[International] China's Use of NVIDIA AI Chips: Smuggling and Leasing Southeast Asian Data Center Computing Power Become Two Major Avenues
RtiTalk 小編11d ago
NVIDIA CEO Jensen Huang recently admitted to an extramarital affair, reportedly with a Chinese executive from a Singaporean company. The US suspects this company smuggled NVIDIA artificial intelligence chips into China. This scandal has drawn public attention to how China obtains or uses cutting-edge chips under US export restrictions.
Why are NVIDIA chips so important?
According to Agence France-Presse (AFP), NVIDIA designs advanced components called Graphics Processing Units (GPUs), originally developed to process 3D graphics for video games.
These GPUs can process vast amounts of digital data, making them widely used for training and running artificial intelligence (AI).
ChatGPT, launched in 2022, demonstrated the potential of AI and sparked an investment boom, with governments and tech giants worldwide racing to further develop this technology.
As chip orders surged, NVIDIA's market capitalization soared to $4.5 trillion, currently ranking it among the top two most valuable companies globally alongside Apple.
Which chips does the US restrict from exporting to China?
Citing security concerns, Washington prohibits the export of NVIDIA's highest-end GPUs, Blackwell and Rubin, to China, as advanced AI can be used for military purposes.
This policy also allows top US AI labs to maintain their global lead.
Huang had lobbied the US government to allow NVIDIA to sell chips to China, warning that restrictions were fostering the rise of competitive Chinese domestic alternatives.
Washington eased controls on NVIDIA's lower-performance H200 chips this year, charging a 25% fee on related sales.
However, conditions set by the US make chip shipments difficult to approve. Reports suggest that as China's domestic chip industry accelerates, Beijing's willingness to purchase H200s has also decreased.
How does China obtain US chips?
Chinese startups have recently launched high-performance models like Moonshot AI, Z.ai, and DeepSeek, putting pressure on their US competitors.
Lian Jye Su, chief analyst at research firm Omdia, told AFP that despite controls, "Chinese companies are still using NVIDIA's cutting-edge chips to train their models."
Chips may be smuggled into China, or Chinese companies exploit loopholes by leasing computing power from data centers located in third countries equipped with numerous NVIDIA chips. These data centers are often in Southeast Asia.
Stephen Wu, founder of investment fund Carthage Capital, explained, "US export controls restrict where chips go, not who uses them."
He told AFP that, for example, NVIDIA GPUs could be stored in Johor, Malaysia, "fully compliant, while still allowing customers in Shenzhen to log in and use them."
White House advisor Michael Kratsios noted this summer that Moonshot AI had used NVIDIA high-end chips located in Thailand.
Additionally, smuggling is a significant issue. Tech media outlet The Information reported in December last year that DeepSeek used thousands of chips, disassembled overseas and then transported into China, when developing its new models.
How do Chinese chips perform?
Poe Zhao, founder of tech analysis publication Hello China Tech, stated, "NVIDIA chips still allow top AI to run faster and easier, but they are no longer the only way."
He added, however, that "the cost issue remains unresolved," as "training with Chinese chips typically requires more time and engineering resources."
Although Huawei lags behind NVIDIA, it remains NVIDIA's top competitor in the Chinese market. DeepSeek is reportedly set to use Huawei chips to train its next major AI model.
Huawei announced in September that its next-generation AI chip would be ready in the first quarter of next year, three quarters ahead of schedule.
Stephen Wu pointed out that to increase the total computing power of Chinese-made chips, AI companies must use more of them.
He said, "The trade-off is higher electricity consumption and costs, but they can afford it because electricity prices in China are cheap, and the chips are subsidized by the government."
Source Link: https://www.rti.org.tw/news?uid=3&pid=1050631
How does this article make you feel?
0 people reacted