SK Hynix warns memory chip demand faces an “unprecedented deterioration” and reports Q3 2022 operating profit fell 60% YoY to ~$1.16B amid soaring inflation
South Korea's SK Hynix Inc (000660.KS) warned on Wednesday of an “unprecedented deterioration” in memory chip demand …
Context & Ripple Effects
The 2022 warning marked the downturn’s opening phase for SK Hynix: falling memory prices were followed by its largest quarterly operating loss in early 2023. The later recovery was uneven rather than a return to the same market conditions; by mid-2024, the company was again reporting sharply higher revenue and profit as chip demand improved.
Subsequent results show why this demand warning matters as a cycle marker. SK Hynix’s later records were explicitly driven by AI HBM demand, and the company was later described as benefiting from an HBM shortage—evidence that high-bandwidth memory became a distinct profit engine from the broad memory market that weakened in 2022.
First-order effects
- SK Hynix enters a period of lower profitability, with third-quarter operating profit down 60% year over year and management flagging a severe deterioration in memory demand.
- The warning increases pressure on SK Hynix to manage the downturn in its core memory business as inflation weakens demand.
Second-order effects
- The subsequent 2023 operating loss indicates that the deterioration extended beyond one quarter, turning a demand warning into a deeper earnings and pricing problem for SK Hynix.
- As demand later recovered, SK Hynix’s results became increasingly tied to HBM: the company’s 2025 and 2026 record profits were attributed to AI HBM demand rather than a uniform rebound across memory.
Third-order effects
- The sequence points to a more segmented memory industry, in which commodity-memory cycles can remain punishing even as specialized HBM commands scarcity-driven growth and margins.
- If that separation persists, memory suppliers’ competitive position will depend less on aggregate chip demand and more on whether their product mix reaches AI infrastructure bottlenecks.
The trend: Memory is shifting from a single cyclical market toward distinct tiers, with AI-oriented HBM increasingly insulating leading suppliers from weakness in conventional memory demand.