China accounts for about 30% of Samsung's HBM chip revenue in the first half of 2024.

Last week, it was reported that Washington was preparing to announce a new export control package this month that would impose stricter restrictions on shipments of semiconductor industrial goods to China, such as HBM chips.

A representative of the US Department of Commerce declined to comment on the new restrictions, but said the agency constantly assesses potential threats and updates export controls to “protect national security and the technology ecosystem.”

images 4.jpg
China is ramping up its stockpiling of Samsung's HBM2 chips. Photo: Extreme Tech

HBM chips are a key component in the development of advanced processors used for AI tasks. Currently, only three major chip manufacturers produce HBM chips for Nvidia GPUs: SK Hynix, Samsung, and Micron Technology.

China's chip demand is mainly focused on the HBM2E model, which is two generations behind the most advanced version HBM3E.

“China’s demand for Samsung HBM has been unusually high, as US AI companies have booked up other brands’ production capacity,” said Nori Chiou, investment director at Singapore-based White Oak Capital Partners.

Baidu, Huawei, Tencent, and chip design startup Haawking are all said to have placed orders for Samsung HBM chips, although it is not possible to estimate the volume or value of these orders.

Chinese companies have made some progress in HBM production, with Huawei and memory chip maker CXMT focusing on developing HBM2 chips, three generations behind the HBM3E model.

While Micron has limited sales of HBM products to China since last year, SK Hynix has focused more on the advanced HBM segment. Therefore, Samsung is the most likely manufacturer for mainland customers.

(According to Bloomberg, Yahoo Tech)

Profits increase by more than 1,400%, Samsung reduces supply of conventional memory chips Samsung Electronics Group said that in the second half of 2024, the company will limit the supply of conventional memory chips, to focus resources on expanding capacity to meet demand for artificial intelligence (AI) chips.