Sakhanda Wire
NVDA $154.00 -1.37% MSFT $506.28 -2.26% GOOGL $280.87 -1.00% META $620.34 -0.97% AMZN $256.88 -0.34%
← Back to the news

Research links Chinese AI models to restricted Nvidia chips

A new analysis claims that leading open-weight AI models developed by Chinese companies show signs of training on Nvidia chips restricted by the United States. Will DePue, a former OpenAI researcher who leads Unemployment Labs, examined the weights of models released by DeepSeek, Moonshot AI, MiniMax and Ant Group. He reported chip-related fingerprints in Moonshot’s Kimi’s K3 releases and DeepSeek’s V4 models, including indications associated with Blackwell GPUs.

The findings do not establish that the models were trained exclusively on Nvidia hardware or that export laws were broken. Chinese companies may have used overseas data centers, including facilities in Malaysia. Epoch AI estimates that roughly two million advanced chips entered China in 2024 and 2025 through efforts to conceal their destination, while 275,000 Nvidia AI accelerators were imported to Malaysia during the same period.

The case has renewed debate over controls based on who uses chips, rather than only ownership or location. Megaspeed and its subsidiary Speedmatrix, led by Alice Huang, have faced US and Singaporean investigations over alleged chip shipments. The article also reports that about 600,000 H200 chips are stored in Hong Kong, though the reason they have not been imported into China remains disputed.

Written by Sakhanda Wire AI from reporting by TechCrunch. Facts, quotes and full context are in the original article below.

Originally published by TechCrunch on

Read the original on TechCrunch ↗

The reporting and the image belong to TechCrunch and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news