ASUS AI: Hybrid Architecture Cuts Inference Costs by 70%

ASUS is integrating a new hybrid AI architecture across its commercial devices. This includes ExpertBook laptops, ExpertCenter desktops, and NUC mini PCs. The goal is to balance performance and cost for businesses using generative AI. This architecture combines Phison’s aiDAPTIV technology with intelligent workload routing. It dynamically distributes AI tasks between local devices and the cloud. This means simpler tasks are handled locally, while more complex ones go to the cloud. What's interesting is this approach can cut token inference costs by up to 70%. This is especially true for mid to large-scale language models. The company says this will make AI adoption more practical and scalable for businesses.