By Hyunjoo Jin and Joyce Lee
SEOUL (Reuters) - Nvidia (NASDAQ:NVDA) CEO Jensen Huang had asked memory chip maker SK Hynix to bring forward by six months the supply of its next-generation high-bandwidth memory chips called HBM4, SK Group Chairman Chey Tae-won said on Monday.
South Korea's SK Hynix said in October that it aimed to supply the chips to customers in the second half of 2025. An SK Hynix spokesperson said on Monday that this timeline was faster than an initial target without elaborating further.
Huang's request for faster delivery highlights demand for higher capacity, energy-efficient chips used in Nvidia's graphic processing units to develop AI technology. Nvidia commands more than 80% of the global AI chip market.
SK Hynix has been leading a global race to meet an explosive demand for HBM chips that aid in the processing of vast amounts of data to train AI technology and are crucial for Nvidia's chipsets. But it is facing increasing competition from rivals like Samsung Electronics (KS:005930) and Micron (NASDAQ:MU).
SK Hynix, which plans to supply the latest 12-layer HBM3E to an unidentified customer this year, also plans to ship samples of more advanced 16-layer HBM3E early next year, its CEO, Kwak Noh-Jung, said at the SK AI SUMMIT 2024 in Seoul.
Samsung said last week that it was making progress on a supply deal with an unidentified major customer after suffering from delays, adding that it was in talks with major customers to produce "improved" HBM3E products in the first half of next year. Samsung also plans to produce the next-generation HBM4 products in the second half of next year.
Shares of SK Hynix jumped 5.1% while Samsung shares firmed 1%. The broader market was up 1.6%. (This story has been corrected to say '16-layer HBM3E', not '18-layer HBM3E', in paragraph 5)