Tau Scaling Law: The Path to Faster AI Inference

The next breakthrough in AI inference may not come from the model itself — it may come from the system underneath it. In this video, @Ronald_vanLoon — globally recognized AI and digital transformation thought leader — breaks down Huawei's Tau Scaling Law, or Her's Law and what it means for enterprise AI infrastructure. At IEEE ISCAS 2026 in Shanghai, Huawei's He Tingbo introduced the Tau Scaling Law, or Her's Law as a new guiding principle for the post-Moore era: instead of making transistors smaller, reduce the time it takes for data and instructions to move across the entire system — from devices and circuits to chips, memory, and interconnects. This shift matters because AI inference is not just a model problem — it is a system-level time problem. Training is about throughput. Inference demands real-time response. And as AI scales, data movement becomes as costly and as limiting as compute itself. The cost per token begins well below the software layer. Ronald van Loon walks through the key technologies @huawei connects to this principle — logic folding, unified bus architecture, and optical interconnects — and explains why the next AI competitive advantage may come not from a bigger model or a faster chip, but from reducing the time lost as data moves through the system. Learn more about Tau Scaling and what it means for the post-Moore era: https://bit.ly/4dNyb8Tg #HuaweiPartner #Huawei #AI #MooresLaw #Cloud #Semiconductors #DataCenter #ChipDesign #DigitalTransformation

from Ronald van Loon https://www.youtube.com/watch?v=Ys2FF8IKhFM