Rise of open-weight AI models expected to accelerate demand for sovereign enterprise AI infrastructureSCX.ai preparing to become Australia’s first ASX-listed pure-play sovereign AI inference infrastructure companyThe company says Australia can expand AI capability by using existing data centres more efficiently rather than building ever-larger AI campuses
Special Report**: Nvidia chief executive Jensen Huang used his first-ever post on X to launch a campaign backing “open-weight” AI models and within 24 hours support had doubled from 25 companies to 50, with OpenAI, Microsoft, Google, AMD, Cisco and dozens of the world’s largest technology companies joining the initiative.**
The campaign reflects what many believe is the next major shift in artificial intelligence.
Rather than relying solely on closed AI models hosted by overseas providers, organisations are increasingly seeking the freedom to download, customise and run AI on infrastructure they control.
Australian AI infrastructure company SCX.ai believes that shift will redefine where value is created across the AI industry.
The company is seeking to raise $40 million through a fully underwritten IPO ahead of its proposed ASX listing under the code SCX, positioning itself as Australia’s first listed pure-play sovereign AI inference infrastructure company.
Founder and chief executive David Keane said open-weight AI was changing the economics of artificial intelligence.
“Businesses are going to have more choice over the AI models they use than ever before,” Keane said.
“But regardless of whether they choose an open-weight model or a proprietary one, they still need somewhere secure, efficient and sovereign to run it. We believe that’s where the next phase of AI infrastructure will be built.”
As AI becomes embedded in everyday business processes, organisations are increasingly asking where their AI runs, where their data resides and how much control they retain over both.
That is the market SCX.ai is targeting.
Inference infrastructure
Rather than developing its own foundation models, the company provides AI inference infrastructure – the computing that powers open-weight models and the everyday AI applications that use them; including chatbots, AI agents, document analysis and enterprise automation.
Inference is the part of artificial intelligence most organisations actually use. Every conversation with an AI assistant, every document summarised, every customer service chatbot and every AI agent relies on inference rather than model training.
SCX.ai’s platform allows those applications to run inside existing Australian commercial data centres using specialised AI processors designed specifically for inference workloads.
The strategy differs markedly from the AI infrastructure narrative dominating global headlines, which has focused on building ever-larger AI factories requiring enormous amounts of power, water and land.
Instead, SCX.ai believes Australia can unlock substantially more AI capability by making better use of the data centres it already has.
“Australia absolutely needs more AI infrastructure,” Keane said.
“But we also have an opportunity to build it more intelligently. If we can deliver significantly more AI from existing commercial data centres using infrastructure designed specifically for inference, we reduce pressure on energy, water and new construction while still meeting rapidly growing demand.”
The focus on efficiency isn’t being driven solely by environmental concerns. Businesses are also discovering that AI can become very expensive very quickly.
Recent analysis by fund manager Ophir estimates Australia’s data centres currently consume about 1.3 billion litres of water annually. If all proposed developments proceed, that figure could increase to between 4.4bn and 5.9bn litres each year by 2031. At the midpoint of those forecasts, data centres would consume more water than power station cooling and golf courses.
At the same time, enterprises are grappling with soaring AI operating costs as token consumption accelerates.
The cost of running AI has become one of the biggest issues facing enterprise customers. Uber revealed earlier this year it exhausted its annual AI budget in just three months, while Westpac has introduced token usage tracking as it looks to manage the growing cost of AI across the organisation.
Interest in open-weight models
Those pressures are driving growing interest in open-weight models, which organisations can run on their own infrastructure at a fraction of the cost of some frontier proprietary models.
It’s also the case that not every task needs frontier level intelligence. Extracting payment details from a document does not need a Fable level intelligence to be done correctly. A small and efficient open-weight model operating at a fraction of the cost can do the task faster and just as accurately.
“Not every AI task requires the most powerful model available,” Keane said.
“If you’re extracting information from a document or running an internal AI assistant, a smaller open-weight model can often deliver the same result faster and at a much lower cost. That’s where enterprises are increasingly focusing.”
The economics are becoming even more compelling as open-weight models rapidly improve.
New models such as Kimi K3 are approaching the performance of leading proprietary systems on many enterprise workloads, narrowing the gap between open and closed models while maintaining a significant cost advantage. Efficiency is also becoming a central investment theme globally.
Last month, SambaNova Systems – whose technology underpins SCX.ai’s platform – raised US$1bn, lifting its valuation to US$11bn. The company develops Reconfigurable Dataflow Units (RDUs), purpose-built AI processors designed specifically for inference.
Unlike conventional GPUs originally developed for graphics processing, RDUs are engineered solely for AI inference. They can deliver substantially higher AI throughput while consuming around 7% less energy, providing responses almost five times faster and using about 99% less cooling water because they eliminate the need for liquid cooling.
“AI chips are evolving at extraordinary speed, so the infrastructure choices companies make now will shape their costs and capabilities for years,” Keane said.
This article was developed in collaboration with SCX.ai, a Stockhead advertiser at the time of publishing.
This article does not constitute financial product advice. You should consider obtaining independent advice before making any financial decisions.
Related Topics #
UNLOCK INSIGHTS
Discover the untold stories of emerging ASX stocks.
Daily news and expert analysis, it's free to subscribe.
By proceeding, you confirm you understand that we handle personal information in accordance with our Privacy Policy.