Inside Huawei’s ISCAS 2026 Reveal: LogicFolding, Tau Scaling Law, and a Wave of Semiconductor Optimism

On May 25, 2026, Huawei's semiconductor chief He Tingbo took the stage at the IEEE International Symposium on Circuits and Systems (ISCAS) in Shanghai with a message designed to reshape the conversation around China's chip ambitions. The company introduced two intertwined concepts: a vertical chip-stacking architecture called LogicFolding and a new design philosophy named the Tau (τ) Scaling Law . Together, they form a public roadmap aimed at producing chips with transistor density equivalent to 1.4 nanometers by 2031—without access to the EUV lithography machines that the US has blocked from reaching China .

The immediate reaction was a powerful rally across Chinese semiconductor stocks as investors bet on a potential sanctions workaround, but the announcement also drew sharp skepticism from industry analysts who noted that the roadmap still lacks physical silicon.

A New Architectural Playbook

At its core, LogicFolding is a departure from the traditional race to shrink transistors horizontally. Instead, Huawei proposes stacking logic circuits vertically to increase transistor density. The company claims this approach can boost density by 55% and improve power efficiency by 41% without needing the most advanced lithography tools from ASML .

The Tau Scaling Law provides the theoretical framework behind this pivot. Rather than following Moore's Law and its focus on geometric scaling—making transistors physically smaller—Huawei's new principle centers on compressing signal propagation delays and data movement time (tau, or τ) across the entire computing stack. He Tingbo said the company spent six years developing this methodology and is already applying it across its chip lines .

"The τ Scaling Law puts the time it takes for signals and data to move through a chip and its surrounding system at the centre of design," Huawei's announcement stated . The idea is that by shortening wiring, reducing latency, and optimizing data flow, Huawei can keep pushing performance forward even when further transistor shrinks are blocked by export controls.

The 2031 Target and Kirin Roadmap

The most headline-grabbing claim was a timeline: 1.4-nanometer-class chips by 2031. For context, TSMC has stated it expects to begin mass production of its own 1.4nm process in 2028 . Huawei's target would close a gap of several generations, though it remains three years behind the industry leader's stated schedule.

More near-term, the company said its upcoming Kirin smartphone processors—expected in late 2026—will be the first to adopt the LogicFolding architecture. Those chips are projected to reach 238 million transistors per square millimeter, with roughly 13% faster clock speeds and the 41% power efficiency gain . LogicFolding is also slated for Huawei's Ascend AI chips by 2030 and for large-scale AI clusters in data centers .

The Market Roars

The stock market response was swift and broad. On mainland China's A-share market, SMIC surged as much as 19%, reaching a total market capitalization of RMB 1.25 trillion . Hua Hong Semiconductor rose by the full 20% daily trading limit. Dongxin Semiconductor, ACM Research (Shanghai), and Yongxi Electronics also hit their 20% limit-up thresholds, while Cambricon soared over 8% . The rally extended beyond foundries into semiconductor, PCB, advanced packaging, and memory chip sectors .

When Hong Kong markets reopened after a holiday on May 26, the momentum continued. SMIC gained as much as 16% in Hong Kong, while Hua Hong Semiconductor jumped nearly 12%. GigaDevice and Innoscience advanced 4% to 7% . Analysts cited "renewed optimism" that Chinese chipmakers could narrow the gap with foreign rivals despite US export restrictions .

The Skepticism That Matters

For all the market enthusiasm, the technical community has raised several red flags.

No working chip. The ISCAS presentation was a design methodology and a projected roadmap. No physical LogicFolding chip has been demonstrated, and the claims about density and efficiency improvements remain theoretical . Turning a chip architecture into a manufactured product requires a viable foundry process for stacking logic layers vertically—a challenge that is itself non-trivial.

The EUV gap hasn't disappeared. While LogicFolding aims to reduce reliance on the most advanced EUV machines, it does not eliminate the need for capable lithography entirely. "It would be easier to manufacture 7nm chips" with LogicFolding than to actually reach 1.4nm-equivalent density without EUV, one industry analyst noted . The approach may help at mature nodes, but achieving true leading-edge density without ASML's most advanced tools remains unproven.

A five-year horizon with high execution risk. The semiconductor industry is littered with ambitious architecture announcements that failed to deliver at scale. The Tau Scaling Law is a heuristic framework—a design philosophy—rather than a physical law, making it harder to validate independently or to benchmark against conventional node transitions .

The sanctions backdrop is intensifying. The announcement came just weeks after US lawmakers introduced the MATCH Act, a bipartisan bill that would further tighten export controls by restricting immersion DUV lithography tools to China . These are the older but still capable machines that Chinese fabs have relied on for advanced work. If the MATCH Act passes, even the fallback equipment that LogicFolding's roadmap may need could become harder to access.

What the Rally Actually Reflects

The stock surge is less a verdict on LogicFolding's technical viability than a signal of how sensitive Chinese semiconductor equities have become to any narrative of self-sufficiency. For investors, the ISCAS keynote represented something tangible to price in: a named architecture, a public timeline, and the imprimatur of Huawei's top chip executive speaking at a major IEEE conference.

Whether LogicFolding and the Tau Scaling Law ultimately deliver 1.4nm-class chips by 2031 remains an open question. The market has already made its short-term bet. The engineering proof is still five years away.