Nvidia has struck a massive long-term memory supply agreement with South Korea's SK Hynix, according to multiple reports out of Seoul and the trade press.
The Korea Herald reported that Nvidia CEO Jensen Huang said the company and SK would announce a $500 billion partnership. HotHardware described the arrangement as Nvidia securing SK Hynix HBM — high-bandwidth memory — as part of a $500 billion AI infrastructure deal, and GuruFocus and The Tech Buzz reported the same $500 billion figure.
There is at least one notable discrepancy in the numbers. Seoul Economic Daily, citing the South Korean presidential office, reported that SK would supply Nvidia with $750 billion worth of product in a long-term deal spanning five years — a larger figure and a longer stated horizon than the $500 billion headline carried elsewhere. The sources available here do not reconcile the two numbers, so the final structure and accounting of the agreement remain unclear.
What is consistent across the reporting is the substance: Nvidia is locking up a supply of the specialized memory chips that sit alongside its AI processors. Those memory stacks are one of the tightest bottlenecks in AI hardware — a graphics processor is only as useful as its ability to feed data in and out quickly, and SK Hynix is a leading supplier of the parts that do it. The Tech Buzz framed the agreement in terms of an ongoing "AI arms race."
It also underscores how much leverage has shifted toward memory makers. A deal of this scale is effectively Nvidia paying for certainty: guaranteed parts, years in advance, at a time when rivals are competing for the same limited output.
Why it matters: if the biggest buyer in AI hardware is willing to commit hundreds of billions of dollars to guarantee memory supply, it signals that the constraint on AI expansion is no longer just chip design or funding — it's who can physically get the components.