Samsung Memory just admitted they might pay TSMC to build their HBM base dies.
Thesis: Samsung is capitulating on its IDM advantage. The admission that they may outsource HBM4 base dies to 'external partners' (TSMC) reveals their internal foundry yields aren't competitive. They are a memory volume player trying to price like a specialist, but HBM3E delays prove they are structurally behind SK Hynix. The profit drop labeled as 'one-off incentives' is likely compensation for these delays. You don't own the #3 player in a winner-take-most market.
Verdict: AVOID — Conviction: MEDIUM
Catalyst: Q4 HBM3E mix hitting 50%. If they miss this target (again), the credibility floor collapses.
Key Risk: Legacy memory pricing spikes due to Chinese supply constraints, lifting Samsung's cash cow business regardless of AI execution errors.
The Tell: The surrender on HBM4 base dies: 'When choosing a foundry partner for production at the base side... We'll be flexible in our approach regardless of whether it involves an internal or external partner.' This destroys the 'One Samsung' synergy thesis.
Friction Level: HIGH_FRICTION — Bulls see a temporary qualification delay before a mix-shift breakout. Bears see structural subordination where Samsung is forced to use TSMC to compete in HBM.
Report not found
The report data is no longer available. Please return to the archive.