Huawei's Tau scaling law challenges Moore's Law with speed focus

Facing US tech export restrictions that cut off access to the world's most advanced chipmaking machinery, Huawei is proposing a fundamental change in semiconductor progress. Instead of obsessing over shrinking transistors to double density (Moore's Law), the company's new Tau (τ) Scaling Law, unveiled by semiconductor chief He Tingbo at the IEEE International Symposium on Circuits and Systems in Shanghai, focuses on reducing time constants—the delays that occur as data travels through transistors, wires, memory, and across entire data-centre clusters. The Greek letter τ represents these time constants, and the new framework argues that users ultimately care about how fast computing tasks get done, not the physical size of individual transistors.

By prioritizing data movement speed over transistor density, Huawei aims to deliver world-beating chip performance even without access to extreme ultraviolet (EUV) lithography tools. This could rewrite the playbook for China's tech survival, allowing domestic chipmakers to compete using different metrics. However, the industry is divided: some see it as a necessary adaptation to restrictions, while others question whether it can truly match the performance gains from traditional transistor scaling. The Tau law's success will depend on whether Huawei can implement designs that minimize time constants at scale, requiring innovations in architecture, packaging, and memory hierarchy rather than pure fabrication advances.