Watch / Semiconductor Australia 2026
AI’s memory bottleneck creating a new opportunity
AI’s memory bottleneck creating a new opportunity
About this video
Get tickets to Australia's premier deep tech investor event of the year. ausbiz subscribers get 20% off with the code AUSBIZ26. Go to semiconductoraustralia.com
Key points:
AI inference is shifting memory demands across data centres and edge devices
Weebit Nano says ReRAM could tackle AI’s growing memory bottleneck
Commercialisation and a first AI customer are key catalysts for FY27
AI is entering a new phase as the industry shifts from training increasingly powerful models towards running them at scale. That transition is putting greater pressure on memory performance and efficiency, particularly as AI moves beyond data centres and into cars, robots and industrial equipment.
Weebit Nano CEO Coby Hanoch explains why the rise of AI inference is changing the requirements for semiconductor memory and where the company’s ReRAM technology could fit into that next wave of deployment. With large amounts of power and processing time spent moving data between memory and compute, the company sees in-memory computing as a potential way to address the growing bottleneck.
The focus now turns to commercialisation, with Weebit Nano already working with major semiconductor manufacturers and seeing its technology progress through sampling. Hanoch outlines the path towards mass production, ongoing technology development and the company’s target of announcing its first AI customer before the end of calendar 2026.