Australian AI lab i14 emerges from stealth with $9 million round to cut the compute cost of AI

16 Nov 2026

Australian frontier AI lab i14 has today emerged from stealth with new model architectures that match comparable transformer models at a fraction of the compute and energy cost.

The launch is backed by a new $9 million round led by Salus Ventures, with participation from IP Group, Tin Alley Ventures, Virescent Ventures, Hostplus, Archangel and private investors. i14 has also partnered with AWS, training its models on AWS Trainium chips.

A spin-out of the University of Melbourne, i14 was founded by CEO Tom McLeod, Chief Scientist Professor Simon Dennis, Dr Kevin Shabahang and Michael Diamond and is based alongside the university's computer science department in Melbourne.

The launch comes as access to chips and power has become the main constraint on the global AI build-out. The International Energy Agency expects data centre electricity use to double by 2030 from about 485 TWh in 2025 to 90 TWh.

Almost every large AI model today is built on an architecture known as the transformer, whose memory grows with every word it processes. On long tasks, and for AI agents, that memory quickly exhausts the capacity of a chip.

i14 has taken a different approach. Its architecture draws on about 30 years of research in computational cognitive science into how the human brain forms and recalls memories so efficiently, and keeps memory fixed however long the task runs.

In i14's testing against a transformer of the same size and quality, one chip handles about 10 times as many sessions, and the models are up to dozens of times cheaper to train and run. That performance holds at up to 120 billion parameters, well above the low tens of billions at which competing architectures reportedly failed.

i14 can convert an existing, publicly available AI model to its architecture in a few hours allowing companies to run private frontier models on their own data and hardware.

i14 CEO and co-founder Tom McLeod said: "This is genuinely world class technology and it has been built in Australia. The world is building data centres as fast as it can, and it still can't keep up with demand for AI. Our architectures do the same work with far fewer chips and far less energy. That changes the economics of AI, and it means companies can own and run their own frontier models. We want Australia to be a country that builds frontier AI, not just one that buys it.

i14 co-founder and Chief Scientist Professor Simon Dennis said: "The transformer came from one line of research: scale up what works with more data and more computing power. We came from a different one, which asked how the brain manages memory so efficiently. i14 brings the two together. We build new designs from the way the brain handles memory, and we train them at the scale of modern AI."

Archie Judd, Principal at Salus Ventures, stated: “Running AI at scale depends on scarce chips, data centres and power that few control. By dramatically reducing what's needed, i14 gives businesses and nations far more choice in how and where they run AI. This is a world-class team out of the University of Melbourne, building the kind of sovereign capability Australia must have.”

ENDS

Media contact

Annabel Clunies-Ross

0428 295 517

[email protected]