Huawei Technologies Co’s top semiconductor scientist warned of the looming physical limits Western chipmaking giants, including Nvidia Corp, face in their pursuit of more powerful processors, in a rare public appearance streamed in China.
Across a marathon four-hour interview aired in late July, Liao Heng expounded on his views about the best path forward in chip design, addressed some of the hardships Huawei faced when US sanctions cut it off from global chipmaking suppliers half a decade ago, and lauded his company’s novel approach, dubbed the Tau Scaling Law.
Liao’s appearance with a local influencer was a fresh signal of confidence from the Shenzhen-based semiconductor engineering leader of the country, following a May keynote address titled New Semiconductor Path in Practice where the company detailed its design philosophy. He made the case that the decades-long practice of continuously shrinking and making integrated circuits more dense will soon run out of runway. Nvidia, today’s most important chip company in the world, is building ever larger chips with billions of transistors and increasing amounts of high-bandwidth memory attached.
"There has to be a limit in how they scale up with ever-increasing compute die and more HBM,” Liao said. "The industry is still pushing, but once they cross that physical limit, there will be an avalanche.”
Liao’s comments suggest he shares a widely held view that chip foundries like Taiwan Semiconductor Manufacturing Co and Intel Corp may eventually run into a wall in attempting to further and further reduce the size of each transistor inside a chip. Both companies, alongside Samsung Electronics Co, have extensive roadmaps stretching years into the future, aided by the indispensable advanced lithography machinery from ASML Holding NV. But Huawei, which is prevented from accessing those systems due to sanctions directed at China as well as the company itself, has had to think more creatively to find an alternative path to catch up.
The company’s Tau Scaling Law proposes focusing on improving transmission speeds between parts in a computer system, as the inherent advantages of shrinking chips start to dissipate. Huawei is set to soon unveil its first smartphone chip designed under that framework, through a technology called LogicFolding.
Liao said the world will have a clear understanding later this year of how Huawei’s alternative approach will help it narrow the gap with rivals when its new smartphone chip is dissected by industry research outfits such as SemiAnalysis and TechInsights. The growing bifurcation between the US and China in how they pursue semiconductor advances, according to Liao, comes as rising geopolitical tension cleaves the global supply chain into two increasingly discrete ecosystems.
"To survive, each side has to build its own complete manufacturing and supply capabilities, even without serious clashes between the two sides,” Liao said.
The uncommon public discussion with a scientist at the heart of China’s semiconductor self-sufficiency drive shows how the country is hardening its resolve to wean itself off of Western technology in the AI era.
During the podcast interview, the Huawei scientist also offered a metaphor for the entire AI value chain, representing it as an 18-level pagoda and one-upping the layer cake framework analogy of Nvidia chief executive officer Jensen Huang. Liao emphasised the need for Chinese players across each layer to support each other and work collectively to overcome external sanctions, especially between chipmakers and AI model developers.
He then showered praise on Liang Wenfeng, the founder of Hangzhou-based AI lab DeepSeek, whose cutting-edge model in 2025 triggered a multibillion-dollar selloff in AI stocks by showing how Chinese developers could train systems at a fraction of the computing resources used by American peers.
DeepSeek’s breakthrough came as Liang and his team sought innovation in the model’s architectural design in view of the limited computing resources they have, Liao said. That momentum has picked up pace this year, as several other Chinese companies have released benchmark-topping models.
He likened Chinese AI innovation to the way people maximise space in a cramped apartment while neighbours and rivals live in a more spacious villa. Huawei itself is developing its AI chips to better support efficiency-focused model architectures through what Liao called a co-design mechanism.
"We have to put in more effort around design to trade higher complexity for less compute resource consumption,” Liao said. – Bloomberg
