Jun Young-hyun (center), vice chairman and CEO of Samsung Electronics, speaks to reporters after meeting with Nvidia CEO Jensen Huang at the guest house of the Shilla Hotel in Jung-gu, Seoul, on the 8th. Joint Press Corps
Samsung Electronics is moving beyond supplying high-bandwidth memory (HBM) for Nvidia’s artificial intelligence (AI) accelerators to discussions on securing next-generation chip orders through its foundry’s advanced processes, including the “Drive AGX Thor” chip for autonomous vehicles and the “Grok” language processing unit (LPU).
Jun Young-hyun, vice chairman and CEO of Samsung Electronics, attended Nvidia’s “Korea AI Ecosystem Reception” held at the Guest House of the Shilla Hotel in Jung-gu, Seoul, on Monday. After holding a private business meeting that evening with Nvidia CEO Jensen Huang, Jun told reporters, “We have collaborated with Nvidia for a long time, and I think we had the best conversation (so far).” He added, “We discussed how to handle HBM4 (sixth generation) supply and foundry cooperation in the short term, and we also talked about co-developing through cooperation over the mid-to-long term.”
Samsung Electronics particularly stressed that it needs to fully supply HBM4, which began this year, as well as SOCAMM, a low-power memory module. The company plans to continue long-term cooperation from next year with HBM4E (seventh generation) and HBM5 (eighth generation).
Jun also said the company is forming a deep relationship with Nvidia on expanding foundry cooperation. “We are discussing production of Nvidia’s autonomous driving chips using 4-nanometer (nm, one billionth of a meter) and 8-nanometer processes,” he said. “Grok, another of Nvidia’s accelerators, is also in cooperation (with Samsung’s foundry), and we are discussing cooperation on the next generation as well.”
Since acquiring Grok, an LPU-specialized fabless (chip design) company, for $20 billion (about 30.6 trillion won) last December, Nvidia has been outsourcing production of its “Grok LPU 3rd generation (LP30)” through Samsung’s foundry 4-nanometer process. Jun’s position is that Samsung’s foundry will also handle the LP40. While some had widely predicted that TSMC, which has strengths in advanced packaging processes, would secure the LP40 order, Jun effectively dismissed that rumor.
Meanwhile, regarding a long-term agreement (LTA) with Nvidia on memory supply, Jun said, “Samsung Electronics will work as hard as possible to help Nvidia succeed as its best partner.”