Huawei scientist: Nvidia will soon hit the same chip wall
▼ Summary
– Huawei’s top scientist warns Nvidia will soon hit the same physical performance limit that Huawei claims is reshaping the chip industry.
– Huawei’s Tau Scaling Law measures progress by signal travel speed rather than transistor shrinkage, framing its approach as a survival strategy under US sanctions.
– Huawei’s LogicFolding design stacks circuitry vertically to boost density, but faces unresolved issues with heat, design software, and manufacturing yield.
– Nvidia’s Jensen Huang calls the folding approach a breakthrough but not a threat, while US export controls have pushed China toward homegrown chips, widening the divergence.
– Huawei’s argument is a strategic bet that physics will flatten performance gains over time, potentially eroding Nvidia’s lead and reshaping the industry’s future.
Nvidia is approaching the same physical barrier in chip performance that Huawei has long argued will redefine the industry, according to one of the Chinese firm’s top scientists, as reported by Bloomberg. This is less a boast and more a calculated wager on the future direction of technology. If every manufacturer is approaching a hard ceiling on how much faster silicon can operate, then Nvidia’s dominant position becomes far less secure than its market valuation suggests.
The argument stems from Huawei’s own redefinition of progress. He Tingbo, the company’s chip chief often referred to as its “chip queen,” introduced a Tau Scaling Law that gauges advancement by the speed at which signals travel across a circuit, rather than by the traditional metric of shrinking transistor size. This reframing is a matter of survival under sanctions. Deprived of access to the most advanced lithography equipment, Huawei has a vested interest in demonstrating that the old path of ever-smaller transistors is nearing exhaustion for everyone, not just for those locked out of the best tools.
Their alternative is to build vertically. A design called LogicFolding stacks circuitry in layers, which Huawei claims sharply boosts transistor density and achieves performance comparable to advanced nodes without the machinery it cannot acquire. The obstacles, however, are significant. Vertical stacking concentrates heat, strains existing design software, and introduces serious questions about manufacturing yield, none of which Huawei has proven it can resolve at commercial scale.
Nvidia is not accepting the premise. Jensen Huang has dubbed the folding method a breakthrough but insists it poses no threat, while also warning that a Chinese champion powered by Huawei chips would be a “horrible outcome” for American technology. This warning lands in a market heavily shaped by policy. US export controls have pushed China’s chip sector away from Nvidia GPUs and toward custom ASICs and domestic components, accelerating the very divergence Huawei describes.
The gap remains wide for now. American assessments place Nvidia years ahead of Huawei, with a performance multiple several times larger, a lead the physical-limit argument aims to chip away at gradually rather than erase overnight. There is a narrower version of the same claim. Gains in AI chips increasingly depend on memory bandwidth as much as raw computing power, and Huawei has been working furiously to develop its own high-bandwidth memory to close that specific gap.
The messaging is intentional. Beijing recently featured Huawei’s secretive chip lab on national television, and executives have leaned into the narrative that Chinese engineering can find a route around American controls. The stakes are immense for both sides. China represents one of Nvidia’s largest potential markets, and Huang has acknowledged his company has largely surrendered it to Huawei under the weight of export restrictions.
Huawei’s assertions deserve as much skepticism as attention. The company has every reason to promote a future where its limitations become universal, and its most ambitious projections stretch to 2031, far enough out to resist easy verification today. Whether that wall arrives on Huawei’s timeline is the open question. If Nvidia continues to extract performance through packaging, software, and memory innovations, the ceiling Huawei is counting on could remain comfortably out of reach.
The timing here is no coincidence. Framing a shared limit is a way to reset expectations while Huawei trails, converting a gap it cannot yet close into a race it claims nobody can keep winning. At its core, this dispute is about who gets to write the next chapter of computing. If gains from shrinking transistors truly are fading, the advantage shifts to whoever can design most creatively around the constraint. The warning is, ultimately, a wager on time. Huawei is betting that physics eventually levels the playing field, and when it does, the company that spent its sanctions years building an alternative route will be the one still moving forward.
(Source: The Next Web)
