Huawei 'chip queen' returns with a new scaling law

What happened?

On May 25, 2025, He Tingbo, known as Huawei's 'chip queen,' reappeared at the IEEE International Symposium on VLSI Technology after seven years of public silence. Her last appearance was in 2019, when the United States cut off Huawei's access to advanced semiconductor technology. In her presentation, He Tingbo introduced the Tau scaling law, a new approach to improving the performance of artificial intelligence chips without relying on smaller manufacturing processes (such as 3nm or 2nm).

The Tau law is based on a technique called logic folding, which reuses circuits over time to multiply computing capacity without increasing chip area. According to The Next Web, the law establishes a linear relationship between performance and the number of logic folds, allowing efficient scaling on existing hardware. Unlike Moore's Law, which relies on reducing transistor size, and Huang's Law, which focuses on hardware specialization (such as Nvidia's GPUs), the Tau law proposes temporal scaling: instead of making transistors smaller, they are made to work more times per cycle. This could allow chips manufactured on mature nodes (such as 7nm or 14nm) to achieve performance comparable to 3nm in AI tasks.

Why is it important?

This announcement is crucial because Huawei has been under a severe technology embargo since 2019, preventing it from accessing the most advanced design and manufacturing tools (such as those from TSMC or ASML). The Tau law offers an alternative path to maintain competitiveness in AI without needing cutting-edge nodes. Moreover, it directly challenges Moore's Law and Huang's Law (from Nvidia), which have dominated the industry for decades.

If the Tau law is confirmed in real products, it could allow Huawei to produce high-performance AI chips using 7nm or even older processes, reducing its dependence on external suppliers and strengthening its technological autonomy. This would have geopolitical implications, as China seeks to narrow its gap with the U.S. in semiconductors. Historically, Huawei already achieved significant advances in 5G despite restrictions, and the Tau law could be a similar step in semiconductors. However, the current context is more complex: sanctions in 2023 further tightened access to lithography technologies, so an architectural solution like logic folding might be the only viable path for Huawei in the short term.

Market consequences

In the short term, the Tau law is primarily theoretical. Huawei has not presented commercial chips based on this technology, and it is unknown whether logic folding is viable for mass production. However, the mere announcement has already generated expectations among investors and analysts. According to conference data, He Tingbo showed simulations indicating up to a 3x improvement in performance per watt compared to conventional designs, but these data have not been independently verified.

• For Nvidia and AMD: It could erode their technological advantage if Huawei manages to produce competitive chips with older processes. However, practical implementation would take years. Nvidia, for example, is already developing its own logic folding architecture in its Blackwell GPUs, albeit with a different approach. The Tau law could accelerate competition in architectural efficiency.
• For the semiconductor industry: The Tau law could open a new research direction, moving away from miniaturization toward architectural efficiency. Companies like Intel and AMD are already exploring similar techniques (such as chipleting and 3D stacking), but logic folding is a more radical concept. If validated, it could shift R&D investment priorities.
• For consumers: In the long term, it could translate into more affordable AI hardware less dependent on advanced lithography scarcity. However, in the short term, no commercial products are expected before 2027-2028, according to industry analysts cited by The Next Web.

What readers should know

Caution is important. The Tau law is an academic concept presented at a conference, not a market-ready product. He Tingbo did not provide concrete performance data or release dates. Additionally, the reliability of the source (The Next Web) is medium-high, but it has not been cross-checked with other independent sources. Until Huawei publishes results on real chips, this remains promising but unconfirmed speculation. It is worth noting that in the past, Huawei has made theoretical announcements that later did not materialize, such as the HarmonyOS operating system in its early days, although it eventually made progress. Therefore, it is advisable to closely follow the next steps: the publication of peer-reviewed papers and the presentation of prototypes at conferences like ISSCC or Hot Chips.

The Tau law could be a technological 'Trojan horse' allowing Huawei to bypass sanctions through architectural innovation, but its success depends on the industrial viability of logic folding.

Historical context

He Tingbo led the development of the Kirin 9000 chip, Huawei's last major chip before sanctions. Her disappearance in 2019 was seen as a sign of the company's difficulties. Her return with a new scaling law recalls other moments when Chinese companies have tried to innovate to circumvent technological barriers, such as Huawei's development of the 5G network despite restrictions. Furthermore, the concept of logic folding is not entirely new: in the 1990s, IBM researchers explored similar ideas to improve CPU performance, but they were never commercialized due to synchronization complexity. The difference now is that AI workloads are more latency-tolerant and can benefit from sequential execution of folds. If Huawei manages to overcome design challenges (such as increased power consumption from oversampling), it could mark a milestone comparable to Nvidia's GPU leap in 2012 with the Kepler architecture. However, the scientific community is divided: while some see potential, others point out that the Tau law could be a reinterpretation of already known techniques, such as temporal computing or time-multiplexing, without true fundamental innovation. Only time will tell if this 'chip queen' has returned to redefine the industry or if her reign fades into academic noise.