Investors have spent months debating whether the AI buildout is becoming a financing problem. The next two Korean earnings reports offer a different test: whether the physical demand underneath that buildout is still arriving.
SK hynix reports third-quarter results at 9 a.m. Seoul time on Oct. 28, according to its SEC filing. Samsung Electronics follows at 10 a.m. Seoul time on Oct. 29. Together they sit near the center of the global memory supply chain that feeds AI servers.
Why memory is a useful AI-demand check
AI data centers need large quantities of high-bandwidth memory and conventional DRAM alongside accelerators. Unlike a multiyear capital-spending promise, memory shipments, pricing and capacity utilization reveal whether customers are actually taking hardware now.
| Company | Report time | Key AI-demand signals |
|---|---|---|
| SK hynix | Oct. 28, 9 a.m. KST | HBM shipments and pricing, DRAM mix, capex, 2027 capacity commentary |
| Samsung Electronics | Oct. 29, 10 a.m. KST | HBM qualification and volume, memory pricing, foundry demand, capex |
SK hynix has another catalyst layered onto earnings
In an Oct. 26 Form 6-K, SK hynix said it is reviewing additional shareholder-return measures but that nothing had been decided. The company said any additional measures, subject to board approval, would be announced with third-quarter earnings.
That means the report can move the stock on two axes at once: the operating outlook for memory and the amount of excess value management is prepared to return to shareholders.
What would strengthen the AI-demand thesis?
For the broader AI complex, the cleanest positive signal would be continued tight HBM supply, firm pricing and confident customer commitments extending into 2027. Strong ordinary DRAM pricing would broaden the story beyond one premium product line.
Conversely, softer pricing, delayed customer ramps or a more cautious capacity plan would matter because the AI-financing debate is already asking whether capital is becoming scarcer than compute. Weak physical orders would turn that from a financing question into a demand question.
Why this matters beyond Korea
Memory is upstream from the U.S. hyperscalers. If Korean suppliers show accelerating real orders, it supports the idea that cloud and model-company capital commitments are being converted into installed infrastructure. If they show hesitation, investors have to ask whether the spending plans embedded in U.S. technology valuations are outrunning near-term deployment.
The two earnings calls therefore arrive as a useful cross-check on the same AI-capex thesis that will also be tested by large U.S. technology earnings this week.
Sources
Markets & News distinguishes verified disclosures, reported market reaction and our analysis. This article may be updated as primary-source information changes.