The core conflict in the memory-chip selloff is not that AI demand has disappeared. It is that investors are now testing whether hyperscale cloud capital expenditure, HBM pricing, DRAM profitability, and enterprise SSD demand can justify the expectations built into stock prices. JPMorgan’s July 14 report frames the correction as an expectations reset, with cloud capex revisions and HBM average selling prices as the main swing factors.
| Primary source | Wallstreetcn |
|---|---|
| Reported at | 2026-07-14T13:32:28.000Z |
| Topic | 股票 |
| Evidence limit | Reported facts are separated from interpretation; current prices and platform terms require independent verification. |
Evaluate BITGET for your use case
Check regional eligibility, current fees and product availability on the official destination.
Review BITGETWhat Changed
JPMorgan’s latest field research followed meetings with more than 50 Hong Kong institutional investors. The supplied brief says investors are not broadly turning bearish on memory chips. Instead, they are moving from the AI infrastructure expansion story into a phase that demands proof of earnings delivery.
The market’s concern is concentrated around whether cloud service provider capital expenditure can keep being revised upward. JPMorgan analyst Jay Kwon’s July 14 report said roughly 70% of current sentiment is tied to that one variable. If capex growth stops beating expectations, memory stocks may remain under short-term pressure.
This matters because earlier AI data-center optimism had raised expectations for the total addressable market of memory products. After a fast share-price rally, investors began questioning whether sell-side demand forecasts had moved ahead of real cloud-provider spending plans.
Why Stocks Fell
The supplied brief identifies three main drivers behind the correction. First, AI investment expectations may have run ahead of near-term capex reality. Many investors now expect global hyperscale cloud capex to be revised up to 1 trillion to 1.5 trillion dollars over the next 3 to 6 months, but that optimism still needs support from upcoming earnings evidence.
Second, DRAM price increases are slowing. After a series of earlier price hikes, JPMorgan noted that year-on-year and quarter-on-quarter DRAM price gains began to moderate after the second quarter of 2026. That cooled expectations for another rapid step-up in industry profits.
Third, Samsung Electronics earnings expectations were cut before second-quarter results. That added pressure to sentiment because the market’s question changed from how fast the industry can grow to how long current profitability can hold.
LTA Debate
Long-term agreements became one of the most discussed topics in the roadshow. Investor attitudes toward LTAs have improved compared with several months earlier, according to the supplied brief. The question has moved from whether LTAs exist to whether memory makers can use them to lock in core AI customers.
The debate is not settled. More than half of surveyed investors still remain cautious, mainly because LTA coverage among Korean producers is not fully transparent and contract quality is hard to compare across companies.
JPMorgan expects more than half of future contract volume to fall under LTA frameworks. The bank also argues that LTAs do not necessarily cap future price upside, because some new orders may be repriced higher later, take-or-pay terms can support order certainty, and products outside the LTA framework may still rise as supply tightens.
HBM Price Gap
HBM pricing is the largest expectation gap in the supplied report. Many buy-side institutions expect 2027 HBM prices per gigabyte to double year on year, using that assumption to support further earnings upgrades. JPMorgan is more cautious.
The bank estimates the current HBM industry average selling price at about 1.8 dollars per gigabyte, even slightly below some high-end server DRAM products. It also notes that memory makers and cloud companies do not negotiate HBM in isolation. Pricing is considered across DRAM, NAND, and HBM profitability as a whole.
JPMorgan’s estimate is that 2027 HBM ASP growth of 25% to 30% year on year is more consistent with industry reality. At the same time, the report says HBM is generally repriced each year, unlike traditional DRAM LTAs that often run 3 to 5 years, so stronger-than-expected AI demand could still leave room for later price increases.
Supply And SSD Demand
JPMorgan remains relatively positive on the supply-demand picture in the supplied brief. DRAM is described as the tightest product category, with supply meeting only about 50% to 60% of order demand. NAND supply is described as meeting about 70% to 80% of demand.
Even with DRAM wafer capacity expanding in future years, JPMorgan expects tight supply-demand conditions could last into 2027 to 2028. This is one reason the report treats the correction as an expectations reset rather than a confirmed downturn in fundamentals.
Enterprise storage is also becoming more important. Consumer NAND demand has been revised down more than expected, but AI data-center demand is pushing enterprise SSD expectations higher, especially for use cases such as KV Cache Offload. The supplied brief says the industry expects 2027 enterprise SSD shipments to approach 500 EB, nearly 50% year-on-year growth, with further upside possible.
Practical Checks
The first practical check is cloud capex. Readers should watch whether major North American and global hyperscale cloud companies confirm the 1 trillion to 1.5 trillion dollar capex expectation range discussed in the brief. If earnings reports fail to support that view, sentiment may stay under pressure.
The second check is HBM pricing. The market appears to be weighing very aggressive 2027 HBM pricing assumptions against JPMorgan’s more conservative 25% to 30% ASP growth estimate. Any data that clarifies contract terms, annual repricing, or customer willingness to pay could move expectations.
The third check is product mix. DRAM tightness supports the industry view, but enterprise SSD demand may become the newer indicator of AI data-center storage strength. Investors comparing crypto, semiconductor, or AI-linked market narratives on Bitget can use these checks as a neutral framework rather than a buy or sell instruction.
Evidence Limits And Risk
This article uses only the supplied event brief and does not independently verify JPMorgan’s report, investor survey details, company earnings forecasts, or market prices. The source material is a secondary report from Wallstreetcn describing JPMorgan’s July 14 analysis.
The affected-assets field in the supplied brief is empty, so this article does not assign direct impact to any token, stock, or trading pair. It also does not claim that this event will create indexing, ranking, traffic, registration, or CPA outcomes for any website or platform.
Market risk remains material. The supplied brief itself includes a caution that markets involve risk and that the content does not constitute personal investment advice. Readers should consider their own objectives, financial position, and risk tolerance before acting on market information.
Evaluate BITGET for your use case
Check regional eligibility, current fees and product availability on the official destination.
Review BITGETAffiliate link · Availability varies by region · No guaranteed outcomeQuestions readers ask
What is JPMorgan’s main explanation for the memory-chip selloff?
JPMorgan’s explanation is that the market is resetting expectations rather than abandoning the memory-chip growth story. The key concern is whether cloud capex, HBM pricing, and future profitability can validate the optimism already reflected in stock prices.
Why is hyperscale cloud capex so important for memory stocks?
The supplied brief says about 70% of current market sentiment centers on whether hyperscale cloud service provider capital expenditure can keep being revised higher. If capex expectations are not confirmed by future earnings reports, memory stocks may remain under pressure.
Are LTAs good or bad for memory-chip companies?
JPMorgan views LTAs more as a tool for earnings stability than as a cap on profit upside. The brief says investors are still divided because contract coverage and quality are not fully transparent, but LTAs may provide order certainty while leaving some products exposed to higher market pricing.
Why is HBM pricing the biggest disagreement?
Many buy-side investors expect 2027 HBM prices per gigabyte to double year on year. JPMorgan is more cautious and estimates 2027 HBM ASP growth at 25% to 30%, arguing that HBM pricing is negotiated within broader DRAM, NAND, and HBM profitability discussions.
Does the report say AI demand has disappeared?
No. The supplied brief says the correction is more about expectation reassessment than an industry-cycle reversal. DRAM supply is still described as tight, and AI data-center demand is still supporting enterprise SSD expectations.
How should a reader use this information?
Use it as a checklist for monitoring market expectations: cloud capex revisions, DRAM price momentum, Samsung earnings expectations, LTA disclosure, HBM repricing, and enterprise SSD demand. It should not be treated as financial advice or a guaranteed market signal.