AI collaboration deepens between SK Hynix and Nvidia

A meeting between the heads of SK Hynix’s parent company and Nvidia on Wednesday indicated the two will forge closer ties in AI hardware. Chey Tae Won, speaking during Nvidia’s CES press conference in Las Vegas, said SK Hynix is ramping up its chip development to meet Nvidia’s growing demands.

According to reports, Nvidia’s co-founder Jensen Huang and SK Group Chairman Chey Tae Won discussed ways to expand AI applications. Both agreed to increase SK Hynix’s high-bandwidth memory (HBM) chips, which are the key elements of Nvidia’s AI accelerators.

Shares of SK Hynix jumped over 5% in Seoul following the news, which has fueled substantial gains for the company so far this year. The Korean memory chip maker was Nvidia’s leading supplier during the AI boom sparked by ChatGPT, leaving its rival, Samsung Electronics behind.

SK Hynix’s HBM3E chips propelled it to the forefront of memory solutions for AI and spurred successive advanced product training.

How Korea gained an advantage in AI manufacturing 

Huang said AI systems that can “reason, plan, and act” could revolutionize all industries worth $50 trillion.

Chey added that South Korea is crucial as a partner in this transformation because of its manufacturing strengths.

Both SK Hynix and Nvidia agreed to explore further collaboration on Nvidia’s recent Cosmos platform to develop physical AI technologies such as robots and autonomous vehicles.

Nvidia introduced its Cosmos platform for generative world models and advanced video tokenizers suited to its GPUs at the CES 2025 event in Las Vegas. Physical AI systems deal with practical tasks such as object manipulation and mobility, unlike ChatGPT-style software AI.

HMB advancements ahead of Nvidia demand   

Nvidia’s generative AI systems need fifth-generation HBM3E chips, with SK Hynix leading the pack. According to Chey, whose company’s HBM is now advancing ahead of Nvidia’s requests, SK Hynix has the technological edge.

SK is expected to introduce HBM3E 16-layer samples and 12-layer HBM4 products during the first half of 2025.

Nvidia is the company’s largest customer, and the entire production of HBM chips for 2025 is already sold out. Currently, SK Hynix supplies Nvidia with HBM3E 12-layer products and is expanding its capabilities to address growing AI demand.

SK Group also prioritizes AI data centers as a growth driver. Chey stressed energy-efficient solutions for power-hungry facilities, such as optimized power supplies, cooling systems, and energy management overall. SK believes there is a huge opportunity to collaborate these efforts with its broader technology-driven transformation.