US Closes Access to Lithography, Huawei Finds New Ways to Develop AI Chips

JAKARTA - Huawei Technologies Co is looking for other ways to catch up on AI chip performance after its access to advanced lithography machines was closed by US sanctions. Its path is no longer just shrinking transistors, but speeding up the connections between components in a single system.

China Daily, quoted Thursday, August 6, reported that the approach began to gain attention when foreign chip companies approached the physical limit in making more powerful processors.

For decades, the chip industry has relied on the old way. Transistors are made smaller so that more of them can be packed into an integrated circuit. But that space is getting tighter as transistor sizes now approach the scale of atoms.

Huawei Semiconductor Chief Scientist Liao Heng said Western companies still have a roadmap for the next few years thanks to US advanced lithography machines.

Lithography is a process of printing very small patterns on a wafer to make chips. A wafer is a basic material plate used in semiconductor production.

Huawei is in a different position. The Chinese technology company cannot access the equipment due to US government sanctions.

Amid the pressure, Huawei is pushing an approach called Tau Scaling Law. The concept proposed in May shifts the focus from shrinking chip size to increasing the speed of transmission between system components.

Huawei is scheduled to introduce its first smartphone chip designed with the framework in September. The chip uses a technology called LogicFolding.

According to Liao, the industry will see more clearly by the end of this year how Huawei's alternative approach helps narrow the gap with competitors. The assessment will come after the new mobile chip is disassembled and researched by industry research companies.

Liao also likened the AI value chain to an "18-tier pagoda". This framework is said to be more comprehensive than Nvidia CEO Jensen Huang's "five-layer cake" model.

Huang's model divides the AI sector into five layers, namely energy, chips, computing infrastructure, cloud or AI models, and applications. The framework explains how business value flows in the AI industry to generate profits at the application layer.

Meanwhile, the "18-level pagoda" Liao looks at AI from the engineering and production sides. The basic part includes theory, energy, mining and raw materials, silicon, transistor design, wafer fabrication, lithography and equipment, as well as advanced packaging.

The middle part includes the architecture of the chip, the environment to run the program, and quantization. Quantization is a technique for simplifying the calculation of AI models so that they can be run more efficiently on computing devices.

The upper layer covers AI training methods, large language models, AI agents, commercial products, and applications. AI agents are systems that can carry out certain tasks more independently based on user instructions.

According to Liao, the performance limit of this sector is determined by the shortest layer, or the weakest point in the chain.

Professor Zhou Jianjun from the School of Integrated Circuits, Shanghai Jiao Tong University, assessed Huawei's move as having major strategic implications.

According to Zhou, Huawei is looking for a different path to improve performance. Not by continuing to shrink transistors, but by making groups of chips and the overall AI system work better together.

Xiang Ligang, Director General of the Information Consumption Alliance, said Huang's model sees the AI industry as a market opportunity that can be seized layer by layer.

Instead, Liao's model sees AI as an integrated engineering challenge. That is, no single layer can be optimized alone.

For Huawei, this view is not just a theory. Access to advanced lithography equipment is still closed.

"Huawei's breakthrough lies in cross-layer collaboration, namely using the advantages of system clusters to close the hardware shortcomings of single chips," said Xiang.

China Daily said that shifts were also seen in China's AI accelerator market in 2025. AI accelerators are chips or hardware designed to accelerate the processing of artificial intelligence.

According to the latest report of the US market research company, International Data Corp or IDC, US export controls have also contributed to major changes in the market.

China's domestic suppliers led by Huawei recorded a surge in combined market share to 41 percent. Nvidia's share fell to 55 percent, far from its previous position of being called a near monopoly.