← Back to stories Industrial factory with metal pipes and special equipment surrounded with fences located against cloudy sky on street with power lines in town
Photo by Brett Sayles on Pexels
凤凰科技 2026-05-27

Huawei (华为) unveils "Tao's Law" chip roadmap, pledges multi‑GHz Kirin and Kunpeng advances — reportedly

What Huawei announced

Huawei (华为) has reportedly introduced a new internal design philosophy dubbed "Tao's Law" (韬定律) intended to steer its semiconductor development away from a strict Moore's Law trajectory and toward architecture‑and‑integration optimization. It has been reported that the company presented a detailed product roadmap at recent forums, promising a string of Kirin (麒麟) mobile chips and Kunpeng (鲲鹏) server/AI processors with sharply higher transistor densities and much higher clock speeds over the coming years.

Technical claims and roadmap

According to the slides that have circulated, it has been reported that a tentative Kirin 9050 this autumn would reach 238 mtr/mm2 transistor density (a 53.5% jump) and run at about 3.1 GHz, with successive Kirin generations climbing to roughly 3.38 GHz (9060), 3.71 GHz (9070) and breaking past 4 GHz by 2029, ultimately aiming as high as 5.0 GHz by 2031. For servers, it has been reported that last year’s Kunpeng 950 rose from 64 to 96 cores and 133 mtr/mm2 density; the next Kunpeng 960 reportedly targets >200 mtr/mm2 and a 4.0 GHz design frequency, with core counts the slides suggest could scale into the 128–192 range. For context, it has been reported that current TSMC 3nm ARM processors run around 4 GHz, sometimes peaking above 4.5 GHz.

Why this matters — and the geopolitical backdrop

If accurate, these claims signal a push by Huawei to close architecture and performance gaps across mobile, PC and AI server domains and to de‑risk China’s chip stack amid U.S. export controls and sanctions that have constrained access to leading‑edge foundry process and tooling. Can Huawei turn PPT figures into silicon that competes with Apple and Qualcomm on mobiles or with established server CPUs in AI workloads? That remains the central question for Western observers and Chinese planners alike — success would bolster domestic self‑reliance; failure would underscore ongoing industrial constraints.

Caveats and source

It has been reported that the material originated on ifeng's platform as user‑uploaded content and carries the platform's storage‑only disclaimer, so the roadmap details remain effectively unverified until Huawei publishes chips or independent benchmarks. Reportedly bold numbers on slides are worth scrutiny: presentation claims do not always translate into production timelines or real‑world performance.

AISpaceTelecom
View original source →