Huawei semiconductor chief He Tingbo presented what the company calls the Tau, or τ, Scaling Law at the IEEE International Symposium on Circuits and Systems in Shanghai on May 25, 2026. Huawei says the framework can improve computing performance by reducing end-to-end delay across devices, chips, software and complete systems rather than relying only on ever-smaller transistors.
Moore’s Law is a benchmark, not a physical commandment
Moore’s Law originated as an observation that transistor counts on economically produced integrated circuits tended to double on a regular cadence. For decades, semiconductor manufacturing improvements made smaller transistors a reliable path to lower cost and higher performance. The pace has slowed as fabrication becomes more physically difficult and enormously expensive.
The industry has not simply stopped innovating. Chiplets, advanced packaging, three-dimensional stacking, specialized accelerators, faster interconnects and software optimization all seek gains that transistor shrinkage alone no longer supplies cheaply.
Huawei’s Tau proposal
He described τ as a measure connected to the time required for useful work and data movement through a computing system. Huawei’s public account divides optimization across semiconductor devices, integrated circuits, chips and systems. At chip level it emphasizes coordinated design of software, architecture and silicon so instruction and data flows suit real workloads.
The company also promoted “LogicFolding,” an approach intended to increase effective density and shorten communication paths through structural design and stacking. Huawei said it had applied elements of the broader methodology in hundreds of chips and planned to introduce LogicFolding in a Kirin processor.
Why the claim matters geopolitically
U.S. export controls restrict China’s access to advanced chips and manufacturing equipment, particularly extreme-ultraviolet lithography tools used at leading nodes. A design method that extracts more performance from available fabrication could reduce—but not erase—the disadvantage created by those restrictions.
System-level efficiency is especially important for artificial-intelligence clusters, where memory bandwidth, communication, power and cooling can limit useful performance. A processor with less advanced transistors can still compete on a defined workload if architecture and software use resources more efficiently.
Why independent verification is essential
Tau Scaling Law is Huawei’s terminology and roadmap, not yet an industry law established by a long public record. Company presentations may select favourable workloads or compare “equivalent” density in ways that do not capture power, yield, cost and manufacturability. Independent benchmarks and physical analysis of shipped products are needed.
Stacking logic also creates engineering challenges. Heat removal, alignment, interconnect reliability, packaging yield and repairability can offset density or latency gains. A concept that works in a demonstration must scale to high-volume production at acceptable cost.
A challenge, not proof of dominance
He’s presentation signalled that Huawei intends to compete through co-design and packaging while working around manufacturing constraints. It may pressure U.S., Taiwanese, South Korean and European firms to accelerate similar system-level approaches, many of which they already pursue.
The announcement does not prove that Huawei has overtaken TSMC, Nvidia or other leaders, nor that Moore’s Law has been “replaced” for the entire industry. It sets a testable proposition: future Huawei products should deliver sustained gains in performance, energy, density and cost consistent with the claimed framework. Shipped silicon and reproducible benchmarks will decide whether Tau is a durable engineering principle or mostly a powerful brand.
source
