Nvidia develops location tracking for AI chips

Nvidia is reportedly testing new tracking software that could allow for the monitoring of the approximate location of its AI chips amid ongoing rumors of chip smuggling. This development comes as the company faces scrutiny over its supply chain and the potential illegal export of its products, particularly to China.

Deepseek, a Chinese AI startup, is reportedly using thousands of smuggled Nvidia chips for training its next major AI model, raising concerns about compliance with international trade regulations. This information comes from sources cited by The Information, highlighting the lengths to which the company is going to enhance its technological capabilities.

President Trump has announced a significant policy shift allowing Nvidia to sell its advanced H200 AI chips to approved customers in China, effectively lifting a previous export ban imposed due to national security concerns. This decision includes a 25% surcharge on sales, which experts have criticized as illogical given the context of the deal.

Starcloud, a startup backed by Nvidia, has successfully launched a satellite equipped with an Nvidia H100 chip, which is now operational and querying responses from Google's AI model, Gemma. This marks a significant milestone as it represents the first instance of training an AI model from space, showcasing the potential of orbital data centers in AI development.

Nvidia has stated that it has not observed any substantiation or received tips regarding chip smuggling through data centers outside of China, following a report by The Information about the Chinese AI startup DeepSeek's use of banned Nvidia chips for its AI models.

DeepSeek, a Chinese AI startup, is reportedly developing its next AI model using several thousand Nvidia Blackwell chips that were smuggled into China through third-party countries. This development raises concerns regarding compliance with international trade regulations and the implications of using restricted technology.

Google’s Tensor Processing Units (TPUs), specifically the TPUv7, are emerging as a formidable alternative to Nvidia’s GPUs, which have dominated AI advancements for over a decade. This shift is exemplified by the training of cutting-edge models like Google’s Gemini 3 and Anthropic’s Claude 4.5 Opus on TPUs rather than Nvidia hardware.

China has added AI chips from domestic companies, including Huawei and Cambricon, to its government-approved supplier list for the first time, coinciding with the Trump administration's decision to allow Nvidia to export its H200 AI chips to China. This marks a significant shift in China's approach to AI technology procurement amid ongoing geopolitical tensions.