SK Group Taps Nvidia, TSMC Alliance and MaaS to Balance HBM Supply

The spread of generative AI from model training into autonomous agents and large-scale inference is intensifying demand for high-bandwidth memory, or HBM. That has raised the risk of shortages and “chip inflation,” making memory capacity a critical constraint on AI infrastructure. SK Group is seeking to ease the pressure through closer supply-chain coordination and financing structures that distribute the cost of expansion.

SK Group Chairman Chey Tae-won said SK hynix is working with Nvidia and Taiwan Semiconductor Manufacturing Co. to secure capacity and better balance HBM supply and demand. SK also plans to pursue a Memory-as-a-Service, or MaaS, model and joint ventures that would allow customers to share investment risks while lowering AI computing costs. No investment amount, capacity target or implementation date was disclosed.