He Tingbo, who runs Huawei’s semiconductor business, gave a rare public talk on Monday and proposed a new principle of physics. She called it the Tau Scaling Law, and Huawei in a separate statement said it is also calling the principle Her’s Law, as a nod to her. Her’s Law replaces Moore’s Law, which has governed the chip industry for sixty years and which Huawei would still very much like to be governed by, except for the part where the United States has cut off Huawei’s access to the tools that make Moore’s Law work.
Tau Scaling Law says that instead of shrinking transistors, the trick is to make signals move faster through them. To make signals move faster, Huawei has invented a chip architecture called LogicFolding, which stacks logic circuits in two layers and shortens the wiring between them, producing what Huawei says is 55 percent higher transistor density and 41 percent better power efficiency. Huawei has quietly mass-produced 381 chips based on the broader principle over the past six years, He Tingbo said, and the first chip to use the new LogicFolding architecture specifically will be a Kirin smartphone chip launching later in 2026, with Ascend, Huawei’s AI accelerator line, scheduled to adopt it by 2030. None of these claims have been independently evaluated. The manufacturing tools needed to do LogicFolding at the advanced nodes Huawei is targeting do not yet exist, which He Tingbo did not dwell on.
Huawei is also targeting transistor density equivalent to a 1.4-nanometer process by 2031. TSMC plans to mass-produce real 1.4-nanometer chips in 2028. The current gap between TSMC and the Huawei-SMIC pair is roughly five years. So in five years of additional work, Huawei plans to be three years behind, which is being reported as a narrowing of the gap, and which is, technically, a narrowing of the gap.
It is also a useful way to say something true and something flattering at the same time. The true thing is that export controls have worked: Huawei cannot get EUV lithography, it cannot match TSMC at the leading edge, and it is openly admitting that the leading edge is still three years away in 2031. The flattering thing is that not getting EUV has produced a principle of physics, named after the chairwoman, that lets Huawei describe being three years behind in 2031 as a leap forward.
The reason any of this is worth paying attention to is that the chip layer was the missing piece. The frontier AI lab is DeepSeek, the funding round around it reads like a directory of Chinese industrial capacity, the semiconductor “Big Fund” sits behind the chip companies, the power supplier is CATL, and the chip supplier is Huawei. Tau Scaling Law is the chip supplier’s contribution to the bit. It does not need to produce 1.4-nanometer chips in 2031 to do its job in 2026. Its job in 2026 is to give the Chinese AI buildout a chip story to point to, and “He Tingbo gave a talk and announced a new law” is a chip story.
Whether Her’s Law produces 1.4-nanometer chips in 2031 is a question for 2031. Whether Her’s Law produces a usable narrative in 2026 is a question for now, and the answer appears to be yes.
–
By the Control Plane Editorial Team