The semiconductor industry is on the brink of a fundamental crisis: the largest AI chip manufacturers, including Nvidia, are approaching the physical limits of scaling. This is the key conclusion that Huawei's chief scientist for semiconductors, Liao Heng, voiced during an extensive four-hour interview with a Chinese blogger. This is not about temporary difficulties, but about a systemic dead end into which the multi-year race to increase computing power is leading.

In my deep conviction, Liao Heng's analysis deserves close attention. He points out that the industry has relied for too long on trivial methods: increasing die size, boosting transistor density, and expanding high-bandwidth memory (HBM) capacity. Today, HBM is a critical component in AI accelerators, providing high-speed data exchange between the processor and the computing system. However, as the scientist emphasizes, this approach has a hard limit.

"There must be a limit to how they scale with ever-larger compute dies and more HBM. The industry keeps pushing forward, but once it crosses this physical limit, an avalanche will occur", Liao Heng stated.

Huawei's Alternative Paradigm: Tau Scaling Law

Instead of the dead-end approach of adding more transistors, Huawei is promoting its own methodology — Tau Scaling Law. This approach shifts the focus from miniaturization to reducing signal transmission time between chip components and within the computing system. A key element here is LogicFolding technology, which distributes critical logic circuits across multiple layers, radically reducing latency.

Huawei officially presented this concept in May at the Institute of Electrical and Electronics Engineers conference in Shanghai. At that time, company leadership stated that it expects to achieve transistor density equivalent to 1.4 nm using Tau Scaling by 2031. Moreover, according to my information, Huawei is preparing to unveil its first smartphone chip based on this technology, which would be a serious bid for technological sovereignty.

Sanctions Pressure and the Bet on Efficiency

The forced search for alternative paths is a direct response to the strict U.S. restrictions in effect since 2019. Chinese manufacturers are denied access to ASML's advanced lithography equipment, which closes the path to producing the most modern chips. However, Liao Heng sees this not only as a problem but also as a stimulus for development. He emphasizes that Chinese companies have to compensate for the shortage of computing resources through more complex architectural design.

"We need to invest more effort in design to trade higher complexity for lower consumption of computing resources", he explained.

In this context, the scientist highly praised DeepSeek founder Liang Wenfeng, whose success, in his view, is directly linked to the forced search for efficient architectural solutions under conditions of limited computing power. This once again confirms the thesis that constraints can spur innovation.

The Formation of Two Ecosystems

The geopolitical divide, according to Liao Heng, will inevitably lead to the formation of two independent manufacturing and technological chains. Each side will be forced to build a full production cycle to survive. This is not just a forecast, but a statement of a new reality in which the global semiconductor industry will finally fragment.

Liao Heng's remarks are yet another powerful signal of Huawei's confidence. The company is not merely adapting to sanctions but is also trying to set its own development vector for the entire Chinese semiconductor industry. Recall that in June, Huawei already promised to release a new generation of Ascend AI chips annually with a twofold increase in performance.

My analysis: Liao Heng's statements are not just a scientific discussion, but a clear strategic manifesto. If the physical limit of scaling is indeed reached in the coming years, then Huawei's bet on architectural innovation rather than process technology could prove prophetic. This could radically change the balance of power in the AI chip market, where Nvidia's dominance would cease to be unshakable.