By Hyunjoo Jin
SEOUL, Jan 26 (Reuters) – Samsung Electronics plans to begin manufacturing of its next-generation high-bandwidth reminiscence (HBM) chips, or HBM4, subsequent month and provide them to Nvidia, an individual acquainted with the matter instructed Reuters on Monday.
Samsung has been making an attempt to meet up with cross-town rival SK Hynix, a major provider for superior reminiscence chips essential for Nvidia’s AI accelerators, after provide delays had hit its earnings and share costs earlier final yr.
Samsung shares climbed 2.2% whereas rival Hynix shares have been down 2.9% in morning commerce.
The individual declined to present particulars comparable to what number of chips it plans to produce to Nvidia.
A Samsung spokesperson declined to remark, whereas Nvidia was not instantly obtainable for remark.
South Korean newspaper Korea Financial Every day reported on Monday that Samsung handed HBM4 qualification exams for Nvidia and AMD and can begin delivery to Nvidia subsequent month, citing chip business sources.
SK Hynix mentioned in October it has accomplished HBM provide talks with main prospects for subsequent yr.
SK Hynix plans to start deploying silicon wafers subsequent month into a brand new fab, M15X, in Cheongju, South Korea, to provide HBM chips, an government on the firm instructed Reuters earlier this month, with out elaborating on whether or not HBM4 can be half of the preliminary manufacturing.
Each Samsung and SK Hynix are set to announce their fourth-quarter earnings on Thursday when they’re anticipated to share particulars of HBM4 orders.
Nvidia CEO Jensen Huang mentioned early this month that the corporate’s next-generation chips, the Vera Rubin platform, is in “full manufacturing,” because the U.S. firm prepares to launch the chips, to be paired with HBM4 chips, later this yr.
(Reporting by Hyunjoo Jin; Modifying by Jamie Freed and Shri Navaratnam)
