SCX.ai Holdings has begun trading on the Australian Securities Exchange as the country’s first listed pure-play AI infrastructure company after completing a AUD$40 million initial public offering. The fully underwritten IPO was anchored by Ellerston Capital, Wilson Asset Management and Frazis Capital Partners, with shares offered at $0.30. Trading commenced at midday on Friday, with the stock opening at $0.26 before falling to $0.20.
A Sovereign AI Inference Model
SCX.ai positions itself as a sovereign AI inference-as-a-service provider that installs purpose-built technology inside existing Australian data centres rather than building new large-scale facilities. The company, chaired by former ANZ executive Wayne Stevenson, uses SambaNova Systems hardware in Sydney to sell AI processing tokens to enterprise and government customers. Its services target organisations that need data residency and control without relying on hyperscale cloud providers.
The company’s model focuses on inference, which occurs when AI models respond to requests, instead of training models or renting out chips. Founder and CEO David Keane said the business simply sells tokens, meaning greater machine throughput translates directly into higher revenue. SCX runs on air-cooled SambaNova chips that consume considerably less electricity and no water for cooling.
Commercial Growth Since the Prospectus
SCX reported contracted annual recurring revenue of $6.5 million as of 31 July 2026, up 20.9 per cent from $5.4 million in May. Its paying customer base grew from 13 at the prospectus date to 49 by the end of July, an increase of 277 per cent. The company also recorded more than 400 active users, and token consumption grew by over 120 per cent month-on-month between April and July.
Paid utilisation on SCX's Node 1 infrastructure reached just under 35 per cent by the end of July. Keane said the key figures are capacity utilisation and conversion, with active users underpinning near-term revenue potential. He added that growth in token consumption indicates products are delivering and customers are expanding their commitments.
Energy Efficiency and Market Positioning
SCX aims to differentiate itself from the trend of building larger AI factories, especially as community concern grows over the environmental impact of data centres. The company uses only a small amount of energy and zero water for chip cooling, and it does not train AI models for other technology companies. Keane said tougher rules on developers are making the company’s unit economics stronger because competitors face additional costs.
SambaNova's SN40L chips have been reported to deliver between 2.5 and 5.6 times the performance-per-watt of comparable GPU alternatives for selected workloads. SCX says the technology uses around 75 per cent less electricity and 99 per cent less cooling water than conventional GPU approaches. SambaNova does not hold equity in SCX but has a $15.2 million agreement to purchase computing capacity over three years beginning in April 2026.
Growth Plans and Partnerships
SCX is targeting the end of calendar year 2026 for its Node 2 deployment to become operational, with plans to expand across additional Australian locations. The money raised in the IPO will support the purchase of more chips for a second data centre. The company has considered a site in Whyalla, South Australia, but a location has not been confirmed.
The company has added global distribution partner LLMGateway, which connects AI infrastructure providers to more than 40 AI token users through a single API. This partnership is designed to sell SCX capacity internationally while SCX continues to focus on Australian enterprise and government demand. Growth priorities for the second half of financial year 2026 include establishing Node 2, converting active users, and building the go-to-market team.
SCX.ai has entered the public market with commercial momentum and an energy efficiency argument that sets it apart from traditional AI data centre operators. Its first trading session was subdued, but the company remains focused on utilisation and deployment milestones. The coming months will test whether its sovereign inference model can convert demand into sustained revenue growth.