SK Hynix

SK Hynix manufactures high-bandwidth memory (HBM) that sits on every leading AI accelerator, plus DRAM and NAND for data centers. HBM revenue surged 300%+ in 2023; tight supply and long lead times. First to ship HBM3E in volume.

SK Hynix supplies the majority of HBM3/HBM3E for Nvidia H100/H200, making it critical to AI accelerator performance.

skhynix.com Memory

South Koreapublic · 000660.KS

At a glance

Total raised

Undisclosed

no public funding on record

Headcount

—

no public estimate on record

Headquarters

Icheon

South Korea

Founded

1983

43 years active

Status

Public

Ticker · 000660.KS

Subsector

Memory

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

91 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

—

Rounds tracked

Latest stage

IPO

Date	Stage	Amount	Valuation	Lead investors
Jan 1, 1996	IPO	—	—	—

Stock performance

How they make money

Hardware (one-time)Pattern

SK Hynix runs a semiconductor hardware business: it sells memory chips and modules to equipment manufacturers and data-center operators on a per-unit basis, with revenue recognized at shipment. The company manufactures high-bandwidth memory (HBM), DRAM, and NAND flash. Its primary customer is Nvidia, which integrates SK Hynix's HBM3 and HBM3E into its H100 and H200 AI accelerators. Other buyers include hyperscale cloud providers and server OEMs who purchase DRAM and NAND for data-center infrastructure. Revenue is transactional — each purchase order is a discrete sale. The economics are driven by fab utilization, wafer yields, and selling price per bit. SK Hynix invests heavily in fabrication facilities and advanced packaging; fixed costs are high, so margin expands when capacity runs full and ASPs are favorable, as they have been during the AI-driven HBM shortage. HBM commands a premium over commodity DRAM because of the advanced stacking technology required. There is no recurring service or subscription component — the model is purely unit-volume hardware, with the cyclicality typical of the memory industry.

Generated by deepseek-v4-flash · June 2, 2026Analytical summary · not investment advice

AI bull / bear

Bull case

SK Hynix is the most irreplaceable supplier in the AI hardware stack. It ships the majority of HBM3 and HBM3E memory that powers Nvidia's H100 and H200 accelerators, and it was the first to deliver HBM3E at volume. With HBM revenue surging 300%+ in 2023 and its chairman stating the AI-driven memory shortage will persist through at least 2030, the demand runway is measured in years, not quarters. The company plans to double memory wafer capacity within five years, a clear signal of confidence that the current ordering frenzy is structural, not cyclical. No other memory maker has matched SK Hynix's HBM production cadence or yield, giving it pricing power and long-term customer lock-in with the largest AI compute buyer on the planet. As long as training and inference workloads scale, SK Hynix sits directly in the flow of hyperscaler capital expenditure.

Bear case

Samsung and Micron are closing the HBM gap fast, and both have announced aggressive HBM3E qualification timelines. If either achieves comparable yields, SK Hynix loses its sole-source premium and faces margin compression. The company also carries real operational risk: a toxic fluorine-gas fire at its Cheongju plant hospitalized seven workers and forced a 3,600-person evacuation, underscoring the fragility of advanced memory fabrication. Doubling wafer capacity in five years requires flawless execution across multiple new fabs, a tall order when a single incident can halt production. Meanwhile, customer concentration is extreme — Nvidia accounted for the bulk of HBM spend in 2023 — and any shift by Nvidia toward dual-sourcing or in-house memory integration would slice directly into SK Hynix's revenue. Finally, the memory industry has never broken its boom-bust cycle, and a correction in AI chip demand could leave newly built capacity underutilized.

Generated by deepseek-v4-flash · June 2, 2026Refreshes quarterly · see disclosures

In the news

Watch

Leadership

Kwak Noh-jung

President & CEO

—

Products

HBM3E High-Bandwidth Memory
SK Hynix's stacked high-bandwidth memory that made it the leading HBM supplier for AI accelerators. Its 12-layer HBM3E (36GB) is the workhorse memory on Nvidia's flagship GPUs, and the company has held over 50% of the HBM market through the HBM3/HBM3E generations. Co-developed closely with Nvidia, HBM3E is the 'fuel' for training trillion-parameter models and the single biggest driver of SK Hynix's record AI-era profits.
HBM4 High-Bandwidth Memory
SK Hynix's next-generation HBM, debuted at CES 2026 as a 16-layer 48GB stack. HBM4 widens the memory interface and raises bandwidth and capacity to feed next-generation accelerators such as Nvidia's Vera Rubin platform. SK Hynix aims to extend its HBM market leadership into the HBM4 generation, where it competes head-to-head with Micron and Samsung for hyperscaler and AI-lab design wins.
Enterprise DRAM, SOCAMM2 & eSSD (incl. Solidigm)
Beyond HBM, SK Hynix supplies server DDR5, low-power LPDDR, and compact SOCAMM2 modules for AI servers, plus high-capacity enterprise NAND SSDs through its Solidigm subsidiary (the former Intel NAND business). This broad memory-and-storage portfolio lets SK Hynix serve the full AI data-center stack, from accelerator-attached HBM to bulk DRAM and QLC storage for AI datasets.

Compare SK Hynix with

vs Memory
SK Hynix vs Micron