Nvidia gets the headlines. SK Hynix gets the contracts.
The South Korean chipmaker has quietly positioned itself as the single most indispensable company in the AI hardware stack, controlling roughly 56% of the global high-bandwidth memory market and serving as Nvidia’s primary supplier of the memory chips that make AI accelerators actually work.
From Seoul to Nasdaq, with a $26.5B detour
The clearest sign of SK Hynix’s arrival as a global force came in late July 2026, when the company completed a US listing that raised $26.5 billion, the largest IPO by a foreign entity in American market history.
American Depositary Receipts began trading on Nasdaq under the ticker SKHY, debuting at $149 per share.
Nvidia has secured multi-year HBM supply commitments from SK Hynix estimated to be worth hundreds of billions over time. That is not a vendor relationship. That is a strategic dependency.
A $1 trillion milestone, then a reality check
Earlier in 2026, SK Hynix stock surged more than 250% year-to-date, pushing the company’s market capitalization to $1 trillion in May.
SK Hynix posted record quarterly profits in Q2 2026, driven almost entirely by AI memory sales. The market, however, had priced in exceptional plus more, and when the results landed short of elevated analyst projections, shares dropped between 9.6% and 18% in post-report trading.
The stock has since fallen more than 50% from its peak.
What this means for the broader AI trade
The concern that rattled markets following the Q2 report was not about SK Hynix specifically. It was about whether the pace of capital expenditure from hyperscalers, the Googles, Microsofts, and Amazons of the world, would remain as aggressive as investors had assumed. Any slowdown in that spending hits SK Hynix almost immediately, given how concentrated its AI revenue is.
Samsung and Micron are both working to close the gap in HBM market share, but SK Hynix’s 56% position did not appear overnight, and it will not disappear overnight either. The company’s relationship with Nvidia runs deep enough that a meaningful supply shift would require Nvidia to actively diversify, which carries its own engineering and qualification costs.