China banned Nvidia 5090D V2 while CEO Jensen Huang was in town, report claims — move comes as Beijing pushes its AI tech companies to use homegrown chips (www.tomshardware.com)
from sp3ctr4l@lemmy.dbzer0.com to world@lemmy.world on 22 May 15:46
https://lemmy.dbzer0.com/post/69287141

China reportedly banned Nvidia’s RTX 5090D V2, an export-friendly version of its top-end RTX 5090 GPU, while CEO Jensen Huang was visiting the country as part of President Donald Trump’s state visit last week, the Financial Times reports. The report claims the chip has been added to a list of banned goods at Chinese customs. Huang was a late addition to Trump’s entourage last week, boarding Air Force One in Alaska after he was initially not included on the President’s guest list.

…

The Nvidia RTX 5090D V2 is a version of the company’s top-of-the-line gaming GPU designed to comply with U.S. export controls. This graphics card, which has less VRAM and lower bandwidth compared to the vanilla 5090, is designed for Chinese gamers and 3D artists. However, AI developers have also been taking advantage of this relatively powerful GPU, especially as they’ve been cut off from Nvidia’s more potent Blackwell-powered AI GPUs.

…

Full article here, clicky linky.

---

So basically, China banned an Nvidia GPU that while yes is technically designed for ‘gamers’, is also widely used as a kind of budget alternative to the higher end Nvidia GPUs geared explicitly toward AI/LLM compute and training.

China did this while Huang was in the country, as part of the Trump visit delegation.

Huang is on the Council of Advisors on Science and Technology of the Trump administration.

(source)

This arguably represents a significant setback for Nvidia, who would really like to have as much physical market presence in China as possible, and the Trump admin’s ‘AI policy goals’, simultaneously.

It also could be taken to indicate that China is quite confident in its approach of producing largely open source LLM models via its own homegrown domestic industry of chips geared toward its own methods of LLM training and operation.

#world