SK Hynix Lands Key Role in Nvidia’s Vera Rubin, Yet Stock Suffers Its Worst Day in Months

SK Hynix shares plunged 10% after Nvidia confirmed HBM4 supply, driven by weak won and US tech selloff. Despite the drop, stock is up 205% YTD on AI demand.

The paradox facing SK Hynix could hardly be starker. The South Korean chipmaker has just secured a coveted spot on Nvidia’s upcoming Vera Rubin AI platform — confirmation that it will supply HBM4 memory alongside Samsung and Micron. But the market response was anything but celebratory. On Friday, shares crashed nearly 10% to 2,070,000 won, their sharpest single-day decline since early 2025.

The selloff was not triggered by company-specific bad news. Rather, a weak South Korean won and a broad pullback in US technology stocks dragged the entire sector lower. With the Seoul exchange closed over the weekend, Friday’s close serves as the reference point for Monday’s open — leaving investors to question whether the Nvidia validation can help reverse the slide.

The Numbers Behind the Whipsaw

Despite the rout, SK Hynix’s long-term trajectory remains extraordinary. Year-to-date, the stock has surged roughly 205% — a gain built on relentless AI demand for high-bandwidth memory. The all-time high of 2,407,000 won was set just three days before Friday’s plunge, on June 2. By contrast, the 52-week low of 491,500 won dates back to October 2025, meaning the current price is still more than four times that trough.

Nvidia’s Vera Rubin Timeline Takes Shape

Jensen Huang personally confirmed during a visit to South Korea that Samsung, SK Hynix and Micron have all been qualified as HBM4 suppliers for Nvidia’s next-generation accelerator platform. Vera Rubin — built around the Vera CPU, Rubin GPU and NVLink-6 switch — is scheduled for commercial availability in the second half of 2026. AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure will be among the first cloud providers to deploy Rubin-based systems.

Nvidia’s deliberate move to a multi-supplier structure for HBM4 reduces its reliance on any single memory maker. For SK Hynix, that cuts both ways: the qualification validates its technology leadership, but the increased competition could eventually squeeze pricing power once supply catches up with demand.

Next-Generation HBM4E and US Listing Plans

While the HBM4 ramp is still over a year away, SK Hynix is already pushing ahead with the follow-on generation. First engineering samples of HBM4E are slated to go to customers in the second half of 2026, with commercial production expected in 2027. Management expects memory supply to remain tight for years, underpinning aggressive investment in new packaging facilities.

Simultaneously, the company is taking steps to broaden its investor base. SK Hynix plans to issue American Depositary Receipts in the US later this year, opening the door to large institutional funds restricted to US-listed securities. The move comes after the company’s market capitalization briefly topped $1 trillion at the end of May — a milestone that underscores its transformation from a cyclical memory player into a cornerstone of the AI infrastructure buildout.

Macro Headwinds Versus Structural Demand

The immediate question for the Seoul market on Monday is whether the Nvidia signal will offset Friday’s macro-driven rout. Short-term, the multi-supplier qualification confirms that SK Hynix remains a key partner in the most critical bottleneck of AI server production: high-bandwidth memory. Longer-term, the real test will be whether demand outpaces the industry’s ability to add capacity — preserving the pricing leverage that has propelled the stock’s meteoric rise. For now, the market is weighing two competing narratives: a technology win that reinforces SK Hynix’s position, and a broader risk-off mood that has little to do with the company’s fundamentals.