Icetana (ICE) Investor update summary

Product and deployment

• AI monitored 20,000+ cameras across 80+ sites in 15+ countries; gross margin ~90%.

• Self-learning identifies normal activity within 24 hours and flags unusual events in real time, alerting teams within 3 seconds without manual setup.

• Operator capacity is ~50 cameras without the software, ~600 with Safety & Security, and over 2,000 with Antara Core and Triage Agent.

• Antara Core runs private generative AI on-prem, keeping footage on site with no ongoing token costs; Triage Agent checks events against standard operating procedures before escalation.

• Configurable event detection supports hazards, aggression and weapons; roadmap includes spill detection, robotics integration and agentic closed-loop workflows, aligned with SmartBX.

Commercial progress

• ARR rose 19% in the quarter to $3.1m at Sep 2026; FY26 revenue was $2.5m versus $1.9m in FY25, up 31% YoY.

• Three Antara Core orders in three months generated AUD 1.29m contracted since the June launch; five-year SaaS terms and ~90% gross margin were stated.

• Millennium orders included $199,000, $419,000 and $675,000; the September order followed a four-week live trial, and four announced orders in 12 months were 60-month SaaS contracts.

• Millennium also featured a $2.1m three-year expansion and a $376,000, 60-month SaaS sale; it co-developed Hazard Detection and serves 100+ shopping centres across Australia and NZ.

• Majid Al Futtaim expanded its March 2026 renewal, increasing camera count and ARR and extending its subscription by three years.

Channels and growth strategy

• SoftBank Robotics has five regional distribution agreements; SBR America signed in June 2026, with live deployments in ANZ and Japan and an active MENA pipeline.

• Estimated opportunity spans 10m–50m cameras across SoftBank Robotics territories and 1.2bn+ cameras globally in 2025; 1% of the global base is cited as 12m cameras.

• One million cameras under licence is an internal ambition, with ~$150 per camera per year used as a planning assumption; it is not a forecast or guidance.

• Australian retail property has an estimated 250,000 cameras; expansion plans prioritize proving the model in Australia before targeting North America.

• Core algorithms were trained on 700m hours of retail CCTV footage; internal modeling estimates Hazard Detection could avoid AUD 400,000 annually in trip-and-fall liability, while first-customer cloud AI savings are stated at up to $4m p.a.