Huawei scientist says Nvidia will soon hit the same chip wall

One of the Chinese company’s top scientists has warned that Nvidia will soon run into the same physical limit on chip performance that Huawei says is reshaping the whole industry, according to Bloomberg.

The argument is less a boast than a bet on where the technology is heading. If every chipmaker is nearing a hard ceiling on how much faster silicon can get, then Nvidia’s commanding lead becomes less durable than its share price assumes.

The claim builds on Huawei’s own reframing of progress. Its chip chief He Tingbo, sometimes called the company’s ‘chip queen’, unveiled a Tau Scaling Law that measures advances by how fast signals travel across a circuit rather than by shrinking transistors.

The point of that reframing is survival under sanctions. Cut off from the most advanced lithography, Huawei has an interest in arguing that the old road, ever-smaller transistors, is running out for everyone, not only for the companies barred from the best tools.

Its answer is to build upward. A design Huawei calls LogicFolding stacks circuitry into vertical layers, which it says lifts transistor density sharply and reaches advanced-node performance without the equipment it cannot buy.

The hurdles are real. Stacking chips this way concentrates heat, strains design software, and raises hard questions about manufacturing yield, none of which Huawei has shown it has solved at scale.

Nvidia is not conceding the premise. Jensen Huang has called the folding approach a breakthrough while insisting it is no threat, and he has warned that a Chinese champion running on Huawei chips would be a ‘horrible outcome’ for American technology.

The warning lands in a market bent by policy. US export controls have pushed China’s chip industry away from Nvidia’s GPUs and toward custom ASICs and homegrown parts, accelerating exactly the divergence Huawei is describing.

For now, the distance is still wide. American assessments put Nvidia years and several multiples of performance ahead of Huawei, a lead the physical-limit argument is meant to erode over time rather than overnight.

There is a narrower version of the same claim. Gains in AI chips increasingly hinge on memory bandwidth as much as raw compute, and Huawei has been racing to build its own high-bandwidth memory to close that specific gap.

The messaging is deliberate. Beijing recently put Huawei’s secret chip lab on national television, and its executives have leaned into the narrative that Chinese engineering can route around American controls.

The stakes are enormous for both sides. China is one of Nvidia’s largest potential markets, and Huang has said his company has largely conceded it to Huawei under the weight of export controls.

Huawei’s claims deserve scrutiny as much as attention. The company has every incentive to talk up a future in which its constraints become everyone’s, and its boldest projections run to 2031, far enough out to be hard to test today.

Whether the wall arrives on Huawei’s schedule is the open question. If Nvidia keeps finding performance through packaging, software, and memory, the ceiling Huawei is counting on could stay comfortably out of reach.

The timing is not accidental, either. Framing a shared limit is a way to reset expectations while Huawei is behind, turning a gap it cannot yet close into a race it says nobody can keep winning.

The dispute is ultimately about who writes the next chapter of computing. If the gains from shrinking transistors really are running out, the advantage shifts to whoever designs most cleverly around the limit.

The warning is, in the end, a wager on time. Huawei is betting that physics eventually flattens the field, and that when it does, the company that spent the sanctions years building an alternative path will be the one still climbing.