Jensen Huang, CEO of Nvidia, and Hong Kong actor Tony Leung were awarded honorary doctorates on the 23rd at the Hong Kong University of Science and Technology. Jensen Huang stated, “We are working as quickly as possible to approve Samsung’s AI memory chip usage” and noted that Nvidia is considering sourcing both the 8-layer and 12-layer HBM3E products from Samsung. HBM3E is the latest fifth-generation high-bandwidth memory used in Nvidia’s flagship AI accelerator, Hopper (H200), and the new product Blackwell. This reflects Nvidia’s urgency in certifying Samsung Electronics HBM3E ahead of Blackwell’s mass production.
According to Bloomberg on the 23rd (local time), Huang made these comments at the honorary doctorate ceremony held at HKUST, indicating that Samsung Electronics’ HBM3E has not been certified yet. Recently, Huang has mentioned SK Hynix and Micron as HBM3E suppliers for Blackwell in presentations and lectures.
Currently, HBM3, a 4th generation memory supplied by Samsung to Nvidia, is known to be primarily used in the H20. The H20 is a China-specific product with reduced performance, created to circumvent U.S. export restrictions. If the U.S. government under Trump halts exports of low-performance chips like the H20 to China, Samsung, which hasn’t delivered the latest HBM yet, will see a reduction in revenue from older HBMs. Samsung Electronics is thus prioritizing the Nvidia HBM3E supply, having completed an important step in quality testing, suggesting sales could expand in the fourth quarter.
During the event where he received an honorary doctorate alongside Tony Leung and others, Huang stated, “Nvidia has been working in China for the past 25 years. China, including Hong Kong and Shenzhen, boasts impressive design centers and thousands of engineers who have made significant contributions to Nvidia and will continue to do so.” Despite potential heightened U.S. sanctions under a possible second Trump administration, Nvidia has no plans to scale back its China operations. He emphasized the importance of global collaboration in open research and indirectly opposed breaking supply chains into U.S. and China segments.
Huang also praised China’s AI capabilities, acknowledging contributions from institutions like HKUST to global AI progress. Reports from media outlets like SCMP noted his remarks on the Greater Bay Area as a unique region for mechatronics and AI prowess, outpacing other manufacturing hubs like Japan and Germany.
Nevertheless, Nvidia’s partners are beginning to “de-China.” According to sources like DigiTimes, Hong Kong-based PC Partner Group, which assembles Nvidia graphics cards, announced relocating its headquarters to Singapore and plans for a secondary listing on the Singapore stock exchange. Previously handling finance and management from Hong Kong, and manufacturing and R&D from China and Taiwan, the company is moving these functions to Singapore and Southeast Asia. This marks the first significant signal of Nvidia affiliates reducing their mainland China operations.
In response to questions during last week’s Q3 earnings call about a potential second Trump term, Huang expressed support for any decision by the new administration, committing to adhere meticulously to regulations while supporting clients and competing in the market. Despite U.S. export restrictions, Nvidia’s revenue from China, including H20 sales, has increased, though its percentage of total sales dropped from 21.2% a year ago to 15.4%, due to explosive growth in non-China markets.