They grew HBM sales 1.9x sequentially and still missed their own internal forecast because customers refused to take the product until it was 'optimized'.
Thesis: Samsung is caught in the 'middle income trap' of semiconductors. They are aggressively cutting legacy capacity (reducing low-end DRAM share from 30% to single digits) to escape Chinese commoditization, but they haven't yet secured the high ground in AI memory. The Q1 'gap' is real: customers are drawing down inventory and waiting for the 12-stack HBM3E that Samsung promises is coming. Until they prove they can ship the optimized product at volume, this is a 'Show Me' story, not a growth story.
Verdict: HOLD — Conviction: MEDIUM
Catalyst: Q2 earnings call confirmation of 'full swing delivery' of optimized HBM3E 12-stack to major GPU customers.
Key Risk: If the 'optimization delays' extend into H2, the 2025 HBM bit supply doubling target becomes a fantasy, and they miss the window for the Nvidia Blackwell cycle.
The Tell: The CFO admitted that while HBM sales grew 1.9x, it was 'slightly below our initial forecast.' You don't miss a forecast on a product growing that fast unless your customers are rejecting your yield or performance. That's not a supply constraint; that's a product quality signal.
Friction Level: HIGH_FRICTION — Bull sees 'optimization' as a temporary Q1 blip before a Q2 ramp. Bear sees 'Say/Do Gap' where Samsung is perpetually one quarter behind SK Hynix on the only chip that matters.
Report not found
The report data is no longer available. Please return to the archive.