SK Group Chairman Chey Tae-won is set to attend Nvidia’s annual technology conference this month, with expectations high for discussions on AI memory chip collaboration with Nvidia CEO Jensen Huang.
Industry sources revealed on Thursday that Chey will travel to San Jose, California, for Nvidia GTC 2026, commencing on March 16. This marks the inaugural attendance of the SK chairman at Nvidia’s prominent developer conference.
This highly anticipated meeting follows their informal “chicken and beer” dinner in February, where both leaders reaffirmed their commitment to partnering on high-bandwidth memory (HBM) supply.
At GTC 2026, Nvidia is expected to introduce its next-generation AI accelerator platform, Vera Rubin. This system is anticipated to incorporate HBM4 – the sixth generation of HBM – intensifying the competition among memory chip manufacturers vying for supply agreements.
Reports indicate that Nvidia has allocated approximately two-thirds of its HBM4 orders for this year to SK hynix, although Samsung Electronics is showing strong early performance.
Samsung announced the world’s first mass shipment of HBM4 in February, and its chips are projected to be integrated into Nvidia’s Vera Rubin platform, set to be unveiled at the conference.
Industry observers foresee Chey and Huang engaging in discussions across various strategic cooperation areas, including HBM4 supply coordination, collaborative development of future HBM technologies, and potential partnerships in AI data centers and energy infrastructure.
SK hynix will also showcase its cutting-edge AI memory technologies at GTC 2026 with a dedicated exhibition booth.
The company plans to present its HBM4 chips, currently undergoing final optimization for mass production, alongside its flagship HBM3E products. The booth will feature physical samples of the memory chips and Nvidia AI systems powered by them, according to industry sources.
This year’s GTC, scheduled from March 16-19, is projected to draw around 30,000 participants from 190 countries and territories, according to Nvidia.
