To The Moon

Samsung Echoes Micron's Beat as the Memory Cycle Firms Up

The memory complex is riding a sustained price-upcycle, and core vendors' earnings have shifted into a high-growth phase.

Samsung Electronics: Q3 profit at a three-year high

Samsung released its Q3 guidance: revenue of ₩86 trillion (up 8.7% YoY) and operating profit of ₩12.1 trillion, well above the roughly ₩10 trillion consensus.

This marks the highest quarterly profit since Q2 2022 (₩14.1 trillion) and a 158% QoQ rebound. While detailed figures are not yet disclosed, analysts estimate the Device Solutions (DS) division—driven mostly by memory (roughly 70% of DS revenue)—delivered about ₩5 trillion in profit, more than 10× Q2's ₩0.4 trillion. Samsung will publish full segment results on October 30.

Micron's late-September results were also strong, with revenue +46% YoY and operating profit +30% YoY. (See Micron's Revenue Soars on AI-Fueled Demand, But HBM Questions Loom)

Memory pricing is strongly rebounding

~DRAM: Per Meritz Securities, over the past month DDR5 16Gb spot pricing rose 20.56%, and DDR4 8Gb rose 21.65%.

~NAND: TrendForce data show 512Gb TLC NAND increased from $2.82 in August to $3.45 in September

~Contract moves: Into Q4, Samsung reportedly lifted some DRAM contract prices by up to 30%, and NAND by 5–10%.

The AI compute boom began with a surge in HBM demand—critical for AI accelerators—and is now spilling over into general-purpose memory. As Samsung, SK hynix, and Micron allocate more DRAM capacity toward HBM4 for Nvidia and others, output of commodity DRAM declines, lifting overall DRAM pricing.

NAND is feeling a similar tailwind. AI data-center buildouts are driving up demand for SSDs (and even HDDs). In SSDs, NAND from Samsung, SK hynix, and SanDisk (Kioxia/WD) is the key component.

AI is setting up a new “memory supercycle.”

The market broadly expects the price upcycle to at least extend into next year. A 2018-style supercycle could re-emerge.

In 2018, memory—especially DRAM—hit record pricing and margins as synchronized smartphone and hyperscale-server demand (more bits per device) collided with tight supply from limited wafer adds and slow node ramps.

Unlike 2018, the key shift in this memory cycle is that AI is driving a sustained rise in demand. The industry narrative is shifting from a price-upcycle driven by production cuts to a structural demand gap driven by AI—suggesting greater durability.

Demand side:

~HDD & SSD. The “train → infer → retrain” loop in large AI models keeps creating huge amounts of data, so storage needs keep climbing. The backbone of data storage—nearline HDDs—is already in shortage, with lead times now over 52 weeks. North American cloud providers are even pre-booking some HDD makers’ 2026 capacity. As inference rolls out more broadly, we expect ongoing growth in demand for both HDDs and SSDs.

~DRAM. HBM demand is rising steadily because each GPU/ASIC carries more memory per device and unit shipments are growing—this also soaks up fab capacity that would have gone to commodity DRAM. At the same time, server DRAM (e.g., DDR5) keeps increasing as memory per server goes up. Looking ahead, if AI PCs and AI smartphones scale, total DRAM demand should get another leg up.

Supply side :

Over the past few years, memory makers haven’t added much brand-new capacity. Instead, they’ve shifted existing lines toward AI-centric products like HBM, DDR5, and QLC NAND. Because new fabs take a long time to build and ramp, near-term “new supply” mostly comes from raising utilization—which is relatively slow and rigid—so the market stays tight.

$Micron Technology (MU.US)$ : A vertically integrated memory maker across DRAM (including HBM) and NAND, with exposure to data center, PC, mobile, and auto. The investment case is leverage to a DRAM-led upcycle plus a structural mix shift toward HBM, which can lift gross margins as cost/bit falls on new nodes. U.S. fabrication and incentives add geopolitical and cost advantages. Watch next: HBM ramp (capacity, yields, customer wins), DRAM/NAND bit growth vs. cost/bit decline, inventory turns, and node transitions (e.g., 1β/1γ) translating to margin expansion.

$SanDisk Corp (SNDK.US)$: Focused on NAND—client/enterprise SSDs, embedded (UFS). The investment angle is high operating leverage to a NAND pricing recovery and premium retail positioning that can support ASPs and mix. Brand strength plus the Kioxia JV scale are key moats. Watch next: NAND contract/spot price trend, enterprise SSD mix, JV fab utilization/wafer adds, and any portfolio actions at WDC that could highlight flash asset value.

$Western Digital (WDC.US)$: A dual-engine portfolio: nearline HDD for hyperscalers and NAND (via the Kioxia JV). This structure hedges cycles and gives re-rating optionality if the company simplifies the portfolio over time. AI-driven data creation supports exabyte growth in nearline HDD, while enterprise SSD recovery provides a second leg. Watch next: Nearline HDD exabyte shipments and capacity points, HDD ASP/margin trajectory, NAND pricing/margins, cash flow discipline, and any strategic moves on separation or partnerships.

$Seagate Technology (STX.US)$: A pure-play nearline HDD leader, pushing HAMR to enable 30TB+ class drives and a lower $/TB for hyperscale storage. The investment case is straightforward operating leverage to AI-era “data lake” expansion and capacity upgrades, supported by a history of shareholder returns. Concentration risk is higher (HDD-centric), so execution on HAMR yields and mix is critical. Watch next: HAMR qualification and yield progress, nearline exabytes shipped, utilization/lead times, gross margin cadence, and hyperscaler capex signals.

Risks: The sector has rallied sharply in the near term, the AI trade is crowded—watch for a potential pullback.

@TigerStars  @CaptainTiger  @TigerWire   @Daily_Discussion  @Tiger_chat  @Tiger_comments  @MillionaireTiger