Samsung Electronics (005930) plans to more than triple its high-bandwidth memory (HBM) production this year compared to last year, allocating over half of output to its latest product, HBM4.
"We are ramping up steeply and have no major issues with production," Hwang Sang-jun, Executive Vice President of Memory Development at Samsung Electronics, told reporters on January 16 (local time) at Nvidia's annual developer conference GTC 2026, which opened in San Jose, California.
"Our goal is to have HBM4 account for more than half of total HBM production," Hwang said. "When supply is somewhat tight, concentrating on premium products is better for the industry overall."
Regarding the global memory shortage, he explained that the company must balance supply volumes by considering partners requiring strategic allocation and those needing mass-produced products.
Samsung also announced it will apply 2-nanometer base die processing—more advanced than the current 4-nanometer—to HBM5, which follows the newly unveiled HBM4E. "There are cost burdens, but utilizing cutting-edge processes is inevitable to meet HBM's product direction and concept," Hwang said.
Production of Nvidia's new inference chip "Grok3," unveiled the same day, will take place at Samsung Electronics' Pyeongtaek campus. Hwang noted that Grok was a Samsung Foundry customer even before signing a licensing agreement with Nvidia. While there were concerns that Grok's deal with Nvidia might end its relationship with Samsung Foundry, Nvidia was satisfied with the products and decided to maintain the production arrangement.
