Top Three Original Makers Suspend DDR5 Contract Quotes: Supply-Cut Signals, a Spot-Price Spike, and a Q4 Build Alert#

Zixing | October 1, 2026

Hello everyone, I’m Zixing.

The build-cost series so far has covered the overall build ledger, GPUs and storage, monitors, handhelds, and the DDR4 life-extension plan — but the previous pieces observed demand-side and price-side phenomena: retail prices rising, alternative options coming back, increases spreading to more categories. This time is different: the sellers have simply stopped quoting prices.

Per Taiwan’s DigiTimes citing supply-chain sources (re-reported by GMT EIGHT on September 14, 2026): Samsung Electronics has taken the lead in suspending DDR5 contract quotes for October 2026, with SK hynix and Micron following, and quote resumption expected to slip to mid-November. Samsung’s message to downstream customers was blunt — “nothing to sell.” With all three original makers suspending quotes in unison, the storage market has escalated from “price rises” to “supply cuts”; that change in nature is the core of this article.

Scope note first: as of publication, Samsung, SK hynix, and Micron have issued no official announcements; every event statement here follows “per supply-chain sources / Taiwan media reporting.” Specific figures involving vendor capacity and yield are mostly second-hand media and supply-chain channels, labeled one by one. Also, during research we found that an almost identical-sounding quote-suspension event occurred in October 2025; this article keeps them strictly separate, using the historical data only for comparison and never mixing it with this episode.

Table of Contents#

1. The Event on a Timeline: Samsung’s ‘Nothing to Sell’ and a Collective Quote Suspension
2. The Price Vacuum: Why Suspending Quotes Is Worse Than Raising Prices
3. Supply-Side Dissection: The HBM4 Squeeze and Long-Term Lockups
4. The Price Transmission Chain: Three Layers of Amplification from Suspension to Retail
5. Q4 Build Decisions: Budget-Tiered Timing Advice
6. Risk Notes: Quote Timeliness and Information Screening

1. The Event on a Timeline: Samsung’s ‘Nothing to Sell’ and a Collective Quote Suspension#

First, pin the event to a timeline. The public source of this suspension is GMT EIGHT’s September 14, 2026 re-report of Taiwan’s DigiTimes, cross-confirmed with identical wording by Eastmoney’s September 26 storage-industry daily.

Event timeline:

• September 14, 2026: GMT EIGHT re-reports DigiTimes’ first story — Samsung Electronics takes the lead in suspending October DDR5 contract quotes; SK hynix, Micron, and other storage original makers follow.
• Around the same time: Samsung directly told downstream customers “nothing to sell” — the most striking sentence of this episode: not “pricier,” but “no stock.”
• The following week: DDR5 spot prices spiked about 25%.
• Market expectations dashed: downstream had expected Q4 contract prices to be settled by the end of October; with Samsung dragging on quoting, Q4 contract talks fell into a “price vacuum,” and downstream module makers and system integrators could only rush to the spot market and pay up for stock.
• Resumption point: quotes expected to resume in mid-November — note, this is a supply-chain estimate, not any original maker’s official schedule.

The three makers’ actions and information bases in this episode, tabulated:

Why “Not Quoting” Is a Landmark Event#

In years of hardware analysis, the storage original makers’ routine has been a cycle of “raise prices — cut orders — cut prices — revive production,” and however tense contract talks got, there was always a quote. This time is a collective refusal to provide a price benchmark: Samsung suspended quotes on “nothing to sell,” with hynix and Micron following, meaning the entire contract market has lost its pricing anchor.

Another notable signal: per DigiTimes, in Q4 the upstream makers are quoting only tech giants or first-tier cloud providers, with almost no DDR5 capacity released to ordinary customers — storage “has fully entered a seller’s market.” The supply chain also judges that the three makers’ allocation will become stricter, quoting only long-term customers — “no quotes from the original makers” may shift from a one-off event to the norm. If that holds, the same script playing out in October 2025 and again in October 2026 is not coincidence but the routine state of a new supply structure.

Year Disambiguation: The 2025 Episode and This One Are Not the Same Event#

A necessary research warning: the script “Samsung suspends October DDR5 contract quotes, hynix and Micron follow, resumption in mid-November, spot up 25% in a week” ran once in October-November 2025 (a DigiTimes report of November 3, 2025, reprinted by Sina Finance and others), and ran again this September (per DigiTimes). The two episodes’ wording is highly similar, but their data must never be mixed:

All data in this article follows the 2026 event; 2025 data appears only in the historical comparison above. The same script repeating one year apart is itself a footnote that this storage supercycle continues — the capacity-siphon logic we unpacked in the GPU and storage price-surge piece has not been broken to date.

2. The Price Vacuum: Why Suspending Quotes Is Worse Than Raising Prices#

A price rise and a quote suspension are two entirely different market states. When prices rise, the market at least knows “how expensive,” and both sides can still bargain over a benchmark; after quotes stop, the contract market enters a “price vacuum” and pricing power migrates wholesale to the spot market. This section unpacks the mechanism.

Collective Action at an 87% Capacity Share#

Per GMT EIGHT (as relayed by Xueqiu; single source), the three original makers control about 87% of global DRAM capacity. That concentration means that when the three act in lockstep, the contract market has no alternative path of “bypassing the three for supply” — the remaining 13% (including newcomers like CXMT) cannot fill the gap in the short term. That is why a unilateral suspension by any one of the three would merely signal higher prices, whereas a collective suspension becomes outright “fasting.”

The Downstream Plight: From Negotiating Prices to Grabbing Spot#

The worst-hit in a price vacuum are downstream module makers and system integrators. Per supply-chain messages relayed by Xueqiu, the market had expected Q4 contract prices to be settled by the end of October; now the talks have no counterparty, and downstream has two options:

1. Rush to the spot market and pay up — that is exactly where the roughly 25% one-week spot spike came from; the more buyers scramble, the more prices lose control;
2. Accept the original makers’ allocation priority — tech giants and first-tier cloud providers first, with ordinary customers queuing or receiving no quote at all.

Module-maker feedback confirms the tightness: Chen Qingwen, general manager of Team Group under ADATA, said original-maker capacity is locked by long-term contracts and AI demand, leaving limited actually-available supply, and maintained the judgment that supply will be even tighter in the first half of 2027; the original makers have also been sending customers “friendly notices” flagging tight 2027 supply (per Digitimes reporting). Note the phrase “friendly notice” — the original makers warning a year ahead shows tight supply is not a one-off event but a long-term arrangement written into capacity plans.

3. Supply-Side Dissection: The HBM4 Squeeze and Long-Term Lockups#

The direct cause of the suspension is “nothing to sell” — so where did the stock go? This section does mechanism analysis on the public facts. A caveat first: the specific figures in this section (yield, long-term-contract ratios, capacity shares) all come from second-hand media and supply-chain channels, not official vendor data, and each citation is tagged “per supply-chain sources.”

The Wafer Economics of HBM4#

Per supply-chain sources, a single HBM4 die consumes about 3 times the wafer area of commodity DRAM. This multiplier is the key to understanding the whole supply crunch: from the same wafer, HBM4 yields only about one-third the output of ordinary DDR5, while HBM sells to AI data centers as high-margin orders — the original makers have no reason to leave premium capacity to low-margin consumer products.

Three key mechanism points:

• Capacity priority: the big three direct premium capacity to high-margin HBM first; per supply-chain sources, Samsung’s HBM4 yield has risen to about 70% — a yield ramp means HBM4 effective supply keeps expanding, and the displacement of commodity DRAM capacity will only intensify.
• Long-term lockups: per supply-chain sources, Samsung has locked about 70% of its storage capacity through 2031 to major AI customers via long-term contracts. A long-term contract is essentially a “capacity future”: capacity is sold before it is even produced, and the share left for the spot and ordinary contract markets naturally shrinks.
• Customer tiering: in Q4 the upstream makers quote only tech giants or first-tier cloud providers; ordinary customers are off the quote list — cause and effect intertwined with the downstream “spot scramble” in Section 2.

Tabulating this allocation logic:

Same-Period Industry Moves: Where the Money and People Are Going#

To judge the supply side’s medium-term direction, the original makers’ capital moves say more than prices. Read together, these late-September 2026 items carry a lot of information:

• SK hynix’s Solidigm plans a US IPO as early as 2027 at a valuation of about 100-150 billion US dollars (per Reuters) — storage assets commanding that valuation show capital markets’ strong confirmation of enterprise storage’s long-term boom.
• Samsung and SK hynix are competing for Sony’s next-generation PS6’s 24Gb GDDR7 memory order (per DigiTimes) — even next-generation console memory has become a high-end order the original makers fight over, pushing consumer capacity further down the priority list.
• China’s CXMT officially announced mass production of its G5 DRAM platform on September 20, raising output per wafer by 50% — a new capacity variable has appeared, but from announcement to actual volume is a long cycle, and the supply-demand picture will not change in the short term.
• Samsung completed the industry’s first 16GB LPDDR6 validation with an I/O rate of 10.7Gbps — high-end mobile memory keeps marching to newer generations, and competition for leading-node capacity will only intensify.
• The US International Trade Commission (ITC) opened a Section 337 investigation into DRAM equipment on September 23; YMTC won its 3D NAND patent case in Munich on September 18 — friction at the equipment and patent levels shows the supply-side contest has extended to the legal battlefield.

These moves point to one conclusion: the original makers’ capex and premium capacity are both concentrating toward high-margin enterprise and AI directions, and consumer DDR5’s position on the capacity allocation sheet will keep sliding back.

4. The Price Transmission Chain: Three Layers of Amplification from Suspension to Retail#

With the supply side understood, see how prices transmit. Storage prices have three layers: contract prices (bulk transaction prices between original makers and major downstream customers), spot prices (immediate transaction prices among intermediaries and channels), and retail prices (the stick prices we see when building). The three layers transmit in a clear order, and the suspension shock amplifies along exactly this chain.

Chain dissection:

1. Contract layer (vacuumed): with the three original makers not quoting, Q4 contract talks lost their counterparty, and the benchmark the market expected by end-October failed to appear. Before the suspension, the market’s prevailing expectation for Q4 DRAM contract increases was 3%-8%; that expectation is now likely to be breached.
2. Spot layer (pricing power amplified): with contract pricing absent, the spot market becomes the only price-discovery venue. After Samsung’s “nothing to sell” message, DDR5 spot rose about 25% in a week; Bank of America expects September DRAM and NAND spot prices still have 10%-20% of upside, and current channel supply fill rates are below 50% — with over half of channel demand unmet, spot prices have no mechanism to turn back.
3. Retail layer (lagging catch-up): retail memory and system quotes adjust with a lag to upstream moves. The specific latest-week retail increase was compiled in the site’s previous piece on a measured basis and is not repeated here; only the direction of transmission is certain — contract and spot increases eventually land on the build invoice.

Consolidating the citable key figures of this episode, each tagged with its data nature (this table is the article’s data foundation and the basis for the build advice that follows):

Two readings:

• The increase is spreading to more categories. Mobile memory up over 80% quarter over quarter, DRAM cost exceeding SoC for the first time as the single most expensive component in flagship phones, automotive-grade storage up 180% in three months — DRAM and NAND tightness is no longer confined to the build market; it is happening simultaneously in every end device that eats storage. The briefing also mentions enterprise SSD contract prices catching up; this article’s research verified only the general direction of NAND increases and did not obtain a specific percentage for October 2026 enterprise SSD contract prices, so the judgment here stays at trend level: enterprise SSDs share the original makers’ capacity and the seller’s-market logic with DRAM, so catch-up increases are mechanically inevitable; the specific magnitude will be tracked in a later piece once contract prices settle.
• Consumer tolerance is near its limit. CINNO Research data shows Q3 mobile LPDDR increases narrowed to 3%-7%, leaving phone makers less and less room to raise prices — which in turn means the original makers’ incremental supply will flow even more preferentially to enterprise customers with deeper pockets, and consumer supply will see no improvement in the short term.

5. Q4 Build Decisions: Budget-Tiered Timing Advice#

Principles first, then the timing table. My personal framework is simple: building is demand-driven and should not be hostage to price swings; but when the price inflection is unclear, the payment timing can be optimized. The advice below rests on this cycle’s verified supply facts; “editor judgment” is labeled explicitly.

Three core judgments:

1. Memory: postpone if you can. Mid-November quote resumption is a supply-chain estimate; only after resumption will Q4 contract prices have a benchmark and retail regain an “anchor.” Until then, retail follows spot, and spot can swing 25% in a week — buying memory at retail now means paying a double premium (spot premium plus panic premium).
2. CPU: the low-price channel window may be October. Market opinion holds that CPU channel prices may hit their last low in October before Q4 increases — note this is market opinion and editor judgment; this round’s research found no independent source for cross-validation, so it is not established fact — but the logic holds: as storage’s share of system costs rises, the whole BOM’s pricing expectations shift upward. Readers with a platform upgrade planned will most likely find better value locking CPU and motherboard in October than waiting for the depths of Q4.
3. Brand prebuilts: usually rise less than parts, an overlooked alternative route. Mechanistically (inference): first-tier system makers are the original makers’ long-term customers and, under a structure where “original makers quote only tech giants, first-tier cloud providers, and long-term customers,” still access contract-price channels, while the parts market eats spot prices only. In a seller’s market, the prebuilt-versus-parts gap passively narrows, and budget-sensitive players should add brand prebuilts or barebones to the comparison.

The budget-tiered timing table follows (the basis column maps to the data foundation above):

Beyond timing, capacity strategy can also be tiered to this supply picture. One principle: before contract prices regain a benchmark, buy “enough” rather than “redundant” — in a market swinging 25% a week, the extra capacity is priced at panic premium. The three capacity tiers:

Two route reminders regular readers know: for very tight budgets, the AM4-plus-DDR4 life-extension plan still stands — the price-gap logic and pitfall checklist are fully covered in the DDR4/AM4 comeback piece and not repeated here; on the GPU side, the shrink-wrap spiral and advice in the previous piece also stand — combined with this article’s memory timing, they form a complete Q4 build decision sheet.

6. Risk Notes: Quote Timeliness and Information Screening#

This final section is for every reader planning to take this article’s numbers comparison-shopping; these points matter more than the conclusions:

• Quote timeliness: the spot increase cited here (about +25% in a week) is a reporting-time figure; spot swings violently intraday, and the actual quote on your order day may differ markedly. Any build decision should be re-verified against same-day real-time quotes; do not extrapolate today’s price from a week-old increase.
• Contract prices unsettled: Q4 contract prices remain in a vacuum; every increase expectation cited (the old 3%-8%, the 30%-50% quarterly speculation, the 30-dollar speculative quote) is an expectation or speculation, not a transaction price. Wait for mid-November resumption and defer to actual contract prices.
• The resumption point is an estimate: “quotes resume mid-November” comes from supply-chain estimates, not any original maker’s official schedule; it could slip either way and should not anchor procurement as a firm date.
• No official announcements: Samsung, SK hynix, and Micron have issued no official confirmations, and media inquiries received no direct response. Every event statement here is limited to “per supply-chain sources / Taiwan media reporting”; should any vendor speak officially, defer to that.
• Data-source tiering: the numbers here fall into four tiers — multi-source cross-confirmed media reporting (the +25% spot move and the suspension event itself), institutional views (BofA, Counterpoint, CINNO), supply-chain relays (the 87% capacity share, 70% HBM4 yield, 70% long-term lockups, all second-hand single sources), and speculation (quarterly increases, the 30-dollar quote). Note the tiering when citing; relayed data carries distortion risk.
• Year basis: to reiterate, this article covers the October 2026 contract-quote suspension; the 2025 same-script data (Q3 2025 DRAM up +171.8% year over year, October 2025 DDR5 16Gb up +102% in a month) is historical comparison only — do not mix it with this episode in comparisons or discussions.

Closing Thoughts#

That this storage cycle has reached “original makers suspend quotes” shows something that used to live only in textbooks is happening: in an industry with extreme concentration, pricing power can withdraw wholesale from the contract market during shortages. The same script in 2025 and again in 2026 is less coincidence than the new normal after AI demand restructured capacity allocation — the consumer market has gradually become “the receiver of leftover capacity” rather than the original makers’ core business.

My advice to builders remains restrained: no panic, no hoarding, sequence by need. Wait for the post-mid-November contract benchmark on memory if you can; if not, buy the requirement without chasing highs; the CPU low window (if market opinion holds) is October; and if parts prices have scared you off, put brand prebuilts on the comparison list. A PC is ultimately for using — don’t let the market table make the decision for you.

I will follow up on the specific enterprise SSD contract increase and the Q4 contract settlement once the data is verifiable. I’m Zixing — see you in the next measurement piece.