NVDA
New playbook, same fight
Huawei didn’t just show up with another shiny chip slide deck. On Monday, May 25th, the company unveiled a new architecture built to sidestep the assumptions behind U.S. export controls. Instead of relying on the old “shrink the transistor, win the future” model, Huawei is leaning on design tricks like its Tau Scaling Law and LogicFolding. Translation: if the door to advanced lithography is locked, Huawei is trying to build a window.
Why investors should care
That matters because Nvidia’s business model assumes the best AI chips stay the best AI chips. But if Chinese firms can keep improving without needing the exact tools Washington is trying to restrict, the moat gets a little less moat-y. That’s especially awkward for Nvidia after CEO Jensen Huang has already warned that if U.S. companies pull back, Huawei and its CANN ecosystem can fill the gap.
The policy vs. physics duel
This isn’t just a China tech flex. The U.S. Commerce Department also issued guidance saying using Huawei’s Ascend AI chips anywhere in the world would violate export controls. So you’ve got Washington tightening the screws while Huawei is basically saying, “Cool story, we’ll take a different route.”
- Huawei says its new chips could reach 1.4nm-equivalent density by 2031.
- Kirin chips due in late 2026 are set to be the first to use LogicFolding.
- Huawei claims it has already designed and mass produced 381 chips using the Tau Scaling Law over the past six years.
Big picture
This is the kind of headline that reminds you semiconductors are part physics lab, part geopolitical cage match. Nvidia still has the performance crown, but the race isn’t just about raw speed anymore—it’s about who controls the software stack, the standards, and the markets that matter most.