According to Reuters and The Information, the U.S. Department of Commerce is finalizing new rules that would restrict Chinese AI companies from using high-end Nvidia chips to train models remotely via data centers in Singapore, Malaysia, Thailand, and other locations. The new rules could be announced as early as this month.
Worried about further tightening of US chip controls, some Chinese AI companies have contacted overseas data centers to reserve high-end Nvidia chip computing power in advance. A large cloud service provider is internally discussing locking in computing power early, and some domestic gray market merchants have begun hoarding restricted chips.
Megaspeed, a Singaporean company co-founded by Kang Wei and his wife, has become an Nvidia cloud partner and is suspected of shipping servers equipped with Nvidia chips to China. The US and Singapore are investigating its related business, while the company denies violating export controls.
On November 11, Reuters reported that Northwestern Polytechnical University purchased two Nvidia H200 chips, which are restricted by the U.S., through a supplier in October for 874,700 yuan and leased computing power, exposing loopholes in export controls.
A residential committee in a Shenzhen community suggested in a group chat that residents go out for trips from November 17 to 19, stating that those who remain at home will be visited by party and government personnel during specific periods. Empty apartments will have electronic seals with chips placed on their doors, which will trigger an alarm upon opening, and residents must be accompanied by staff to return home. The APEC meeting is being held in Shenzhen from November 18 to 19.
On November 9, after Jensen Huang admitted to an extramarital affair, former Global Times editor-in-chief Hu Xijin posted that he has a background in intelligence studies and, to his knowledge, the Chinese intelligence community never specifically uses honey traps, nor does it ever require female intelligence officers to complete missions by sleeping with targets.
A report released by Bloomberg Economics on November 6 stated that if the Democratic Party slows down AI investment and data center construction after a major victory in the U.S. midterm elections, China's growth, which relies on AI hardware exports, will be hit. Net exports contributed nearly 40% of China's economic growth in the third quarter, and chips and computer products accounted for 40% of the year-on-year increase in exports in the first nine months.
A Reuters exclusive report states that about 600,000 NVIDIA H200 AI chips are stored in Hong Kong waiting to be shipped into mainland China, with prices surging due to Beijing's delayed approval for large-scale imports.
According to the Financial Times, as the U.S.-China trade truce continues and high-end AI chips remain a point of friction, Oracle has canceled a $7 billion chip leasing contract with Tencent under pressure from the Trump administration.
According to a report by the Financial Times on November 4, Oracle canceled a five-year, approximately $7 billion chip leasing agreement with Tencent under pressure from the Trump administration. The original agreement allowed Tencent to use about 100,000 advanced AI chips in Southeast Asian data centers. During the same period, senior officials from Guangdong visited Tencent and demanded that it switch to domestic chips.
Jensen Huang recently announced his formal resignation. His involvement in chip sales and his connection to Chinese spy Huang Le have sparked widespread attention and discussion both inside and outside the Great Firewall.
A desktop computer launched by Huawei, equipped with its proprietary chip and operating system, took over five minutes to boot during testing and frequently failed with endless loading icons. Its system compatibility and performance have sparked heated discussion and ridicule among netizens.
In response to foreign media reports alleging that Chinese businesswoman Huang Le contacted Nvidia CEO Jensen Huang to assist in bypassing chip sanctions, Chinese Foreign Ministry spokesperson Guo Jiakun stated that the claims are purely slanderous and smear tactics.
According to an exclusive Reuters report on October 28, ChangXin Memory Technologies (CXMT), China's largest DRAM manufacturer, plans to start limited production of HBM3E high-bandwidth memory chips within the next two months. The yield is only about 20%-30%, far lower than the approximately 70% level of SK Hynix, reflecting the shortage dilemma in China's AI chip industry. Huawei, Cambricon, and others have already raised prices by 20% to 50% due to rising storage costs.
On October 20, Reuters reported, citing people familiar with the matter, that ByteDance, Alibaba and Tencent have collectively ordered more than 3 million Nvidia RTX PRO 5500 graphics cards, with potential orders worth over $35 billion. Because U.S. restrictions limit exports of high-end AI chips to China, the three companies have turned to this workstation graphics card for AI inference, but Nvidia cannot guarantee supply, and some partners may receive only 5%–10% of the quantity they need.