SAN FRANCISCO — Nvidia has secured a multi-year high-bandwidth memory supply agreement with SK Hynix, a deal that forms a cornerstone of the company's $500 billion AI infrastructure strategy and guarantees access to one of the scarcest components in advanced GPU production.
HBM is essential for AI accelerators because it allows processors to move data at the speeds required for large language model training and inference. SK Hynix, the leading HBM producer, had already reported selling out its entire output through the end of 2025. Nvidia's agreement locks down a substantial share of that constrained supply.
The deal tightens Nvidia's grip on its AI accelerator business, protecting shipments of its highest-margin products while raising the cost of entry for competitors attempting to scale rival hardware at volume.
Nvidia shares traded at $206.84, down 0.9 percent, against a Nasdaq decline of 0.6 percent. The company's move to control supply chain chokepoints reflects its read on sustained long-term demand for AI infrastructure.
For SK Hynix, the agreement provides guaranteed revenue across its advanced memory lines as cloud providers and enterprises accelerate data center build-outs. Memory chipmakers are operating near capacity as AI workloads drive sustained demand for specialized components.
