Chey Tae-won, chairman of South Korea's SK Group, said at CES 2025 on Wednesday that the two sides exchanged views on topics such as high bandwidth memory (HBM) and "physical AI," as he described a meeting with Nvidia's chief executive officer (CEO) Jensen Huang during the show.

SK hynix, the chip division of SK Group, is exclusively supplying HBM to Nvidia, and was the first to supply the fifth-generation 8-layer HBM3E product in March last year, and then mass-produced the world's first 12-layer-stacked HBM3E memory in October of the same year.

Chey said that during this time SK hynix's HBM development was slightly slower than Nvidia's, and the other party (Nvidia) asked SK hynix to speed it up, while recently (SK hynix's) R&D speed has begun to catch up, and the two companies are accelerating their HBM development.

"Currently, we are developing at a slightly higher speed than what Nvidia (asked for)," he said. "And while that may change, we are now developing at a comparable pace."

SK hynix is Nvidia's primary HBM supplier, having an industry-leading fifth-generation HBM3E chip.

The South Korean chip giant earlier announced that it had sold out of HBMs scheduled for production in 2025, with Nvidia being the largest customer.

Chey also revealed that he and Huang discussed and confirmed the HBM timeline that was worked out at a working level, and that "this year's supply has been finalized, but I don't remember the exact numbers."

In addition to HBM, Chey also emphasized that SK Group will focus on AI data centers as a major growth driver in the future.