- SK Hynix raised $26.5 billion in the largest first-time US listing by a foreign company.
- Conventional DRAM contract prices rose roughly 93% to 98% quarter over quarter in Q1 2026.
- AI infrastructure funding reached $12.53 billion across just 27 disclosed rounds in the past year.
SK Hynix began trading on the Nasdaq on July 10 after raising $26.5 billion, the largest first-time US listing by a foreign company. The number that should worry AI founders more than the listing itself: the chipmaker holds 56.4% of the global high-bandwidth memory market, and that market is sold out for the rest of the year.
Tech Funding News has already covered how GPU scarcity is squeezing early-stage AI startups. The memory shortage compounds that problem rather than replacing it. Samsung, SK Hynix, and Micron control more than 95% of global DRAM output, and all three have been reallocating wafer capacity away from general-purpose memory toward HBM, the specialised chips that sit beside AI accelerators, because HBM commands two to three times the margin of standard DDR5.
The price data is stark. TrendForce data shows conventional DRAM contract prices rising 58% to 63% quarter over quarter in the second quarter of 2026 alone, with NAND flash climbing 70% to 75%. IDC expects total DRAM supply to grow just 16% this year, against AI demand capable of absorbing that entire increment and more. For any startup renting GPU capacity, memory is not a side cost. It typically makes up a meaningful share of the underlying hardware bill of materials, and unlike falling token prices, it is moving in the wrong direction.
Why founders can’t just wait it out
The instinct to treat this as a temporary supply hiccup runs into a harder truth: this is not the industry’s usual boom-and-bust cycle.
Analysts polled across TrendForce, IDC, and Gartner converge on the same point — memory makers are prioritising HBM because it is more profitable, not because consumer and general-purpose supply is a rounding error they will fix once demand eases. Micron has said its HBM output is effectively sold out for 2026. Intel has pointed to 2028 before conditions normalise.
That structural framing is exactly why investors are backing companies that attack the memory bottleneck directly rather than betting on it easing. Seoul-based XCENA, founded by engineers who cut their teeth at Samsung and SK Hynix, raised a $135 million Series B in May at a $570 million valuation, betting that placing compute closer to DRAM can cut the costly round trips between processors and memory.
It is one of a small but growing cohort of infrastructure startups treating memory efficiency as the next frontier, now that raw GPU supply has become a more familiar problem for investors to underwrite.
What the money is for
SK Hynix said proceeds will fund new fabrication capacity and equipment, including EUV lithography scanners from ASML, as it races to keep up with what industry watchers now describe as a structural, not cyclical, memory shortage.
The company’s Yongin Semiconductor Cluster in South Korea begins coming online in 2027, alongside its first US manufacturing site, a $4 billion advanced packaging plant in West Lafayette, Indiana, eligible for up to $458 million in CHIPS Act grants.
That expansion sits inside a broader AI memory arms race TFN has tracked closely. Micron has committed $200 billion to new US fabs to meet the same demand, while Samsung crossed a $1 trillion valuation in May on the back of its own HBM4 push. SK Hynix’s memory has also shown up as strategic infrastructure in AI funding itself — it was named as a partner in Anthropic’s $65 billion Series H, alongside Micron and Samsung.
The risk investors are pricing in
Memory has historically been one of the most cyclical corners of the chip industry, swinging between shortage and oversupply. SK Hynix and Samsung now account for more than 40% of South Korea’s benchmark Kospi index between them, a concentration some analysts warn leaves the market unusually exposed if AI infrastructure spending slows or supply catches up with demand.
For now, Counterpoint Research director MS Hwang says demand shows no sign of easing: hotels near SK Hynix’s South Korean facilities are reportedly booked solid with cloud providers and chipmakers lining up to sign long-term supply contracts.
