if DXN can execute cleanly on the pilot. In small-cap land, the difference between a promising contract and a company-making platform often comes down to boring things: delivery deadlines, quality control, working capital, customer acceptance and repeatability. The sizzle is AI; the steak is manufacturing and commissioning without a mishap.

Built in Welshpool, bound for the US

Manufacturing is due to start immediately at DXN’s Welshpool facility in Western Australia, with on-site commissioning expected at the customer’s US mainland facility within about six months of signing. The modular system will incorporate integrated power, direct-to-chip liquid cooling, fire suppression and building management systems. It is designed to support GPU rack densities of up to 150kW per rack, consistent with DXN’s standard AI HPC module range.

That last detail is important. AI data centres are not merely conventional data halls with a few more graphics processors dropped in like Christmas baubles. High-density AI workloads bring fierce power and cooling requirements, and direct liquid cooling has shifted from specialist curiosity to practical necessity in many advanced deployments.

DXN is pitching its advantage around speed-to-deployment, modular scalability and factory-tested infrastructure. In a market where traditional data centre builds can take years, DXN argues its prefabricated AI HPC modules can be deployed in roughly six to eight months from contract signing.

Management calls it a defining milestone

Managing director Shalini Lagrutta described the contract as “a defining milestone” and said securing a publicly listed US-based neo-cloud operator validated the company’s strategic direction over the past three years. She said DXN had been productising its modular data centre capability specifically for the high-density AI inference market.

Lagrutta also pointed to the customer selection as a reflection of DXN’s differentiation in rapid deployment, modular scalability and liquid cooling capability. For investors, the quote matters less as corporate flourish and more as a clue to the company’s intended positioning: not as a generic builder of boxes, but as a specialist supplier of high-density, AI-ready compute infrastructure.

The AI infrastructure tide is running hard

DXN’s timing is handy. The company says the global data centre GPU market is estimated at about US$99 billion in 2025, growing at roughly 14 percent a year, while AI data centre infrastructure is forecast to grow at a compound annual growth rate of 27.5 percent through to 2034. It also says AI inference has overtaken training as the dominant data centre workload and is the fastest-growing segment of GPU infrastructure demand.

The inference angle is worth dwelling on. Training large AI models gets the headlines, but inference is where models are used repeatedly in real-world applications. That can mean relentless demand for distributed, power-hungry compute capacity. Neo-cloud operators, which rent GPU capacity and AI infrastructure to enterprise and developer customers, need speed, density and scalability. DXN’s modular approach is aimed squarely at that pain point.

What investors should watch next

The near-term investment case now turns on execution. First, can DXN deliver the pilot within the indicated six-month commissioning window? Second, can it manage the working capital demands of a contract that is large relative to its market value? Third, does successful delivery convert into the much larger campus opportunity flagged by the customer?

The contract gives DXN credibility in a sector where credibility is hard won and easily lost. It also gives investors a measurable catalyst: delivery of the pilot and any subsequent expansion order. For now, this is a significant commercial win with a tantalising follow-on opportunity, rather than a guaranteed transformation.

Still, for a microcap modular data centre specialist, landing a US neo-cloud customer in the middle of an AI infrastructure boom is about as close as it gets to finding a high-voltage socket in the desert.