Huawei has outlined an alternative semiconductor scaling roadmap that it says could deliver chip technology equivalent to the 1.4nm-class process node by 2031. Presented at the 2026 IEEE International Symposium on Circuits and Systems (ISCAS), the company’s so-called tau scaling principle essentially moves to supplant 2D scaling with optimizations on various levels.
“In recent years, Moore’s Law has faced severe physical limits and diminishing economic returns. The global industry has been increasingly constrained by the slowdown in the geometric scaling of transistors and the erosion of cost-per-transistor benefits. It must now tackle the urgent and common challenge of overcoming the physical constraints of traditional processes and finding a new, sustainable evolution path that can match surging computing demands. This is where the tau scaling law comes into play,” Huawei writes in a press release.
Based on this law, the Chinese firm has developed multi-level co-optimization mechanisms that span semiconductor devices, circuits, chips and systems. These include optimizing resistance and parasitic capacitance of transistors and interconnects, so-called Logicfolding to shorten critical-path wiring, expanding system-level parallelism and reducing system communication latency.
According to Huawei, the methodology has already been applied commercially. The company says it has designed and produced 381 chips using elements of the approach over the past six years. Its flagship Kirin chips, scheduled to launch later this year, will be the first to adopt Logicfolding.
Critics argue that Huawei is describing existing industry practices rather than a fundamentally new scaling paradigm. Advanced packaging, chiplets and system-level optimizations have already become mainstream strategies, which complement rather than replace dimensional scaling. As the company is lithographically constrained by export restrictions, the approach may allow Huawei to extract more performance from mature process technologies through packaging and architecture, but it seems unlikely that it will be able to compete with the industry’s world-leading companies.
