Nam Hyun-woo has worked as a staff writer at The Korea Times since 2013, mostly covering business and politics. He currently belongs to the Business Desk where he covers topics such as emerging tech, AI, ICT and Korea's chaebol community. Prior to joining the team, he was the paper's correspondent for the presidential office of Korea during the Yoon Suk Yeol and Moon Jae-in administrations.
Ban on Nvidia graphics card exports to China adds concerns for Samsung
Nvidia CEO Jensen Huang, right, signs a Samsung Electronics' GDDR7 memory chip during his visit to the company's booth at GTC 2025 in San Jose, Calif., March 20. Yonhap
By Nam Hyun-woo
Nvidia’s reported halt of flagship GeForce graphics card exports to China is raising concerns about the profitability of Samsung Electronics’ memory business, adding to fears of a slowdown amid the ongoing trade war between Washington and Beijing.
U.S. tech outlet Tom’s Hardware recently reported that Nvidia has effectively confirmed its RTX 5090D GPUs cannot be ordered in Q2, with all existing and pending orders temporarily canceled. The RTX 5090D is a China-specific variant of Nvidia’s flagship GPU.
Nvidia has not commented, but multiple tech media reports suggest the 5090D still violates new U.S. restrictions, as its GB202 Blackwell chip remains too powerful despite the downgrade.
The ban has sparked concerns about ripple effects on Samsung Electronics, which supplies Graphics DDR7 (GDDR7) memory chips for the RTX 50 series.
Nvidia CEO Jensen Huang holds a Blackwell GeForce RTX 50 Series GPU as he delivers a keynote address at CES 2025 in Las Vegas, Jan. 6. AFP-Yonhap
Samsung is the primary memory supplier for the RTX 50 series, although Nvidia also sources chips from SK hynix and Micron. Nvidia confirmed in January that the initial batch of RTX 50 series cards would exclusively use Samsung’s GDDR7 memory chips.
In March, Nvidia CEO Jensen Huang visited Samsung’s booth at GTC 2025 in San Jose and wrote on a Samsung’s GDDR7 chip: “SAMSUNG,” “GDDR7 ROCKS” and “RTX ON!”
The gesture underscored the close collaboration between the two companies, highlighting Samsung's pivotal role in Nvidia's next-generation GPU lineup.
However, the suspension comes as Samsung Electronics is already grappling with U.S. restrictions on China-bound shipments of artificial intelligence (AI) accelerators, particularly Nvidia’s H20, which uses Samsung’s high-bandwidth memory (HBM) chips designed for AI applications.
Nvidia has warned that U.S. export restrictions on its H20 chips to China could result in losses of up to $5.5 billion in the first quarter of 2025 alone. Samsung supplies its fourth-generation HBM3 memory for the H20.
Samsung is already trailing SK hynix in supplying fifth-generation HBM3E and sixth-generation HBM4 to leading AI accelerator manufactures, while its HBM3 shipments to China were blocked following U.S. regulation changes earlier this year.
Reuters on Friday reported that Nvidia plans to release a downgraded version of the H20 for China in the next two months, and industry officials expect that the downgraded version will likely use GDDR instead of HBM to avoid U.S. restrictions.
As a result, GDDR sales were initially expected to help offset potential losses from restricted HBM exports. However, with even graphics cards now facing export bans, concerns are growing that GDDR may also be vulnerable to U.S. regulatory risks.
"If the sales ban on Nvidia’s RTX 5090 in China is officially confirmed, it could result in some damages to Samsung's GDDR7 sales. However, as seen with the U.S. export regulations targeting Huawei in the late 2010s, companies always explore alternative channels to make up losses," an industry source said.
"Though current U.S. sanctions on HBM and GPUs are not likely to extend to GDDR, the high level of uncertainty means that the industry steel needs close monitoring."
During an earnings call in January, Samsung Electronics said it is ramping up the manufacturing of high-value-added products such as HBM and GDDR7, adding that it plans to sharply reduce the revenue share of legacy DRAM products from the low-30 percent range in 2023 to single digits this year.