This new facility in Melbourne represents Maincode’s strategy to accelerate the development of purpose-built models for business use. It also challenges the dominance of global tech infrastructure by choosing not to rely on Nvidia’s chips. The AI start-up, backed by the billionaire founder of Stake.com, is responding to strong demand from local organisations that prefer large language models tailored to specific sectors over generic, off-the-shelf chatbots.
Maincode, a well-funded Australian AI firm, gained visibility after introducing its proprietary language model “Matilda.” This system was trained using Australian data and infrastructure. Initially piloted with the Australian Football League, it enabled fans to interact with customised data through a conversational platform. The strong positive response prompted investors to back a larger expansion into local AI development, powered fully by infrastructure located in a Telstra-operated data centre.
The new facility, known internally as “MC-2,” will rely on AMD’s GPU and server technologies. It positions itself as a viable alternative to Nvidia’s dominance in AI hardware. The setup includes Instinct MI355 accelerators and EPYC processors, all supported by open-source ROCm software. This configuration allows Maincode to scale from single projects to a reusable production setup where clients can create and manage their own models.
Maincode’s broader goal is to secure a sustainable role in the global AI market through systems designed to fit local needs. Instead of pursuing a general-purpose strategy like more prominent platforms, its focus is on delivering AI tools for use in areas such as government, aviation and business. Interest from international companies and national governments indicates growing demand for AI that is trained for specific applications and based on sovereign infrastructure.
