Huawei announces Tao's Law, Jensen Huang says TSMC leads by 10 years!

Jensen Huang is right, but only half right
Tao's Law is indeed a major innovation by Huawei under blockade conditions, but Huang's use of "10 years ahead" to downplay the threat is a clever but not entirely honest PR tactic.
1. What exactly is Tao's Law doing?
He Tingbo's core proposition presented at ISCAS 2026 is very clear: replacing "geometric miniaturization" (shrinking transistor size) with "time miniaturization" (compressing signal propagation delay τ).
The essence of traditional Moore's Law is "making parking spots narrower"—eventually, car doors won't open. TSMC's 3D packaging (CoWoS/SoIC) is about "building two independent garages and adding an elevator." Huawei's LogicFolding approach is **"kitchen directly above the dining room, with a hole in the floor"**—replanning the 3D layout of circuits from the design stage, allowing signals to pass vertically rather than traverse the plane.
This is not homogeneous competition; it is a different level of technological route:
TSMC = backend integration (stacking after chip fabrication)
Huawei = frontend restructuring (changing architecture logic from the design stage)
2. Jensen Huang's "10 years ahead"—the data is factual, but the implication is misleading
Huang says TSMC is 10 years ahead in 3D packaging, which is correct. CoWoS has been mass-produced since 2016, and technologies like SoIC, InFO have already served top global clients like NVIDIA Blackwell, AMD MI series, Apple Silicon.
But he deliberately avoids several key points:
Huawei is doing the same technology incorrectly. LogicFolding is an internal chip architecture restructuring, not stacking independent chips; the fundamental technology is different. TSMC's technological path could extend indefinitely with uncertainty.
3D packaging itself faces thermal, yield, and cost ceilings; Tao's Law cannot catch up. Huawei has mass-produced 381 chips over 6 years, with Kirin 2026 density reaching 238 million transistors per mm² (close to initial 3nm), aiming for 1.4nm equivalent by 2031—no EUV lithography machine needed.
TSMC's vulnerability lies precisely in geopolitics—controlling advanced process nodes globally by a single company is a systemic risk.
3. The strategic significance of Tao's Law
From an investment and technological strategy perspective, Tao's Law's value is not in "surpassing TSMC," but in:
First, opening a third path. When Moore's Law approaches physical limits, and TSMC's advanced packaging route is blocked by US export controls, Huawei proves there is another way to continuously improve performance without EUV lithography. This signals far more to China's entire semiconductor supply chain than a single technological breakthrough.
Second, redefining the competitive dimension of "advanced process." In the past, chip competition was measured by "nanometers." Tao's Law shifts the competition from spatial (geometric size) to temporal (signal delay), meaning even if Huawei lags 2-3 generations in process technology, it can still narrow the gap in actual system performance.
Third, the mass production of 381 chips is a hard metric. This is not just PPT technology. Huawei claims to have mass-produced 381 chips based on Tao's Law, covering phones, AI, automotive, infrastructure. If Mate 90 (expected to launch in fall) with Kirin 2026 is successfully verified, it will be the first large-scale commercial validation.
4. The real risks
Fairly speaking, Tao's Law also faces severe challenges:
Thermal dissipation: heat dissipation in stacked layers is a physical challenge, and Huawei admits this is a core bottleneck.
Toolchain: Huawei's chief architect admits a complete 3D design toolchain "may still take years."
Manufacturing complexity: logic folding significantly increases post-processing steps; yield and cost are key variables for mass production.
The 2031 goal of 1.4nm equivalent—whether this can be achieved depends on the speed of solving these issues.