Nvidia and SK hynix formalise multi-year HBM supply pact as shortage outlook extends
Nvidia CEO Jensen Huang warns memory scarcity will persist for years, shifting industry focus from cyclical tightness to long-term capacity planning as SK hynix secures dominant share of HBM4 requirements.
Nvidia and SK hynix have expanded their strategic partnership to secure multi-year high-bandwidth memory (HBM) supply for Nvidia’s artificial intelligence accelerators. The agreement includes multi-year volume commitments, accelerated production ramps, and yield optimization partnerships, designed to address the persistent bottleneck in the AI infrastructure buildout.
Nvidia CEO Jensen Huang characterised the global memory shortage as a structural constraint likely to persist for several years. This assessment shifts the market narrative from temporary cyclical tightness to long-term scarcity, signalling that the semiconductor industry is moving away from typical boom-bust inventory corrections toward long-duration planning cycles.
SK hynix is estimated to supply between 50 and 70 percent of Nvidia’s HBM4 requirements. This level of concentration is unusual in the semiconductor sector, reflecting the reality that only a handful of suppliers can manufacture HBM at leading-edge performance and yield. SK hynix currently serves as the leading supplier of HBM3E and HBM4 used in Nvidia’s Blackwell accelerators and next-generation platforms such as Vera Rubin.
The deepening relationship between the two companies involves industrial co-design initiatives. These include the potential use of Nvidia Omniverse for semiconductor plant simulation and efficiency gains. Nvidia’s prior collaboration with SK Group introduced the concept of “AI factories” in Korea, combining manufacturing, simulation, and AI optimization in a continuous loop to synchronise production systems.
SK hynix has outlined plans to roughly double its memory capacity over a five-year horizon. The expanded deal locks Nvidia into multi-year supply visibility while guaranteeing SK hynix structural demand growth for the duration of the AI cycle, effectively turning scarcity into scheduling discipline for both firms.
