Highlights

  • SCX.ai has launched a fully underwritten IPO to raise about $40 million and list on the ASX under the proposed ticker SCX; the shares are not yet trading.
  • The offer is planned at 30 cents per share for roughly 133.3 million new shares, with Henslow and Canaccord Genuity named as joint lead managers.
  • SCX.ai builds 'sovereign' AI inference infrastructure, keeping Australian data onshore using energy-efficient ASIC chips from technology partner SambaNova rather than power-hungry GPU training farms.
  • SambaNova, SCX.ai's core hardware partner, recently completed the first close of a US$1 billion Series F at an US$11 billion valuation, led by General Atlantic.
  • Investors should watch for the prospectus, final listing timetable, conversion of contracted revenue and presold capacity, and the pace of the planned national data-centre rollout.

SCX.ai, the Sydney-based operator of what it calls Australia's first "sovereign" artificial-intelligence inference network, has launched an initial public offering to raise about $40 million and list on the Australian Securities Exchange under the proposed ticker SCX. If the float proceeds as planned, it would give public investors a rare pure-play exposure to onshore AI computing just as data sovereignty becomes a boardroom and government priority.

The offer is positioned as an infrastructure story rather than a speculative technology punt. SCX.ai runs the physical and software layers that let Australian organisations run AI models on local soil, using chips from SambaNova, the Californian hardware group just valued at US$11 billion. It is important to be clear from the outset: SCX.ai is an upcoming listing, not a trading stock, and the shares will not change hands until the IPO completes and the company is admitted to the ASX.

What Has Happened

SCX.ai has announced a fully underwritten IPO seeking to raise approximately $40 million. Under the planned raise, the company intends to issue around 133.3 million new ordinary shares at 30 cents each, with Henslow and Canaccord Genuity named as joint lead managers and underwriters. On completion, the company would trade on the ASX under the proposed code SCX.

The raise is designed to fund a national build-out of AI "inference" capacity — the computing used to run trained AI models, as distinct from the enormous GPU clusters used to train them. SCX.ai has already deployed an initial node inside Equinix's SY5 data centre in Sydney and plans to expand nationally using existing commercial data-centre space rather than building new facilities from scratch.

Because this is a pre-listing company, the pricing, share count and timetable are planned terms formalised through the prospectus and can change. No debut share price exists yet; the market value of SCX will only be established once the company lists.

Why It Matters

The pitch behind SCX.ai (proposed ASX:SCX) rests on a simple but increasingly urgent idea: that Australian enterprises, government agencies and regulated industries want to use modern AI without sending sensitive data offshore or depending entirely on foreign hyperscale cloud providers. "Sovereign AI" keeps data, processing and operational control within national borders.

That demand has been building. Privacy regulation, national-security sensitivities and commercial caution have made many organisations wary of routing confidential information through overseas AI services. SCX.ai aims at that gap, offering a full stack — inference hardware, data-centre space, private interconnection and workload-management software — that it says lets customers deploy AI while keeping data resident in Australia. A dedicated sovereign-AI infrastructure listing is also unusual on the ASX, offering a more direct way to back the theme than software or diversified data-centre stocks.

Inside SCX.ai's Sovereign Inference Model

SCX.ai — which also operates under the Southern Cross AI banner — was established after founder, chief executive and managing director David Keane relocated to Sydney in late 2025, and is chaired by a director reported by the surname Stevenson.

The technical heart of the offering is a deliberate bet on inference efficiency. Rather than deploying general-purpose graphics processing units (GPUs) — the chips that dominate AI training and carry heavy power and cooling demands — SCX.ai uses application-specific integrated circuits (ASICs) from SambaNova, purpose-designed for running AI models rather than training them.

The Efficiency Argument

The commercial logic flows from that hardware choice. SCX.ai says the ASIC-based approach can be up to 10 times more energy-efficient and up to three times faster at inference than comparable GPU systems, and that the chips run on air cooling alone — removing the water consumption and costly retrofits water-cooled GPU farms typically require. That lets the company slot capacity into existing data centres such as Equinix SY5 rather than financing brand-new facilities, with customers able to access leading models from Meta, OpenAI and DeepSeek. These performance figures are the company's own claims; independent, at-scale verification is something investors would want confirmed over time.

The SambaNova Connection and the $11 Billion Valuation

Much of the strategic narrative around SCX.ai is tied to its core technology partner. SambaNova Systems, whose hardware underpins SCX.ai's nodes, recently completed the first close of a US$1 billion Series F at an US$11 billion valuation, led by General Atlantic.

That valuation matters for SCX.ai because it signals deep-pocketed institutional conviction in exactly the category SCX.ai is commercialising in Australia — energy-efficient inference silicon as an alternative to the GPU-centric model. SambaNova's backers reportedly include Intel, BlackRock, T. Rowe Price and the Qatar Investment Authority.

For SCX.ai investors, the relationship cuts both ways: it lends credibility and supply confidence, but leaves the company reliant on a single hardware supplier.

Key Financial and Operational Details

SCX.ai has offered some early traction metrics alongside the float. As at the end of May, it reported roughly $5.4 million in contracted annual recurring revenue and about $14.6 million in unbilled total contract value, with close to 300 active users — including trial participants — by late June. It has also said nearly one-third of its initial node capacity was presold.

On the physical footprint, the first operational node sits within Equinix's Sydney facility, and the company has flagged further sites through 2026, including a planned expansion into regional South Australia powered by renewable energy. On the demand side, ASX-listed Decidr AI Industries has selected SCX.ai as a sovereign AI inference infrastructure partner. These figures are early and modest, consistent with a business at the start of a ramp the IPO is intended to accelerate.

Potential Catalysts

Several developments could shape sentiment around the planned listing. The most immediate is confirmation of a firm listing timetable, which would move the offer from announced intention to a completed float; beyond that, conversion of the contracted backlog and presold capacity into billed revenue would be a key proof point. Additional catalysts include new nodes coming online in the national rollout, fresh enterprise or government customer wins that validate the sovereign-AI thesis, and continued momentum at partner SambaNova.

Material Risks and Uncertainties

The risks are substantial and typical of an early-stage, pre-listing infrastructure company. First, the IPO must actually complete; planned terms can change and offers can be withdrawn or repriced. Second, SCX.ai is small, investing ahead of scale, and dependent on executing a capital-intensive rollout on time and on budget.

Reliance on a single hardware partner in SambaNova concentrates supply-chain and technology risk. The company's performance and efficiency claims, while central to its pitch, are largely self-reported and await broad independent validation. Competition is intense and well-funded, spanning global hyperscalers, GPU-based providers and other sovereign-AI entrants, and as a newly listed play SCX would likely trade with high volatility once quoted.

What Investors Should Watch Next

For SCX.ai (proposed ASX:SCX), the near-term checklist is clear. Watch first for the prospectus and a confirmed listing date, which will set the terms and open the offer. Then look for evidence that early commercial traction is real and growing — billed revenue against the reported backlog, utilisation of the Sydney node, and the pace at which presold capacity converts. Beyond that, the national rollout is the story: new nodes coming online, the renewable-powered regional site progressing, and fresh enterprise and government contracts. Developments at SambaNova are also a useful barometer of the wider inference-hardware category.

Conclusion

SCX.ai's planned $40 million float would give ASX investors an uncommon, direct route into the sovereign-AI infrastructure theme, backed by a technology partner in SambaNova that the market has just valued at US$11 billion. The proposition is coherent: energy-efficient inference, kept onshore, aimed at genuine and growing demand for data sovereignty.

Yet SCX.ai (proposed ASX:SCX) remains an early-stage company at the start of a capital-intensive rollout, with modest current revenue, a single core hardware supplier and performance claims still to be proven at scale. The offer is a planned, not-yet-completed listing, and the shares are not trading. For investors weighing it up, the focus is on the prospectus, the listing timetable, and whether early contracted revenue and presold capacity translate into a durable, scaling business once SCX.ai reaches the boards.