Goldman's Samsung Triple-Hit: Three Flow Catalysts Land on Oct 8 — the Earnings Story Is Secondary

Goldman Sachs framed the warning around three mechanical selling pressures landing simultaneously on Thursday, Oct 8 — the same day Samsung Electronics is also due to release preliminary Q3 2026 results. The convergence of flows, not fundamentals, is what makes this rare. According to Goldman's note, it is uncommon for one stock to see so many independent flow catalysts on a single trading session.

Goldman Sachs lowered its Samsung Electronics Q3 2026 operating profit forecast by ~5%, from W112 trillion to W106 trillion — and pinned the entire cut on the Korean won, not on chip demand. The firm's revised FX assumption is USD/KRW 1,418 for the quarter versus its prior 1,460. The won is ~3% stronger than the bank had modeled, which mechanically shrinks the won value of Samsung's US-dollar-priced memory sales without changing bit volumes or pricing.

Street consensus has moved in the same direction but from a higher peak: it has slid from roughly W114 trillion in August to W105.5 trillion today, suggesting the FX overhang is broadly priced into expectations. Critically, Samsung Electronics HBM bit shipments are still expected to grow close to 50% quarter-on-quarter in Q3 2026, driven by HBM4 ramp — the only thing being repriced is the currency translation.

The HBM half of Samsung's story is what makes the Goldman target of KRW 490,000 still make sense. Goldman projects Samsung's HBM revenue to jump from roughly US$20 billion in 2026 to US$74 billion in 2027 — a 274% surge driven by HBM ASP of ~US$3.6/Gb (+130% year-on-year) and shipments of ~21 billion Gb (+60%). HBM's share of Samsung's DRAM revenue rises from 8% in 2026 to 19% by 2027 on these figures, pulling the company's DRAM mix decisively up the value curve.

That said, Samsung Electronics remains the share challenger, not the share leader, in HBM. UBS projects SK Hynix at roughly 70% of NVIDIA's HBM4 supply for the Rubin platform, with Samsung and Micron splitting most of the rest. Counterpoint's Q2 2026 numbers put Samsung at 38% of total DRAM but a much smaller slice of HBM specifically — the ramp into HBM4 is the path to closing that gap. Samsung passed final HBM4 qualification tests for both NVIDIA and AMD in early 2026, began mass production in February, and is expanding HBM supply to ASIC customers such as Google's TPU alongside the NVIDIA pull.

The foundry leg of the triangle is where Samsung Electronics is most exposed to a longer-cycle story — but also where the inflection is now visible in the data. Samsung's 2nm process yield has climbed from around 30% in mid-2025 to nearly 60% by late September 2026, per multiple industry trackers. The company's Taylor, Texas fab is ramping, with Tesla's AI5 and AI6 chips reportedly first in line on a deal sized around US$16.5 billion.

The scale gap with the foundry leader is still stark — and the Q2 2026 numbers say so plainly. TSMC generated US$40.2 billion of foundry revenue at a 72.5% share; Samsung Foundry generated US$3.26 billion at a 5.9% share, down from 6.5% a quarter earlier. Samsung is still tracking a 130% projected increase in 2nm-related orders in 2026 versus 2025, but the gap to TSMC is roughly a dozen-fold in absolute revenue.

The third leg of the triangle is the one the HBM narrative usually leaves out — what happens when Apple, Samsung's largest traditional memory customer, develops a credible alternative supplier. Per the Financial Times and follow-up coverage, Apple began testing DRAM from China's state-backed ChangXin Memory Technologies (CXMT) in mid-2026 for iPhones and MacBooks sold within China. CXMT reportedly rejected Apple's attempt to negotiate lower pricing for that volume — a striking position for a still-emerging supplier to take.

The implications cut two ways for Samsung Electronics. On one hand, Apple still needs Samsung and SK Hynix for the bulk of its DRAM volumes and is reportedly capacity-constrained through 2026 — it has publicly signaled it is working to double capacity by 2028. On the other hand, having CXMT at the table as a bargaining chip compresses Samsung's pricing power on China-bound mobile DRAM and gives Apple optionality the next time memory contracts come up for renewal. It also adds a geopolitical layer that sits behind the cycle — CXMT is on the U.S. entity list, which limits how deeply Apple can integrate Chinese DRAM into global SKUs.

The ETF rebalancing is the cleanest part of the trade because the flows are visible in advance. Samsung Securities estimates that the rotation out of Samsung Electronics above the cap generates buying demand of about W90 billion for SK Hynix and W80 billion for SK Square, plus roughly W20 billion each into Hanmi Semiconductor, Jusung Engineering, Isu Petasys, and Wonik IPS.

The article splits cleanly across two time horizons, and the article's central tension is that they point in opposite directions.

The synthesis: Goldman's triple-hit warning is a flow call, not a fundamental call, and the market's job on Oct 8 is to separate the two. If foreign buyers return and Q3 prelim clears the W100T threshold with the HBM mix story intact, the trade through year-end is to fade the Oct 8 noise and lean into the 2027 HBM ramp. If both fail at once, even Goldman's KRW 490,000 target is in play.