How China gets the Nvidia chips it needs to develop AI
Sign up now: Get insights on Asia's fast-moving developments
Washington bans the export of Nvidia’s top-end Blackwell and Rubin GPUs to China, citing security concerns.
PHOTO: REUTERS
- China still gets Nvidia AI chips despite US export controls, through smuggling and data centres in third countries such as Malaysia and Thailand.
- Nvidia’s top Blackwell and Rubin chips are banned in China, while the eased H200 still faces approval hurdles and weak demand from Beijing.
- Chinese chips are improving fast, with Huawei leading the race, but they still need more time, engineering, power and state support to match Nvidia.
AI generated
TOKYO – Nvidia chief executive Jensen Huang has admitted to having an affair, reportedly with a Chinese executive at a Singapore-based company that the US suspects of smuggling artificial intelligence microchips into China.
The scandal has focused attention on how China, a major AI power in itself, obtains the cutting-edge chips despite US export restrictions.
Why are Nvidia chips so important?
Nvidia designs advanced components originally created to handle 3D visuals in video games, called graphics processing units (GPUs).
These have the power to crunch colossal amounts of digital data, so they are also used at scale to train and run AI.
ChatGPT’s 2022 launch showcased the potential of AI, sparking an investment gold rush as governments and tech giants raced to develop the technology further.
With chip orders pouring in, Nvidia’s market valuation has soared to US$4.5 trillion (S$5.75 trillion) – currently on par with Apple as the world’s two most valuable companies.
Which chips does the US restrict?
Washington bans the export of Nvidia’s top-end Blackwell and Rubin GPUs to China, citing security concerns as advanced AI has military applications.
The policy also allows the US’ top AI labs to maintain their global lead.
Huang has lobbied the US government to allow Nvidia chip sales to China, warning that the restrictions are fuelling the rise of competitive home-grown alternatives.
In 2026, Washington eased controls on Nvidia’s previously banned – and less powerful – H200 chip, with the US taking a 25% cut of sales.
But US conditions have made it difficult for shipments to be approved, with Beijing also reportedly reluctant to buy H200s as its own chip industry gains momentum.
How does China get its chips?
Recent launches of high-performing models from Chinese start-ups, including Moonshot AI, Z.ai and DeepSeek, have turned up the heat on US rivals.
“Chinese vendors are still using cutting-edge Nvidia chips to train their models”, despite the controls, Lian Jye Su, a chief analyst at Omdia, told AFP.
Chips may be smuggled into the country or accessed through a loophole that allows Chinese firms to rent out computing power at data centres stuffed with Nvidia chips in a third nation, often South-east Asian.
“US export controls restrict where a chip is shipped, not who uses it,” explained Stephen Wu, founder of the Carthage Capital investment fund.
That means Nvidia GPUs can sit in Malaysia’s Johor region, for example, “in full compliance while a customer in Shenzhen logs in”, Wu told AFP.
White House adviser Michael Kratsios said this summer that Moonshot AI had accessed high-end Nvidia chips in Thailand.
Washington has also accused Chinese AI labs of training systems to mimic the capabilities of US models.
Then there is the big question of smuggling.
Tech outlet The Information said in December 2025 that DeepSeek developed its new model using thousands of chips dismantled overseas and brought into China.
DeepSeek did not respond to a comment request at the time. Nvidia told The Information that it had not seen evidence of such smuggling, which “seems far-fetched”.
How about Chinese chips?
“Nvidia chips still make top-level AI work faster and easier, but they are no longer the only path,” said Poe Zhao, founder of analysis publication Hello China Tech.
“The open question is cost”, as “training on Chinese chips usually takes more time and more engineering”.
Huawei is China’s top Nvidia competitor, although it lags behind the US titan, and DeepSeek will reportedly train its next major AI model on Huawei chips.
“Chinese-made chips are improving fast”, Su said, with more collaborations likely to come.
Huawei said in September that its next-generation AI chip would be ready in the first quarter of 2027, three quarters earlier than planned.
To boost the power of Chinese-made chips, AI companies need to use more of them, Wu said.
“They pay for it in power and cost – which they can afford, because Chinese electricity is cheap and the chips are state-subsidised.” AFP