According to a report from Ctee, Nvidia isn’t just tweaking existing silicon to comply with Washington’s performance caps—it’s preparing to design new chips within China, relying on the country’s supply chain for everything from packaging to HBM memory and process nodes.

The stakes are massive. Nvidia’s AI hardware has been flying off the shelves in China, with revenues far outstripping official figures thanks to layered trades routed through places like Singapore and Malaysia. The H20 chip was shaping up to be a local hit until the White House brought the hammer down, and with Huawei muscling in with its 910C GPUs, Nvidia’s window is closing fast.

Insiders say the plan includes establishing an R&D hub in China fully embedded in the local AI ecosystem, tailored to work with the big model frameworks used by domestic developers. It’s a strategic gamble: by rooting development locally, Nvidia can curry favour with regulators and keep the yuan rolling in.

Ctee’s report hints that Nvidia may pivot its global R&D footprint, originally planned to be based in Taiwan, to align more closely with this China-focused direction.

Nvidia’s dominance in software, particularly CUDA, still gives it an edge over Huawei and the local startups scrambling to fill the void. Even if new China-bound chips are neutered compared to their global siblings, the ability to run CUDA and maintain compatibility with Nvidia’s toolchain could be enough to retain market traction.

But the move’s not without risk. Relying “100% on China” for development puts Nvidia on shaky geopolitical ground, particularly as tensions between Washington and Beijing show no signs of cooling. If anything, it raises eyebrows about whether the US will turn its ire on Nvidia next for trying to skirt export controls through localisation.

With an estimated $12 billion at stake in China, even under current restrictions, Nvidia sees no other option but to bet on a homegrown strategy.