Huawei's "LogicFolding" Chip End-Around: What It Means For The ASML / TSMC / NVDA Stack
Huawei Technologies announced Monday at an event in Shanghai that it has developed a workaround — branded "LogicFolding" — that lets it make chips matching the transistor density of those manufactured at the 1.4-nanometer process node. It plans to be at parity by 2031, going head-to-head with Intel (INTC), Taiwan Semiconductor (TSM), and Samsung Electronics on the leading edge.
The headline detail: Huawei claims it can do this without ASML's (ASML) extreme-ultraviolet lithography machines, which the U.S. has blocked it from buying since 2022. He Tingbo, president of Huawei's chip arm, called the approach "feasible and affordable."
What's already in production: 381 chip models built on earlier iterations of LogicFolding-style architecture. What's launching this fall: Huawei's next-generation Kirin smartphone chip, the first commercial product on the new architecture.
The Technical Story In Plain English
Conventional chip manufacturing squeezes more transistors onto a single flat silicon wafer. The physics is running out of room — transistors can't get much smaller without quantum-tunneling problems. EUV lithography (ASML's monopoly) is what allows that miniaturization to keep going.
Huawei's response: stack the circuits. Build the chip vertically. Cram more compute into the same silicon footprint by going up instead of getting smaller. The challenge is overheating — stacked layers trap heat — and the need to write more sophisticated code to coordinate across the layers. Huawei says LogicFolding solves both.
Analyst Lian Jye Su at Omdia in Singapore was measured: "Whether Huawei will gain a distinct advantage here remains to be seen, but it's at least an alternative path forward, a breakthrough Huawei managed to find while facing supply chain challenges."
What This Means For The U.S. Stack
Three reads, in order of confidence:
ASML (ASML) thesis weakens at the margin. We've held ASML as the toll-collector on every leading-edge chip globally. If Huawei's alternative path works, ASML's revenue ceiling in China gets capped permanently. The U.S./Korea/Taiwan demand stays intact, but the optionality of "ASML wins the China market once tensions cool" goes away. We're keeping the position but flagging it as a watch-list multiple-compression candidate.
TSMC (TSM) and Samsung remain the dominant fabs. Huawei isn't matching them at the leading edge until 2031. Through the rest of the decade, TSM and Samsung still own the high-end. We're keeping TSM as a core overweight.
Western-fab tooling becomes more valuable, not less. The U.S. CHIPS Act + Korea + Taiwan capex cycles need to keep ramping if the West wants a one-decade head start. That makes Lam Research (LRCX), KLA (KLAC), and Applied Materials (AMAT) structural beneficiaries. We're adding them to the watch list this week.
What This Doesn't Change
The Nvidia (NVDA) and Advanced Micro Devices (AMD) design moats are unaffected. Huawei's competing class is the H20 / H200 generation, not Blackwell or successors. The CUDA software lock-in is also unaffected — Chinese AI developers continue to use Nvidia's ecosystem wherever they can get the silicon.
The pricing power on leading-edge inference and training chips stays with the U.S. names. The geopolitical premium on rare-earth-independent and Western-fab silicon goes up, not down.