Maincode Secures $30M to Build Australia’s AI Factory With AMD

A significant $30 million investment has been secured by Maincode to construct Australia’s most sophisticated artificial intelligence production facility. Slated to open in Melbourne during January 2026, this new center, designated MC-2 internally, represents a major step in the company’s objective to establish Australian-made AI as a competitive player on the world stage. The facility will significantly enhance Maincode’s capacity to create and manage highly specialized AI models. These models are engineered to tackle structured, rule-based, and context-sensitive tasks that often prove challenging for conventional, general-purpose large language models.

This new chapter builds upon the success of MC-1, the factory responsible for producing Matilda, which was Australia’s inaugural fully developed and trained large language model. The company is now intensifying its core strength: crafting precise, high-performance models that deliver practical solutions to customer’s operational challenges.

Dave Lemphers, Co-Founder and CEO of Maincode, emphasized the project’s significance. “MC-2 is more than an expansion; it’s a declaration of our conviction,” he stated. “Our goal isn’t merely to create larger models. We are developing advanced, focused models that deliver tangible results, built with the precision, efficiency, and pragmatism that are hallmarks of Australian engineering.”

The facility’s core will be powered by next-generation computing and storage infrastructure utilizing advanced GPUs and platforms from AMD. The integration of AMD Instinct MI355 accelerators, AMD EPYC 9575F server CPUs, and the AMD ROCm software stack is designed to provide high-performance training and efficient inference. This integrated environment prioritizes reliability, scalability, and a rapid return on investment.

A critical differentiator for MC-2 is its deep focus on the token layer, the stage where intelligence is fundamentally shaped. Sophisticated software systems at this layer control how models are trained, adapted, and deployed for specific problems. This approach allows Maincode to transcend the limitations of generic, one-size-fits-all LLMs. It enables the construction of purpose-built models from their foundation, delivering genuine capability without relying on the often-frustrating process of extensive prompt engineering.

“Anyone who has attempted to coax a general-purpose model into following complex rules or logic through prompt engineering understands its inherent limitations,” Lemphers explained. “This is precisely where we provide a solution. MC-2 empowers us to develop and operate specialized models from the ground up, custom-built to execute the precise decision-making and reasoning tasks our clients require, completely eliminating the guesswork.”

Peter Chambers, Senior Director of OEM Sales for APJ and Country Manager for Australia at AMD, commented on the collaboration. “At AMD, our vision is to democratize AI by bringing open, high-performance computing to developers, start-ups, and enterprises globally. With our latest generations of GPUs, CPUs, and software, MC-2 is engineered to deliver cutting-edge AI performance with outstanding energy efficiency. We are proud to support Maincode in fostering locally-led innovation with international impact and to contribute to the advancement of Australia’s AI ecosystem.”

The genesis of MC-2 stems from the real-world frustrations experienced by many of Maincode’s partners. Their initial forays into AI often involved experimenting with off-the-shelf LLMs, only to encounter significant obstacles when projects demanded precision and control. MC-2 is designed as the answer to that next phase. It offers these companies a dedicated environment to build their own purpose-designed AI models, with Maincode handling the development, training, and management within a framework optimized for rapid iteration, reliability, and measurable business value.

“We constructed our first factory out of necessity,” Lemphers recalled. “We needed AI that could think and reason in alignment with actual business operations. MC-2 is the natural evolution, a place where we can build those models more rapidly, more intelligently, and at a much larger scale.”

MC-2 integrates computation, data, research, and deployment into a single, cohesive environment. This creates a robust foundation for the continuous creation and refinement of AI. It will power the next iterations of the Matilda models, develop customer-specific systems, and support applied AI research programs, all managed comprehensively by Maincode from start to finish.

This new infrastructure also accelerates progress in the emerging field of agentic AI. By constructing models from scratch for specific purposes, rather than adapting general chat systems, Maincode enables AI that can reason, plan, and execute actions within defined parameters. This paves the way for adaptive, goal-oriented systems poised to transform operational workflows across numerous industries.

The $30 million commitment from Maincode stands as one of the largest private investments in AI infrastructure within Australia. It signals a profound, long-term belief that the future of AI resides in applied intelligence designed for specific tasks, rather than in the pursuit of artificial general intelligence.

“What we are constructing with MC-2 is pure capability,” Lemphers affirmed. “We are leveraging the world’s best technology and combining it with Australian ingenuity and inventiveness. We are not a chatbot company. We are an AI company that builds completely new models from the ground up, systems engineered to manage the work that general-purpose models simply cannot, regardless of how much prompt engineering is applied. MC-2 is not about scale; it’s about precision. This is where the next generation of useful, practical AI will be manufactured.”

![Image: A modern, high-tech data center with server racks and cooling systems, representing the advanced infrastructure of the MC-2 AI factory.]

(Source: ITWire Australia)