Huawei chief semiconductor scientist Liao Heng said in a recent public livestream that major Western chipmakers led by Nvidia are getting close to physical manufacturing limits in the push for more powerful processors. She also said Huawei’s first smartphone chip designed under its in-house Tau Scaling Law framework will arrive later this year, giving the market a first look at how that alternative path performs in practice.
Liao says conventional scaling is running into hard constraints
According to Bloomberg, Liao made the remarks during a four-hour livestream, where she said the traditional route followed by foundries such as Taiwan Semiconductor Manufacturing Co. and Intel — continually shrinking transistor sizes — is hitting a physical bottleneck. She said Nvidia, still the most important chip company globally, continues to build larger chips with billions of transistors and more high-bandwidth memory, or HBM, but the marginal gains from that strategy are falling quickly.
Liao said that once the industry reaches those physical limits, it will trigger what she described as an avalanche-like shift, forcing the sector to look for a new direction of technological progress.
She also acknowledged that companies such as TSMC and Samsung have detailed roadmaps built around ASML’s advanced lithography systems. Huawei, by contrast, has been unable to access those key tools because of U.S. sanctions and was pushed onto what she called an alternative route. She said that path will eventually show the world how fast the gap between China and the West is narrowing.
Tau Scaling Law and LogicFolding face a market test later this year
The alternative route Liao referred to is Huawei’s Tau Scaling Law, or τ, framework, which the company formally presented in May at the IEEE International Symposium on Circuits and Systems, or ISCAS.
The core idea is to replace traditional geometric scaling with what Huawei calls time scaling. Rather than focusing only on shrinking the physical size of transistors, the framework aims to keep performance improving by cutting signal transmission time and increasing transfer speeds between components across the system. Huawei has said it designed and mass-produced 381 chips under this principle over the past six years.
The key implementation technology behind the framework is LogicFolding. Liao said this method significantly improves performance and transistor density by shortening on-chip interconnects. In the livestream, she said Huawei is preparing to launch its first smartphone chip built on the framework and using LogicFolding, with a release expected later this year. She described that product as a crucial step in moving Tau Scaling Law from theory into market validation.
DeepSeek cited as proof that design can substitute for brute-force compute
Liao also praised DeepSeek founder Liang Wenfeng, calling him the clearest example of the idea that better design can offset raw computing constraints. She said DeepSeek trained a top-tier large language model under very limited computing resources through innovation in model architecture, showing that Chinese developers can build systems at a comparable level using only a fraction of the compute required by U.S. peers.
“We must invest more effort in design, using greater complexity to trade for lower computational resource consumption,” Liao said. She added that the trend has accelerated this year, with multiple Chinese companies releasing benchmark models with leading performance.
Huawei sees the AI and semiconductor supply chain splitting in two
On geopolitics, Liao said she was pessimistic. As tensions between China and the United States intensify, she said, the global semiconductor supply chain is likely to split into two separate ecosystems, with the two countries diverging more sharply in how they pursue chip advances.
“For survival, even if there is no serious conflict between the two sides, each side must build its own complete manufacturing and supply capabilities,” she said. The article said that view aligns closely with China’s broader long-term strategy of reducing dependence on Western technology in semiconductors.
It also noted that industry research firms including SemiAnalysis and TechInsights had already given the market a clearer picture of Huawei’s alternative approach earlier this year after tearing down the company’s new smartphone chip. Liao’s latest public remarks signaled that Huawei does not see the route as a temporary workaround, but as a long-cycle technology path it intends to keep advancing.