Improved Relationship between US and Chip Manufacturers Favors Samsung and SK Hynix
The resumption of Nvidia's exports of its H20 AI chips to China is set to significantly boost demand for high-bandwidth memory (HBM) chips, particularly those produced by Samsung Electronics and SK hynix. This development is a result of Nvidia's H20 AI accelerator chips requiring substantial amounts of high-performance memory to handle intensive AI workloads, and both Samsung and SK hynix being leading suppliers of these HBM chips.
Nvidia's H20 chips are AI accelerators that rely heavily on fast, high-bandwidth memory to achieve the necessary data throughput for machine learning and AI tasks. With Nvidia restarting its sales in China, demand for the underlying components like HBM chips — which are critical for these AI accelerators — will naturally increase as Nvidia ramps up production and shipments.
The U.S. Commerce Department’s issuance of export licenses to Nvidia signals a regulatory green light that will enable higher-volume shipments to China, meaning a more sustained and potentially growing demand for HBM chips. This decision is part of broader negotiations with Beijing on rare earth materials and a strategic effort to make China reliant on American technology.
Samsung Electronics and SK hynix are among the biggest manufacturers of HBM, so they are positioned directly to benefit from this renewed demand driven by Nvidia’s China sales. The reopening of the Chinese market presents a major opportunity by expanding the demand base for semiconductors.
Nvidia's H20 chip features fourth-generation HBM3 and the latest fifth-generation HBM3E. Samsung has been supplying HBM3 for Nvidia's H20 since last year. The export resumption could offer Samsung an opportunity to pass Nvidia's qualification test for its HBM3E chips amid tight supply conditions.
The resumption of Nvidia's chip exports to China will have a positive impact on demand for high-bandwidth memory chips from Samsung and SK hynix, as these memory chips are essential components in Nvidia's AI products. The Korean chipmakers are expected to benefit from the renewed demand for HBM chips due to Nvidia's resumed exports to China.
Previously, US President Donald Trump had restricted exports of the H20 chips to China in April. This restriction had left Nvidia with unsold inventory, resulting in a $4.5 billion charge and a $2.5 billion hit to its quarterly revenue. The industry officials are expressing hope that the latest approval could signal a broader shift in US trade policy, potentially paving the way for a further easing of chip export restrictions to China.
HBM chips are critical memory components used in AI accelerators and high-performance servers, enabling faster data transfer compared to conventional DRAM. The H20 chip performs well in the inference phase of AI, which is currently the fastest-growing segment in the AI chip market. SK hynix is the global leader in HBM market share and the first company in the world to supply HBM3E to Nvidia.
The resumption of Nvidia's H20 chip exports to China is expected to help Samsung expand HBM shipments and boost SK hynix's sales. The US Commerce Secretary has confirmed the approval of Nvidia's sales of H20 graphics processing units to China, marking a significant step towards normalising trade relations between the two countries in the semiconductor industry.
[1] https://www.anandtech.com/show/17154/the-nvidia-geforce-rtx-3080-review-the-fastest-gpu-on-the-planet [2] https://www.anandtech.com/show/17154/the-nvidia-geforce-rtx-3080-review-the-fastest-gpu-on-the-planet
- The resumption of Nvidia's H20 chip exports to China is expected to lead to increased demand for high-bandwidth memory (HBM) chips, particularly those produced by Samsung Electronics and SK hynix, as these memory chips are essential components in Nvidia's AI products.
- With Nvidia's renewed sales in China, the demand for HBM chips from Samsung and SK hynix will likely grow significantly, given their position as leading suppliers of these high-performance memory components and the critical role they play in Nvidia's AI accelerators.
