Samsung's Q2 operating profit wasn't a conglomerate achievement — it was one division. DS semiconductors contributed 94% of total operating profit; smartphones, displays, and appliances combined for the remaining 6%.
The DS division chief told staff internally that 2026 chip profit would exceed the division's cumulative earnings across its entire 40-year history in semiconductors. The rest of Samsung is a rounding error in this story.
Samsung, SK Hynix, and Micron diverted capacity to HBM at a 3:1 wafer-conversion penalty against conventional DRAM — shrinking supply precisely as AI demand spiked. IDC projects 2026 DRAM supply growth at just 16%, NAND at 17%. Both are structurally undersupplied. This profit is a pricing story, not an innovation one.
Profit up 1,810% year-on-year. Stock down 8% the same day. A supply chokepoint, not a product moat, doesn't hold — and the market priced that in before the ink dried.
— The market's verdict, July 7, 2026
New fab capacity won't arrive at scale until late 2027 or 2028 — that date, flagged by TrendForce, Morgan Stanley, and TechInsights, is the structural ceiling on Samsung's pricing window.
Every major industry forecast treats this as a supply-constrained cycle, not a permanent handoff from compute to memory. The 'Samsung dethroned Nvidia' framing borrows permanent language to describe a temporary shortage.
Nvidia built the engine. Samsung owns the only gas station on the highway right now, and it's charging accordingly. Don't confuse the toll booth for the highway.
