Memory Price Reversal 2027: Acer vs. Big 3 Debate

The argument, in one minute

The memory industry is having a public argument about its own future — and buyers are caught in the middle. On one side, Acer chairman and CEO Jason Chen says the shortage is already over, Chinese capacity is flooding in, and memory prices will start coming down in the second half of 2027. On the other side, SK hynix is reported as warning that the shortage gets worse next year with pricing normalizing only around 2030, and Adata's chairman is reported as suggesting DRAM tightness could last another decade. (Reported via Taiwan financial press and ZeroHedge citing Tom's Hardware; no primary company statement located as of October 1, 2026.) Both sides cannot be right about timing, but both can be right about the forces they describe. This article lays out each side's evidence, then gives you the only thing that matters for procurement: what to do under each scenario. As of October 1, 2026, neither outcome is certain — so the smart move is to plan for both. This article takes no side and makes no deterministic prediction.

What Jason Chen actually said (September 19, Taiwan tech forum)

Chen spoke to reporters at a technology forum in Taiwan on September 19, and his message was blunt. Only high-end DDR5 parts such as LPDDR5X-9600 and niche CPUs like Nvidia's N1 and N1X remain in short supply, he said; the broader memory market no longer faces serious supply constraints. His words, as reported by Taiwan financial press (BigGo Finance relay): with Chinese capacity continuing to come onto the market, "the supply shortage problem has already completely disappeared." (Sources: BigGo Finance, 2026-09-19~20; Sohu/ITHome, 2026-09-20.) His evidence rests on three pillars:

• Chinese capacity is the price breaker. Chen called Chinese memory makers "price disruptors" whose continued expansion makes it hard for prices to stay elevated. The centerpiece: CXMT, China's largest DRAM maker and now the world's fourth-largest behind Samsung, SK hynix, and Micron, announced that its fifth-generation DRAM technology platform had entered mass production — a platform the company says produces at least 50% more gross dies per wafer than its previous generation, and on which it is already mass-producing two 24-gigabit LPDDR5X products. Acer, HP, and Asus have all started using CXMT chips in some products, and some Lenovo models sold in Germany have been found carrying YMTC SSDs. (Sources: Reuters, BEIJING, Sept 20, 2026, by Che Pan and Eduardo Baptista, via ZeroHedge 2026-09-21; BigGo Finance.)
• The shortage is already uneven — and partly over. Chen drew a distinction buyers will recognize: LPDDR5 and high-end DDR5-9600 remain tight, but older DDR4 faces the opposite problem — "more sellers than buyers." CPU supply, outside a few entry-level and mid-range models tied to specific projects, has largely normalized. Acer itself still carries high memory inventory, and Chen expects the memory and SSD costs Acer holds in 2027 to come in below this year's levels. (Sources: BigGo Finance; Sohu/ITHome.)
• Suppliers have a commercial motive to talk up scarcity. Chen's sharpest claim was about incentives: memory makers, he suggested, naturally want to keep margins high for as long as possible, and because antitrust rules prevent them from coordinating prices directly, they use public statements as a kind of "signal" to one another — "let me tell you, prices won't come down until the year 20-whatever." (Source: ZeroHedge relaying UDN, 2026-09-21.)
• His timeline, in his own framing: PC prices rise 5–20% toward the end of 2026, plateau — possibly rising further — through the first half of 2027, with component prices peaking around mid-2027 before starting to ease in the second half. But he was explicit about the uncertainty: "Nobody can know exactly when the reversal comes. A fall in PC prices is strictly a hope." (Sources: BigGo Finance; ZeroHedge.)
• One more warning from Chen that buyers should not miss: even as he argues memory tightness is fading, he noted that price increases are becoming their own trend across other components — SSDs, PCBs, and the fiberglass cloth used in motherboards are all seeing hikes. The memory debate is not the whole cost story. (Source: ZeroHedge.)

What the other side says — and their evidence

The opposing camp's case is structural, and it comes with numbers:

• SK hynix, as reported secondhand, warns the shortage will be worse next year and that pricing may not normalize until around 2030. Adata's chairman, likewise via press reports, has suggested the DRAM shortage could persist for another ten years. Their shared foundation: AI hyperscaler demand for high-bandwidth memory (HBM) remains strong, and the new fabs under construction are not expected to come online until the 2030s. (Source: ZeroHedge, 2026-09-21, citing Tom's Hardware.)
• TrendForce, the most-cited independent referee, sits closer to the suppliers' camp on timing: it expects DRAM supply tightness to extend into 2027, driven by AI server and HBM demand outpacing supply expansion. TrendForce's latest research, as reported by Commercial Times on October 1, 2026, put numbers on the near term — conventional DRAM contract prices up 10–15% quarter over quarter in Q4 2026, NAND flash up 15–20% — with enterprise SSD the one category where price increases are expected to accelerate. For perspective on how violent this cycle has been: TrendForce's data shows conventional DRAM contract prices surged 90–95% in a single quarter (Q1 2026), and retail DDR5 module prices in Japan ran more than five times their pre-surge levels (JEITA data via BigGo Finance). (Sources: BigGo Finance relaying TrendForce; Commercial Times/Economic Daily News, 2026-10-01.)
• Micron's behavior is the suppliers' argument made in cash. The company has signed 26 strategic customer agreements with financial commitments now totaling $32 billion — mostly cash deposits — and says more than 75% of its 2027 output is already allocated to customers under and outside those agreements, with most customer discussions already shifting to 2028 supply. Buyers do not hand over billions in deposits years in advance for a shortage that is "already over." (Source: Micron fiscal Q4 2026 earnings call, 2026-09-30 — per call transcripts, e.g. MarketBeat; results confirmed in Micron's official press release.)
• As Chen himself argues, note the asymmetry: the people warning of a decade of tightness sell memory; the person calling the top buys it. That does not make either side wrong — it means you should weigh each claim against checkable facts, which is what the next section is for.

Scenario A — Chen is right (reversal starting H2 2027)

What has to be true: CXMT's fifth-generation platform ramps as claimed and its output actually reaches the open market rather than being absorbed domestically; AI demand growth cools enough that HBM stops cannibalizing conventional DRAM capacity; and the inventory sitting in channels (Acer says its own stocks are still high) bleeds out without a restocking panic. What you do:

• Do not lock long at the top. If you are being offered multi-year fixed-price agreements now, compare them against Chen's timeline before signing. A three-year lock signed at a 2026 peak is the most expensive mistake this scenario punishes.
• Keep inventory lean and buy in tranches. Cover confirmed demand, not forecast hope. In a falling market, every extra week of stock is a writedown.
• Get a second quote from the China-capacity chain. Whether or not you buy Chinese-origin memory, CXMT/YMTC-based alternatives are now a legitimate pricing lever — distributors carrying those lines will sharpen everyone else's pencil. Ask your suppliers explicitly what they can do on those options.
• Watch DDR4 as the canary. Chen's "more sellers than buyers" description of DDR4 is checkable today. If DDR4 spot keeps softening while DDR5 holds, the bifurcation he describes is real — and it tells you which BOM lines to negotiate hardest.

Scenario B — The Big 3 are right (tightness into 2028–2030)

What has to be true: HBM and AI server demand keeps absorbing the capacity that would otherwise serve conventional DRAM; new fabs genuinely cannot deliver meaningful volume before the late 2020s; and the deposit-backed long-term agreements (Micron's $32 billion is the template) keep an ever-larger share of output off the open market. What you do:

• Lock price where you can, on your highest-volume lines. Multi-quarter or annual agreements with defined pricing bands are exactly what the hyperscalers are doing — at buyer scale, the equivalent is getting written price validity and allocation commitments from your distributors now, not in Q1 when everyone is scrambling.
• Dual-source every memory line on your BOM. If one source allocates you, the other keeps the line running. This is also the moment to qualify alternates (including CXMT/YMTC-based modules where your quality process allows) — qualification takes months, and you cannot start it during a stockout.
• Front-load 2027 coverage. With most 2027 output reportedly already spoken for at the top of the market, small and mid-size buyers who wait for spot availability will be choosing from leftovers. Place the RFQs for next year's memory sooner than feels comfortable.
• Budget the rest of the BOM too. Chen's warning cuts across both scenarios: SSDs, PCBs, and laminate materials are rising on their own track. A memory-only sourcing plan misses the inflation sitting in the rest of your board.

How to tell which scenario is winning (signals, not predictions)

You do not need to pick a side today. You need a dashboard. Check these four signals monthly; they are the earliest places where either scenario shows up in data rather than in quotes:

• Contract-price momentum (TrendForce quarterly). The Q4 2026 forecast is +10–15% for DRAM, +15–20% for NAND. If the Q1 2027 forecast prints flat or the increase keeps narrowing, Scenario A is gaining. If enterprise SSD keeps accelerating, Scenario B is.
• DDR4 vs. DDR5 spot divergence. Persistent DDR4 softness alongside firm DDR5 supports Chen's bifurcation story. If DDR4 firms back up, the "shortage is over" claim is losing.
• CXMT output reaching your suppliers. Ask distributors directly whether CXMT-based modules are quotable in volume with stable lead times — not press releases, but actual quotes. That is the moment Chinese capacity becomes your pricing lever.
• Hyperscaler capex and memory deposit behavior. More $32-billion-style deposit deals mean the top of the market is still buying insurance against Scenario B. Watch earnings-call commentary, not headlines.
• Posture note: both scenarios agree on the next two quarters — prices stay high through at least mid-2027. The disagreement is entirely about what happens after. So the near-term actions (written price validity, dual sourcing, qualifying alternates) are identical under both scenarios. Do those now; let the signals above tell you which long game to play.

Frequently asked questions

Quick answers to the questions buyers ask most about the memory outlook.

• Q: Did Acer's CEO really say the memory shortage is over? A: At a Taiwan tech forum on September 19, 2026, Acer chairman Jason Chen said that with Chinese capacity continuing to enter the market, "the supply shortage problem has already completely disappeared" — while noting high-end DDR5 (e.g., LPDDR5X-9600) and certain CPUs remain tight. He expects component prices to peak around mid-2027 and ease in the second half, but called the timing "strictly a hope." (Sources: Economic Daily News via BigGo Finance; UDN via ZeroHedge.)
• Q: What do SK hynix and other suppliers say? A: SK hynix is reported (via press relays) as warning tightness worsens next year with normalization only around 2030; Adata's chairman is reported as suggesting DRAM tightness could last another decade; TrendForce expects DRAM supply tightness to extend into 2027 on AI server and HBM demand. (Sources: ZeroHedge citing Tom's Hardware — secondhand relay; TrendForce via Commercial Times, 2026-10-01.)
• Q: Why is CXMT central to this debate? A: CXMT, China's largest DRAM maker (world's fourth), announced mass production of its fifth-generation DRAM platform, claiming at least 50% more dies per wafer than its prior generation, with two 24Gb LPDDR5X products already in mass production. Acer's CEO calls Chinese makers "price disruptors"; SK Group's chairman Chey Tae-won has publicly expressed concern about the competitive threat from Chinese suppliers. (Sources: Reuters via ZeroHedge, 2026-09-21; BigGo Finance.)
• Q: Should I sign a long-term memory supply agreement now? A: It depends on which scenario you believe — that is the point of this article. Under a reversal scenario, long locks at peak prices are the costliest mistake; under an extended-tightness scenario, allocation commitments are protection. Either way, get price validity in writing, dual-source each line, and start qualifying alternates now.