Liao’s core claim was that the Western industry’s established route—ever-smaller transistors, ever-larger processors, and more attached memory—will eventually hit hard physical limits. Huawei’s answer is to seek system-level speed and density gains without relying solely on the most advanced lithography tools. ![]()
![]()
Physical ceiling and Nvidia’s example
- Liao argued that Nvidia, TSMC, Intel, Samsung, and others cannot indefinitely gain performance by shrinking and packing in more transistors; the approach faces a physical ceiling.
![]()
![]()
- He used the direction of Nvidia’s AI accelerators—larger processor dies or packages with growing amounts of high-bandwidth memory (HBM)—as evidence that conventional scaling is becoming increasingly difficult: expanding chip and memory scale also runs into limits of area, power, heat, interconnects, and manufacturability.
![]()
- This was Liao’s argument, not a demonstrated industry consensus; independent experts have differed on how novel or consequential Huawei’s proposed alternative will prove to be.
![]()