华为的「韬定律」，是创新还是噱头？｜ Bonus

Summary

This What’s Next｜科技早知道 bonus episode has Zhang Haijun / 张海军 explain Huawei’s Tau Law as a shift from process-node and transistor-size language toward system-level time reduction. The episode is cautiously positive: it rejects the view that Tau Law is pure marketing, but it also says the most ambitious part, Cell-to-Cell Logic Stacking, is still unproven in shipped products. Its main addition to the wiki is the toolchain boundary: true cell-level logic folding would require new [[ElectronicDesignAutomation|EDA]] flows, packaging capability, and system engineering, not just a new slogan.

Key Claims

• Tau Law is framed as replacing a pure nanometer-node metric with time, delay, and system-task completion speed as the practical measure of semiconductor progress.
• Zhang Haijun / 张海军 says the idea is system-level rather than a single device, process, or packaging technique; gate-level, chip-level, and system-level changes can all count if they reduce time.
• The episode distinguishes ordinary die-to-die Semiconductor 3D Stacking from Cell-to-Cell Logic Stacking, where logic-cell placement and vertical connection would be considered from the start of chip design.
• Cell-to-Cell Logic Stacking is treated as the most distinctive technical idea in the episode, but Zhang says current public evidence does not show a real product using it.
• The claim that 381 Huawei chips were designed and produced “based on Tau Law” is interpreted broadly: those chips may follow the lower-time design methodology without necessarily using cell-to-cell stacking.
• [[ElectronicDesignAutomation|EDA]] is a central bottleneck because cell-level folding would need new design tools rather than only back-end packaging changes.
• Advanced Packaging still matters after design because three-dimensional logic routes raise process, connection, and manufacturing requirements.
• The episode says 3D stacking itself is not unique to Huawei; TSMC CoWoS, AMD 3D V-Cache, Nvidia super-node designs, and Google TPU interconnect improvements are used as comparison points.
• The source treats the reported 2031 “equivalent 1.4 nm” target as a system-performance comparison rather than literal physical line width.
• Chinese constraint conditions make an alternate route more strategically attractive for Huawei, but the source still requires later proof through tools, chips, cost, power, and reliable production.

Key Quotes

“不是噱头” — Zhang’s short answer to whether Tau Law is only marketing.

“Cell-to-Cell 的堆叠” — the episode’s distinction between Huawei’s ambitious logic-folding idea and more common chip-to-chip stacking.

“目前全球还没有哪个产品真正使用这种技术” — Zhang’s caveat about public product evidence for cell-level stacking.

Connections

• What’s Next｜科技早知道 — show context for the bonus episode.
• Zhang Haijun / 张海军 — guest and semiconductor engineer explaining the claim.
• Huawei and Tau Law — company and central doctrine under discussion.
• Cell-to-Cell Logic Stacking — source-specific technical caveat around logic folding.
• Semiconductor 3D Stacking and Advanced Packaging — broader packaging and vertical-integration context.
• Electronic Design Automation and Synopsys / 新思科技 — toolchain layer that would have to change for cell-level folding.
• Moore’s Law and Constraint Driven Engineering Strategy — incumbent scaling frame and constraint-driven alternate route.
• TSMC, Nvidia, Google, and TPU — comparison cases for advanced packaging, system architecture, and interconnect improvement.

Contradictions

• No direct contradiction found with the existing Huawei Keji Luandun source. This source supports the same cautious position that Tau Law is more defensible as a system metric than as a literal replacement for Moore’s Law.
• It narrows the existing interpretation by separating die-to-die stacking from Cell-to-Cell Logic Stacking and by warning that “based on Tau Law” does not prove the most ambitious cell-level implementation has shipped.
• It also qualifies public hype around “domestic chip breakthrough” claims: the source sees a plausible technical direction, but not enough product, EDA, yield, cost, or power evidence for a final verdict.