- Huawei’s “chip queen” He Tingbo announced the Tau Law at a Shanghai conference — a chip-stacking technique that builds circuits in two vertical layers to achieve 55% higher density without EUV lithography
- Independent semiconductor analyst Jimmy Goodrich reviewed Huawei’s technical paper and found that TSMC will reach the same performance target in 2028 — and be a full generation beyond it by 2031, when Huawei hopes to arrive; the realistic gap is 6-8 years, not 3
- Huawei’s paper explicitly opens by noting that “access to the most advanced lithography is constrained” — the clearest public admission yet that the company has accepted it cannot break through the EUV export-control barrier on any near-term horizon
- A practical problem undermines the stacking approach: Huawei’s factories produce usable chips only ~20% of the time; stacking requires bonding two chips with extreme precision — doubling the defect-rate challenge
What Happened?
At a semiconductor conference in Shanghai last week, Huawei’s semiconductor chief announced what the company framed as a successor to Moore’s Law: the Tau Law. The technique stacks two layers of circuitry vertically rather than continuing to shrink planar chip dimensions — achieving 55% higher density than Huawei’s previous chip generation. Huawei claimed the approach would deliver cutting-edge performance by 2031. WSJ chief China correspondent Lingling Wei reviewed the technical paper with independent analyst Jimmy Goodrich, whose conclusion upends Huawei’s framing: the paper’s own opening language acknowledges that “access to the most advanced lithography is constrained” — an implicit admission that US-led export controls on ASML’s EUV machines have worked.
Why It Matters?
The Tau Law announcement is designed to be heard by two audiences simultaneously. For Washington, it argues export controls aren’t working — useful ammunition for anyone lobbying to ease restrictions. For Beijing, it justifies turning down the Nvidia H200 chip licenses Trump recently authorized: “We don’t need American technology.” But the paper’s contents contradict both arguments. Goodrich’s analysis reveals TSMC will hit Huawei’s 2031 performance target in 2028 — and be a generation beyond it by the time Huawei arrives — because TSMC applies the same stacking techniques on chips made with EUV, which Huawei cannot access. The 20% yield rate at Huawei’s factories is the practical kill shot: stacking requires precision bonding of two chips; if one in five chips is defective individually, the compound defect rate for stacked pairs is far worse. Huawei is innovating impressively within its constraints — but this is a workaround, not a breakthrough.
What’s Next?
On the same day Huawei made its announcement, the Bureau of Industry and Security issued guidance clarifying that export-control license requirements always applied to offshore operations of Chinese companies — an implicit acknowledgment that firms had been circumventing restrictions through overseas subsidiaries to smuggle advanced chips. The dual development signals that the export-control regime is tightening, not loosening, even as Huawei’s announcement provides political cover for those arguing otherwise. The CXMT memory chip company’s Shanghai listing review clearance (reported separately) suggests China is doubling down on domestic semiconductor investment. Watch for whether the Trump administration uses Huawei’s paper to justify further restricting chip exports, or whether it becomes ammunition in lobbying efforts by chipmakers seeking to restore H200 sales to China.
Source: The Wall Street Journal
