Adelaide-based AI robotics startup Metacognition has secured an additional $5 million in funding from Airtree, extending its pre-Seed round to $15m just a month after a $10m raise. The company builds reliable AI agent and robot software, including an operating system named Zeos.
Extending the Pre-Seed Round in Adelaide
The fresh capital arrives following a $10m pre-Seed injection secured by the startup last month. Founded earlier in the year by professors Anton van den Hengel and Stephen Gould alongside Dr Paul Dalby, Metacognition focuses on developing navigation models, reasoning mechanisms, and memory tools designed for physical machinery.
The newly expanded $15m war chest coincides with the release of NavGPT-3. Developed alongside collaborators, the company states that NavGPT-3 sets a world benchmark for robots navigating by following verbal directions.
The Risk of Relying on Overseas AI Models
Airtree co-founder Craig Blair warned that Australia risks losing the economic benefits of AI if it simply hosts technology developed overseas. “If all we do is keep the lights on for models trained elsewhere, the value walks out the door with the tenants,” Blair said.
Blair pointed out the necessity of backing local talent working on foundational components. “We don’t need to win everything, but we do need to be in the game, which means backing the people building the missing pieces,” he said. “Anton and Steve are leading one of the world’s best research teams, tackling some of the hardest unsolved problems in AI right here in Adelaide.”
Targeting Reliable Industrial Automation
Metacognition targets industrial environments where software must govern real-world physical consequences. Early applications included building simulations allowing operators to manage open-cut mines or electricity grids using spoken commands. The startup’s Zeos operating system, navigation models, and memory tools are engineered to let machines store instructions, react safely to emergencies, and improve over days and weeks.
Van den Hengel emphasized the current limitations of deployed artificial intelligence systems. “If agents and robots are to be our trusted co-workers, they need to be able to have the flexibility to talk in natural language and carry out tasks without surprising us,” van den Hengel said.
Public Co-Investment and Private Capital Projections
To support the broader ecosystem argument, Airtree highlighted modeling sourced from the Australian Academy of Technological Science and Engineering. The data suggests that targeted public co-investment in AI enablers could mobilize between $27 billion and $49bn in new private capital.
