Nvidia and SK Hynix Sign Multi-Year AI Chip Deal

SK Hynix announced on June 7 that it has signed a multi-year agreement to supply Nvidia with next-generation memory chips, deepening a partnership that has become central to the global AI supply chain.

The South Korean chipmaker said it will begin delivering high-bandwidth memory 4 (HBM4) chips to Nvidia in the second half of 2026, following months of qualification testing. The chips are critical components in Nvidia's most advanced AI processors, including the B200 and upcoming Rubin architecture, where they sit alongside the GPU to handle the massive data throughput that large language models require.

SK Hynix already supplies the majority of HBM3E chips that go into Nvidia's H200 and B200 processors. The company said the new agreement covers multiple product generations and extends through 2030, though financial terms were not disclosed. The deal also includes collaboration on what the two firms are calling "AI factories," large-scale data centres purpose-built for training and running AI models.

The announcement resolves some uncertainty that had lingered around SK Hynix's HBM4 timeline. Industry watchers had questioned whether the company could maintain its lead over rival Samsung Electronics, which has been racing to qualify its own HBM4 chips with Nvidia. Samsung has yet to publicly announce a similar supply agreement.

Nvidia CEO Jensen Huang, speaking at Computex in Taipei last week, had flagged that HBM4 supply would be tight through 2027, signalling the importance of locking in long-term contracts early. SK Hynix shares rose 3.2 percent in Seoul trading following the announcement.

The partnership underscores how the AI boom has reshaped the memory industry. HBM chips, once a niche product, now command premium pricing and account for a growing share of revenue at both SK Hynix and Samsung, with demand expected to triple by 2028.