The partnership includes a 2-gigawatt AI cloud data center and long-term high bandwidth memory supply for Nvidia’s GPUs.
Nvidia and SK Hynix finalized the largest memory deal in history, securing high bandwidth memory (HBM) supply for Nvidia’s AI chips. SK Hynix’s subsidiary, SK Telecom, will build a 2-gigawatt AI cloud data center in Korea using Nvidia’s Vera Rubin Platform.
The agreement addresses a critical bottleneck in AI infrastructure, as HBM demand surges due to its role in reducing latency and improving power efficiency. However, HBM production faces constraints, including limited EUV lithography machines from ASML and the need for three times the wafer capacity of standard DRAM.
Both companies will co-develop future AI memory generations, aiming to bolster supply amid rising demand for advanced AI hardware.