Samsung Tops DRAM Again (38% vs 29%): What the New Share Map Means for Buyers (2026)

An all-time record DRAM quarter, Samsung back on top by a wide margin, and your next RAM kit still costs five times what it did last year. Those three facts are the same story.

A QuasarZone-relayed report making the rounds in November 2026 put Samsung's Q3 2026 DRAM revenue share at 38% (first) and SK Hynix at 29% (second). That's real, and it traces back to Counterpoint Research's "Q3 2026 Global Memory Tracker," released November 5, 2026. But the headline share numbers hide the part buyers actually care about: which segments each vendor's capacity is going to, and what that means for whether you can find DDR5 DIMMs, server RDIMMs, or HBM at any sane price.

Here's the verified state of the market, and how the 38/29/21 slice maps onto what you'll actually encounter as a buyer.

TL;DR

• Counterpoint Research (Nov 5, 2026) put Q3 2026 DRAM revenue at ~$200B, up 369% YoY — driven by price, not volume: bit shipments rose only ~4% while ASP hit $1.90/Gb.
• Samsung held #1 at 38% revenue share (vs 33% a year ago), SK Hynix was #2 at 29% (up from 25% in Q2, revenue +40%+ QoQ), Micron #3 at 21% (down from 24%), and China's CXMT at 9%.
• Samsung reclaimed share through two stacked engines: a fast HBM4 recovery (HBM4 revenue more than tripled QoQ per its earnings call) plus the commodity DDR5/LPDDR price surge (DDR revenue +500% YoY counterpoint-wide), which favors Samsung's large conventional output.
• SK Hynix's AP lagged because its long-term agreements (14 major customers, >half of 2027–28 capacity locked) were priced before the spot market exploded — TrendForce flags exactly this LTA structure as its weakness this cycle.
• For buyers: the segment determines the vendor. Micron is leveraging server RDIMM mix the hardest, HBM is allocation-only and SK Hynix still leads there, and consumer DIMMs are whatever leftover wafer the module houses can grab — hence the persistent retail shortage.

The Q3 2026 share table

All figures from Counterpoint Research's Q3 2026 Global Memory Tracker (via Yonhap News Agency, Korea JoongAng Daily, and Counterpoint's own insights post, all Nov 5, 2026):

Rank	Vendor	Q3 2026 revenue share	vs Q2 2026	vs Q3 2025
1	Samsung	38%	roughly flat (wire reports cite 38–39% in Q2)	up from 33%
2	SK Hynix	29%	up from 25%	down from 33%
3	Micron	21%	down from 24%	down from 26%
4	CXMT	9%	down from 10%	up from 5%
—	Others (Nanya et al.)	~3%	—	~3–4%

Context numbers worth keeping: global DRAM revenue hit an all-time high of roughly $200B for the quarter, up 369% year over year. Bit shipments grew only ~4% sequentially — revenue came from price. Average selling price reached $1.90 per gigabit. Counterpoint senior analyst Jeongku Choi called it a "pricing super-cycle, not yet a volume cycle," where buyers now "compete for supply rather than negotiating on price." (Counterpoint Research, Nov 5, 2026)

One nuance: Samsung's exact QoQ move is fuzzy — Counterpoint's own quarterly table lists Q2 2026 at 38%, while Yonhap wire copy says Q3 was "down slightly from 39%." I read that as roughly flat with a rounding disagreement; the YoY story (33% → 38%) is unambiguous.

Why Samsung regained share: two engines, not one

The naive read is "Samsung's HBM caught up." That's true but incomplete, because the other engine this quarter was commodity DRAM itself.

Engine 1: HBM4 recovery. On Samsung's Q3 earnings call, memory VP Kim Jae-jun said HBM4 revenue more than tripled from the previous quarter and that HBM4 now makes up ~60% of Samsung's total HBM revenue. Samsung also claimed it hit its goal of bringing HBM share in line with its total DRAM share (~38%) ahead of schedule — which implies Q3 HBM share in the mid-to-high 30s (the company didn't publish the exact figure). Independent confirmation is still pending: Counterpoint's last published HBM split is Q2 2026 — SK Hynix 50%, Samsung 33%, Micron 18% — with brokerages (KB Securities, Bernstein) projecting Samsung approaching 40% in Q4. Bernstein estimated Samsung's Q3 HBM revenue at ~$8.7B vs SK Hynix's ~$7.6B; that gap is an analyst estimate, not a company-reported figure, so treat it as directional. (Samsung earnings call via Aju Press; Bernstein via Bloomingbit; KB Securities via Herald Business)

The catalyst behind Samsung's HBM gain: SK Hynix's HBM4 ramp was slowed by Nvidia's Rubin accelerator timing and interface tuning, while Samsung improved HBM4 mass-production yield from below 60% to roughly 80%. (TrendForce via Chosunbiz; businesskorea) UBS even projects Samsung taking the HBM bit-share lead in 2027 — 41% vs SK's 39% — though that's a forecast, not a result.

Engine 2: commodity DRAM prices, which outgrew HBM. Counterpoint's Brady Wang noted record commodity prices: DDR revenue up ~500% YoY, LPDDR up ~400% YoY. HBM revenue nearly doubled YoY, but its share of total DRAM revenue actually fell because DDR/LPDDR grew faster. That favors Samsung, whose wafer base pivots most cleanly back to conventional DRAM — and removes SK Hynix's structural edge from 2025, when HBM was the only growing segment. A Sangmyung University professor cited by Aju Press made exactly this point: Samsung's Q3 surge may owe more to conventional memory pricing than to HBM alone.

And SK Hynix? It's fine — just contract-timed. SK Hynix posted record quarterly revenue of ₩104.62T and grew DRAM revenue more than 40% QoQ. Its share dipped over earlier quarters partly because its long-term agreements with 14 major customers (cloud providers included) covering more than half of its 2027–28 capacity were struck at pre-spike price levels. TrendForce's Q3 analysis says it explicitly: "SK hynix lags in price growth due to long-term contract structures." Counterpoint adds that HBM prices lagged commodity DRAM this year and are now catching up, with LTAs supporting prices going forward. In other words, SK's share haircut this quarter is largely an accounting timing effect, not a demand loss — its customers are requesting over 2x the volume it can supply, per its own sales chief.

What "enterprise/PC/smartphone prioritization" does to each segment

TrendForce's Q3 industry report lays out the squeeze cleanly: AI training and agentic AI demand pulls HBM, LPDDR5X, and high-capacity RDIMM to the front of the wafer queue; PC and smartphone demand is deliberately deprioritized because the buyers are priced out; and end-product buyers pivot to lower-capacity modules to control cost. Every vendor is making the same trade — Samsung's memory chief said outright they'll "raise the sales share of HBM, server DRAM, and SSDs by one notch" in Q4 — but the three big vendors aren't identically exposed:

• Micron shifted its product mix hardest toward high-margin server DRAM (TrendForce, Q3). It just began high-volume HBM4 shipments for a leading customer platform (its fiscal Q4 revenue was $54.2B). If you're procuring server RDIMMs through OEM/ODM channels, Micron allocation is most visible here.
• SK Hynix is strongest where contracts are locked: HBM for Nvidia-class accelerators and hyperscaler cloud. Its 2027–28 volumes across DRAM and HBM are "effectively sold out" per its call.
• Samsung spans both, which is why it's #1 in DRAM overall — largest conventional wafer base plus a recovering HBM4 line now feeding the same next-gen accelerators.

Why this maps to consumer availability (the part that hurts)

Here's the wedge: HBM packs many DRAM dies per stack and eats far more wafer capacity per delivered terabyte than a UDIMM does. Every wafer died to HBM or RDIMM is a wafer not becoming consumer RAM. That's not a side effect; it's the design of the cycle.

Concrete evidence from this quarter and the months around it:

• PC-grade DDR4 8Gb chips hit an all-time-high fixed transaction price of $27.50 in October, about 4x the $7 a year earlier (DRAMeXchange, via MoneyToday).
• Server-side is worse: 64GB DDR5 RDIMM contract price was $1,500 as of mid-September, 5.5x the $272 a year before (Seoul Economic Daily).
• A module maker's POV: Team Group's CEO said 16GB DDR5 contract chips climbed from $8.70 (Oct 2025) to $48 (Sept 2026) — more than +550% — with spot prices running 30–50% above contract, and warned 2027 will get "worse" with no downward price trend; Team Group expects balance no earlier than H2 2028 (Tom's Hardware, Nov 4, 2026).
• Retail confirms: a community price tracker of eBay DDR5 listings (Sheley.org) showed best-value new listings near $2.12/GB with a median around $7.75/GB in early November 2026 — versus well under $1/GB norms a year ago [third-party tracker, treat as a gauge, not an index].

And here's where vendor share intersects the retail shelf: market share by revenue says nothing about which brand of die lands inside a Corsair or Kingston kit. Module houses buy whatever wafer they can source at whatever price — often at spot, hence the 30–50% spot premium. The same Samsung that leads DRAM revenue is the same Samsung that's rationing consumer DDR5: its consumer-facing module prices in Korea doubled or tripled within two months this cycle, and distributors were limited to ~10 modules per DDR5 sku per shipment window (TheElec). So "Samsung regained #1" does not mean "Samsung DIMMs are easy to find" — quite the opposite. Revenue rank tells you where the wafer went, not what's in stock.

CXMT is the wildcard most buyers hope saves them, but Team Group's chair pushed back on that hope directly: even at nearly 10% share, CXMT's chips track market pricing closely, and high-end supply still routes through Samsung, SK Hynix, Micron, and Kioxia. Counterpoint expects CXMT's share to keep shrinking as HBM4 supply starts in Q4, since CXMT doesn't play in HBM. Budget relief isn't coming from China this cycle.

Wrap-up

• Share map, verified: Q3 2026 DRAM revenue share was Samsung 38%, SK Hynix 29%, Micron 21%, CXMT 9% (Counterpoint), in a ~$200B record quarter driven almost entirely by price — Samsung regained #1 on HBM4 recovery plus the commodity DRAM surge; SK Hynix's dip is largely LTA pricing lag, not lost demand.
• Buyer guidance by segment: server/RDIMM buyers will see Micron's mix shift most aggressively; HBM remains SK Hynix-led (50% share as of Q2) with Samsung closing fast on HBM4; consumer DIMMs are the residual allocation — whatever's left over at spot prices, so module brand tells you little about die brand or availability.
• Don't wait out the cycle: every vendor's own guidance puts tight supply at least through 2028 (Samsung) — SK Hynix says 2030, Citi says 2031 — and Team Group expects no meaningful PC-side relief until 2028–2029. If a build or fleet refresh can wait, waiting costs nothing yet, but don't bet on H1 2027; if you must buy soon, lower-capacity modules (TrendForce's noted buyer pivot) are the current escape hatch.

Sources: Counterpoint Research's "Q3 2026 DRAM Super-Cycle Nears $200B" and its quarterly DRAM/HBM share tracker; Yonhap News Agency's wire report on Q3 share; Korea JoongAng Daily coverage; TrendForce's 3Q26 DRAM Industry Analysis; Aju Press on Q3 HBM race; Bloomingbit on Bernstein's HBM estimates; Chosunbiz on SK Hynix HBM4 delays; MoneyToday on PC DRAM pricing; Seoul Economic Daily on server DRAM pricing; Tom's Hardware on Team Group's shortage warning; TheElec on Samsung consumer module pricing; The Register's Q3 revenue summary; and the DDR5 retail price tracker.