Memory is a boom-and-bust business. The records of today sit at the *top* of a cycle that has always, eventually, turned.
↓ scroll to read
✦ The bottom line
Four real risks: memory cyclicality (profits swing enormously), customer concentration (HBM rides on a small set of AI-chip buyers), geopolitics (a Korean maker exposed to US–China export controls), and heavy capex into a cyclical demand curve. The offset is the fortress balance sheet from Chapter 2.
↓ the brief below
Operating profit · ₩ trillion · the swing is the risk
₩7.4T
₩19.2T
₩37.6T
1Q25
4Q25
1Q26
Operating profit ran ₩7.4T → ₩19.2T → ₩37.6T in a year. The same steep climb that makes the growth chapter exciting is the cyclicality risk: memory profits move in huge swings, and history says up-cycles this sharp eventually reverse. We won't put a number on the next downturn — but the amplitude is the warning.
Cyclicality is the risk inside the product. The next two risks are about who buys it and where it's made. HBM's growth is tied to a concentrated set of AI-accelerator customers — when a handful of buyers drive most of the demand for your hottest product, their spending decisions become yours. SK hynix doesn't disclose a named-customer percentage, so we won't guess one; the point is the dependence, not a figure.
✦ Teach me
China / export-control geopolitics
SK hynix is a Korean company that sells globally and has manufacturing exposure in China. Advanced chips and the machines that make them sit at the center of US–China tensions: governments restrict which equipment and chips can be sold where, and those rules change with the politics.
For a memory maker, that means a fab or a key customer relationship can be helped or hurt by decisions made in Washington, Beijing, or Seoul — not just by supply and demand.
Wall Street calls this
Export controls / geopolitical exposure
When part of your manufacturing and a chunk of the end market sit on opposite sides of a trade fight, *policy* becomes a real swing factor on the business. SK hynix discloses no single number for this, so treat it as a *qualitative* risk to watch, not a quantified one.
Capital spending · direction for 2026
Rising
Management guided that "this year's investment scale will increase significantly" — ramping the M15X fab, the Yongin cluster, and EUV equipment. Big fixed-cost building into a cyclical demand curve is the risk; the offset is the fortress balance sheet (₩35T net cash) that funds it from strength. The release gave no exact capex figure, so none is stated here.
Source · earnings-release · 1Q26 prepared remarks — investment outlook (no specific capex figure disclosed) · Q1 2026 · Filed Apr 22, 2026
One forward-looking item, reported as news rather than disclosed as a number: in late March 2026, CNBC reported that SK hynix had confidentially filed a Form F-1 with the US SEC for a potential American depositary-receipt (ADR) listing. If it happens, US investors could one day buy the shares more directly — but it's a reported plan, not a completed deal, and no figures here rest on it.
⚠
Watch
Real risks: a steep memory cycle, dependence on a few AI-chip buyers, US–China export-control exposure, and heavy spending ahead. The cash to ride it out is there — what you can't see yet is when the cycle turns.