SK Hynix seeks record $28 billion Nasdaq IPO betting on AI memory shortage

The IPO and the core numbers

SK Hynix, the world's largest supplier of HBM (High Bandwidth Memory), started marketing its Nasdaq listing under the ticker SKHY on July 6, 2026, according to reporting from Bloomberg, TechCrunch and SiliconANGLE. The offering seeks to raise up to $28 billion, with 17.79 million shares on offer. If completed on these terms, it would be the second largest IPO in history, behind only SpaceX's IPO, which took place in June 2026, and the largest IPO by a foreign company ever done in the United States. According to Bloomberg, pricing is expected on July 9, with shares set to debut on Nasdaq on July 10, 2026. Institutional investors such as Baillie Gifford and Coatue have already signaled interest in buying up to $7 billion in shares, according to TechCrunch.

Why HBM memory became a bottleneck and an opportunity

HBM memory is used in Nvidia GPUs, including the H100, H200 and Blackwell, and has become one of the most contested links in the AI hardware supply chain. SK Hynix holds between 56% and 58% of the global market for this type of memory, a dominant position that has put it at the center of the race for AI computing capacity. SiliconANGLE reports that the company's HBM revenue already topped $10 billion in the first half of 2026 alone, up 198% year over year, with operating margins above 70% in the first quarter. Those numbers show that the memory bottleneck, as much as the GPU bottleneck, has become one of the most profitable and scarcest points in AI infrastructure.

What this signals for buyers of AI capacity

For smaller companies and automation teams that depend on cloud services and AI model APIs, SK Hynix's move is an indirect cost signal. When a supplier of a critical component like HBM memory can post operating margins above 70% and grow revenue 198% a year, it shows that demand for AI capacity still outstrips available supply. In practice, that means the price of running AI models in production could stay under pressure for longer, even with billions of dollars being poured into new capacity, because the bottleneck isn't only in GPUs: it's also in the memory components that make those GPUs work.

Analytical close

The size of SK Hynix's IPO shows that financial markets are betting heavily not just on AI model companies, but on whoever supplies the hardware that underpins that infrastructure. A $28 billion IPO for a memory maker, the second largest in history, signals that investors see component scarcity as a long-term thesis, not a temporary bottleneck that will resolve quickly.