i14, a Melbourne company that develops foundation AI models, raised A$9m (about $6.3m) in a seed round announced on 5 November 2026.
Salus Ventures led the round, according to Dealroom. Tin Alley Ventures, the superannuation fund Hostplus, London-listed IP Group, Archangel Ventures and Virescent Ventures also invested, according to SmartCompany and Dealroom, which both cited The Australian Financial Review. Virescent confirmed its own commitment, made from its second fund. The company did not disclose a valuation.
i14 was spun out of research at the University of Melbourne and founded in June 2025, according to Dealroom.
What i14 does
Most large language models, including ChatGPT, Claude and Gemini, are built on the transformer architecture, which compares every word in a prompt with every other word. Virescent notes that this means the work grows with the square of the input length: doubling a prompt roughly quadruples the computation, and long documents can tie up several expensive GPUs.
i14 takes a different approach with what it calls the Laplace Sequence Model. Instead of keeping an exact record of every earlier word, the model keeps a compressed gist of what came before, modelled on how the human brain stores and recalls memories. According to Virescent, this gives the model a fixed memory footprint that can be dozens of times smaller than a comparable transformer, and a flat cost for each word however long the context grows.
Virescent said early tests of i14's 1 trillion parameter model matched comparable models while using a fraction of the compute and energy. It gave no benchmark details. SmartCompany reported that i14 says its products can cut costs by up to 85%, depending on the task. The company also says its architecture can be built into existing models.
The team
The model is the result of about 30 years of research by co-founder and chief scientific officer Simon Dennis, a cognitive scientist who builds computational models of human memory. Chief executive Tom McLeod is a two-time software founder with a master's degree in natural language processing. Co-founder Michael Diamond works across cognitive science, data science and applied AI.
Kevin Shabahang, the vice-president of AI research, did his PhD in language modelling under Dennis and was previously co-founder and chief technology officer of legal tech company Time Credit AI. SmartCompany and Dealroom list him among the founders; Virescent does not.
Who it is selling to
i14 is aiming at neoclouds, hyperscalers and other heavy users of AI that want to lower their compute bills, Dealroom reported. "The world is building data centres as fast as it can, and it still can't keep up with demand for AI," McLeod said, according to Dealroom.
Virescent framed its investment around shortages of chips and power that are limiting how fast AI can be built out. It said i14 does not need to outspend larger labs because it can build on existing models' training while cutting running costs. It also noted the risks: no new architecture has yet proven itself at the frontier, and the incumbents have far more money.
What comes next
i14 has not said how it will spend the money. Dealroom reported that the founders kept a significant stake and are considering raising again in about six months.