\u00a0Jensen Huang\u00a0, the co-founder and CEO of \u00a0Nvidia\u00a0, has recently made headlines with his visit to China, where he aimed to advocate for his company’s interests amidst increasing restrictions from the U.S. administration. At the end of last week, Huang met with \u00a0Donald Trump\u00a0 to discuss ways to navigate these limitations, particularly concerning Nvidia’s operations in a country where it has significant revenue streams. In the last fiscal year, which ended on January 26, 2025, China accounted for approximately \u00a013%\u00a0 of Nvidia’s total revenue, translating to about \u00a0$17 billion\u00a0\u2014making it the company’s third-largest market, behind the \u00a0U.S.\u00a0 and \u00a0Taiwan\u00a0.<\/p>\n
<\/p>\n
Huang is evidently committed to preserving Nvidia’s foothold in China, as showcased by his recent remarks at the \u00a0International Exhibition of the China Supply Chain\u00a0 in Beijing. His statement carries weight, especially given the current geopolitical climate surrounding technology companies.<\/p>\n
<\/p>\n
During his speech, Huang praised \u00a0China’s open artificial intelligence models\u00a0, specifically mentioning \u00a0Deepseek\u00a0, \u00a0Qwen\u00a0, and \u00a0Kimi\u00a0. He stated, \u201cThese are the best reasoning models in the world today. They are very advanced.\u201d This endorsement wasn\u2019t casual; it comes at a crucial time for Nvidia, showcasing his strategic approach to maintaining a presence in China’s booming tech industry. Huang’s remarks resonate with several reports that highlight the competitiveness of these \u00a0Chinese AI models\u00a0, making his praise both timely and significant.<\/p>\n
Huang\u2019s commendation of these models is crucial at a transformation point for Nvidia. The company has long relied on \u00a0Chinese partnerships\u00a0 for the sale of its \u00a0GPUs\u00a0. His recent talks with Trump have reportedly led to promising changes; Nvidia appears poised to resume selling its \u00a0H20 GPU\u00a0 to Chinese clients shortly. Despite having some limitations compared to its high-performance counterparts, the H20 chip has received a positive initial reception, with sales increasing by \u00a050% quarter-over-quarter\u00a0 since its launch in mid-2024.<\/p>\n
<\/p>\n
The approval of the H20 chip’s sale in China initially seemed promising, but complications arose due to new restrictions from the U.S. Department of Commerce.<\/p>\n<\/p><\/div>\n<\/div>\n
However, circumstances have not been entirely favorable for Nvidia. Just months after the promising news, the U.S. Department of Commerce imposed renewed restrictions on the H20’s export, significantly affecting its access to Chinese markets. With impending competition, Huang’s situation becomes even more precarious, particularly against the backdrop of \u00a0Huawei\u2019s\u00a0 resurgence in the semiconductor space.<\/p>\n
<\/p>\n
Mere days after the latest sanctions were implemented, Huawei announced its \u00a0Ascend 920 GPU\u00a0\u2014a strategic move that aims to fulfill the demand for AI capabilities left unaddressed by Nvidia in the Chinese market. The \u00a0Ascend 920\u00a0 is designed utilizing \u00a06nm integration technology\u00a0, a collaborative effort between Huawei and \u00a0SMIC\u00a0 (Semiconductor Manufacturing International Corporation). While detailed specifications remain under wraps, expectations are high that this new chip will leverage advanced techniques like \u00a0Multiple Patterning\u00a0 and promises an impressive \u00a04 TB\/s transfer speed\u00a0 for the memory subsystem, thanks to the incorporation of \u00a0HBM3 chips\u00a0.<\/p>\n
<\/p>\n
Image | Nvidia<\/a><\/p>\n