The memory trade has been one of the strongest semiconductor themes of 2026.
AI data centres have driven extraordinary demand for memory and storage, while limited supply has pushed NAND and DRAM prices sharply higher. That combination helped send memory stocks to some of their strongest gains in years.
This week, though, the trade showed a fresh sign of stress.
$SanDisk Corp.(SNDK)$ fell 6.8% on Thursday after investors reacted negatively to its outlook, while $Western Digital(WDC)$ dropped 13%. SK Hynix lost around 5%. $Micron Technology(MU)$ initially fell more than 7% before recovering most of the decline and closing just 1.3% lower.
At the same time, J.P. Morgan’s more cautious view on the 3D-NAND cycle has raised another concern:
If expectations for NAND start coming down, could analysts eventually lower forecasts for the wider memory sector too?
That is particularly relevant for $Micron Technology(MU)$.
But the answer is not as simple as “NAND down = Micron down.”
🐯🪙 Do you think this is just a reset after a huge rally, or an early warning for memory stocks? Share your view below — thoughtful comments may receive Tiger Coins.
📉 Why Did Memory Stocks Sell Off?
The earnings themselves were not weak. The problem was that the market had already priced in exceptional conditions.
$SanDisk Corp.(SNDK)$ and $Western Digital(WDC)$ both reported solid results, but guidance did not clear the very high bar set by investors. $SanDisk Corp.(SNDK)$ guided gross margin to 83%–85%, slightly below the prior quarter’s 84.6%, while investors also questioned whether the extraordinary pace of NAND pricing and margin expansion can continue.
After a large rally, “good” numbers are often not enough. The sector had been priced for tight supply, rising prices, margin expansion and repeated earnings upgrades. Any sign that one of those drivers is slowing can trigger a sharp reset.
The Fundamental Story Is Still Strong
There is little evidence that memory demand has collapsed. In fact, the pricing backdrop remains unusually strong.
Industry estimates point to exceptionally strong memory pricing in 2026. Conventional DRAM contract prices were projected to rise around 90%–95% QoQ in Q1, while NAND Flash prices were expected to increase 55%–60%. For Q2, the estimates pointed to further gains of roughly 58%–63% for DRAM and 70%–75% for NAND Flash.
These numbers show that the fundamental backdrop remains strong. But they also explain why investor expectations have become so difficult to beat. When pricing has already risen this quickly, the market starts focusing less on whether prices are still increasing and more on whether that pace can continue.
Why the 3D-NAND Call Matters
One analyst call does not change an industry cycle. But after a major rally, it can change expectations.
A more cautious Wall Street view on the 3D-NAND cycle has added to concerns that expectations for NAND pricing and margins may be approaching a peak.
|
NAND price assumptions ↓ |
Revenue estimates ↓ |
Margins / EPS ↓ |
Price targets ↓ |
Sector multiples reset |
💾 Why 3D-NAND Is Important
3D-NAND is the core storage technology behind products such as:
-
enterprise SSDs;
-
cloud and data-centre storage;
-
PCs;
-
smartphones;
-
automotive storage.
AI has significantly increased demand for enterprise storage because increasingly large datasets need to be stored and accessed alongside GPUs and servers.
That demand helped tighten the NAND market.
$Micron Technology(MU)$ said in March that NAND demand was “significantly in excess” of its available supply for the foreseeable future, after data-centre NAND revenue more than doubled sequentially.
So the bearish argument is not:
“Nobody needs NAND anymore.”
It is:
“Can NAND pricing continue improving quickly enough to justify current earnings expectations?”
Those are very different arguments.
Could Micron Be Next?
Possibly — but not automatically. $Micron Technology(MU)$ is not a pure NAND trade. Its earnings exposure is spread across NAND, conventional DRAM and HBM, with HBM now one of the most important growth drivers of the AI buildout. Micron’s latest filing shows why the comparison matters: in fiscal Q3, DRAM generated about US$31.3 billion of revenue, versus US$9.9 billion from NAND.
Micron’s current AI story is heavily tied to DRAM and HBM, not NAND alone.
Micron has already begun volume shipments of HBM4 designed for Nvidia’s Vera Rubin platform.
That gives Micron a different exposure profile from Sandisk.
This may help explain Thursday’s price action.
Micron initially followed the sector lower but recovered sharply, while NAND- and storage-heavy names remained under much greater pressure.
That divergence matters.
Micron’s much smaller closing decline may suggest that investors are beginning to distinguish NAND-heavy storage names from companies with greater DRAM/HBM exposure.
But if Micron begins falling alongside them, that would suggest investors are reassessing the broader memory trade.
The Bull Case Has Not Disappeared
Micron still describes memory supply as tight, particularly in NAND and AI-linked DRAM/HBM.
It is also moving toward multi-year Strategic Customer Agreements with binding volume commitments. When its planned agreements are completed, Micron expects half or more of company revenue to be covered by SCAs, giving it greater demand and pricing visibility than in previous memory cycles.
That does not eliminate memory cyclicality, but it may reduce some of the extreme volatility historically associated with spot-driven supply cycles.
👀 What Traders Should Watch Now
1. More analyst revisions: One cautious call matters less than a trend. Several banks lowering NAND pricing, margin or EPS assumptions would be a stronger signal.
2. NAND pricing: Watch whether prices are still accelerating, merely rising more slowly, flattening or falling. Those scenarios imply very different earnings outcomes.
3. Micron versus Sandisk / WDC: Relative performance can show whether the market sees this as NAND-specific or a broader memory reset.
4. HBM demand: If HBM remains structurally tight, Micron may retain an earnings buffer even if NAND expectations soften.
5. Supply expansion: The classic memory risk remains the same: high prices encourage new capacity, which can eventually weaken pricing power.
💬 What’s your view?
A. Just a healthy pullback after the rally
B. 3D-NAND is cooling, but DRAM/HBM remain strong
C. The broader memory trade is due for a larger correction
🐯🪙 Share your view and reasoning — thoughtful comments may receive Tiger Coins.
Comments
To me, this looks more like a healthy reset in expectations than the beginning of a new downcycle. After such a strong rally, memory stocks were priced for near-perfect execution, so even solid earnings and guidance weren't enough to satisfy investors.
The bigger question isn't whether NAND is slowing—it's whether that weakness spreads to DRAM and HBM. So far, AI demand hasn't changed. Hyperscalers are still investing aggressively, HBM supply remains tight, and AI servers continue to require more high-performance memory.
That's why I think Micron is in a different position from pure NAND players. Its AI growth is increasingly driven by DRAM and HBM rather than NAND alone. Unless we start seeing analysts cut DRAM/HBM forecasts or AI capex slows meaningfully, I'd view this pullback as a valuation reset rather than the end of the memory trade.
@AI_FocusedTrader [DOGE]
目前更像是存储内部开始分化,而不是整个内存周期同时转弱。NAND此前涨幅较大,一旦定价增速放缓,闪迪和WDC的盈利预期会更容易被下修;但HBM受AI服务器需求支撑,供需结构依然更紧,美光因此可能具备更强的盈利缓冲。
我的策略是暂时降低NAND链的追涨意愿,重点观察NAND合约价和分析师盈利修订,同时继续跟踪HBM订单与产能。如果DRAM/HBM仍保持强势,我不会把这轮调整定义成全面的存储周期见顶。