To The Moon

Record Earnings—So Why Did Memory Stocks Fall Together?

Recently, $SNDK$，$WDC$，$MU$，$SKHY$，$STX$ all experienced sharp pullbacks.

Strictly speaking, Micron, SK hynix, and Sandisk are producers of DRAM, HBM, or NAND memory chips, while Western Digital and Seagate mainly focus on enterprise hard disk drives. However, under the broader AI data-center investment theme, the market often treats all five companies as part of the same “AI storage trade.”

This selloff does not mean that demand for AI storage has suddenly disappeared. Instead, it appears to be a broad repricing of the sector as elevated valuations, high expectations, and excessive leverage cooled simultaneously.

1. Macro Environment: Lower Rate Pressure, but Capital Rotated Out of Memory Stocks

U.S. nonfarm payrolls unexpectedly fell by 23,000 in July, far below market expectations for an increase of 80,000. This led investors to reduce expectations for a Federal Reserve rate hike in September. The U.S. 10-year Treasury yield fell to approximately 4.66%, while the Nasdaq gained 1.3% and the S&P 500 reached a record high.

Impact on memory stocks: Lower interest rates would normally support high-valuation technology stocks. However, memory stocks fell even as the broader technology market advanced. This suggests that the decline was not caused by a market-wide risk-off move, but rather by capital rotating out of memory stocks—which had already recorded substantial gains—and into software and other semiconductor companies.

In other words, the macroeconomic environment was the backdrop, not the main trigger for the memory-sector selloff.

2. The Main Reason the Sector Fell: Strong Earnings Were No Longer Enough

2.1 The Market Was Not Pricing in Growth—it Was Pricing in Continued Acceleration

Both Sandisk and Western Digital reported revenue and EPS above analysts’ consensus estimates. Their next-quarter guidance also exceeded some published market forecasts.

However, investors were expecting much more:

* NAND, DRAM, and enterprise HDD prices to continue rising sharply;

* Data-center revenue to maintain rapid sequential growth;

* Gross margins to reach new record highs;

* Next-quarter guidance to beat every major consensus estimate by a wide margin;

* Management to raise its longer-term earnings expectations further.

Therefore, the sector did not decline because the earnings were weak. It declined because the results failed to exceed the market’s much higher “implied expectations” or “whisper numbers.”

2.2 Valuations Had Already Priced In a Near-Perfect Scenario

Before earnings, SNDK had risen nearly 470% year to date, WDC had approximately tripled, and MU had at one point gained more than sevenfold over the preceding year.

When share prices already reflect booming AI demand, supply shortages, rising memory prices, and expanding margins, even strong results may not be enough. If the pace of improvement does not accelerate further, investors may reduce the valuation multiples they are willing to pay.

2.3 Investors Began Trading the Question: “When Will Memory Prices Peak?”

The market is not mainly concerned about current demand. It is looking several quarters ahead and asking:

* Will the pace of NAND and DRAM price increases begin to slow?

* Can exceptionally high gross margins be sustained?

* Could long-term supply agreements limit earnings upside from further spot-price increases?

* Could new capacity after 2028 lead to oversupply again?

As a result, the market’s focus has shifted from “How strong is demand?” to “How much longer can pricing and margins continue to accelerate?”

3. Individual Earnings Were Not Weak—but Each Stock Fell for Different Reasons

Micron Technology (MU): Initially Rose After Earnings and Was Mainly Dragged Down by the Sector

Micron’s latest quarterly results and guidance both exceeded market expectations. Customers also committed $22 billion to secure future supply, while the company signed 16 strategic customer agreements representing approximately $100 billion in remaining performance obligations.

Management expects tight memory supply-demand conditions to continue beyond 2027, supported by AI demand and structural supply constraints. Following its earnings release, MU shares initially rose as much as 12% in after-hours trading.

When the earnings reports from SNDK and WDC triggered a broader sector selloff, MU initially fell by more than 7% but ultimately closed only around 1.3% lower.

Why was Micron relatively resilient?

* MU benefits primarily from DRAM and HBM, not only NAND;

* Demand for HBM in AI servers remains tight;

* Long-term customer agreements improve revenue and cash-flow visibility;

* Its decline appeared to be driven more by sector contagion than by weaker company guidance.

SK hynix (SKHY): Record Profit, but Still Below Extremely High Expectations

SK hynix’s second-quarter operating profit increased more than sixfold year over year and reached a record high. However, it still missed analysts’ expectations, and the company’s shares closed 9.6% lower in Seoul.

The main concerns included:

* Delayed shipments of certain advanced products limited improvements in DRAM pricing;

* HBM price increases were weaker than those of conventional memory products;

* Long-term supply agreements improved order visibility but could also limit further pricing upside;

* The company did not immediately announce the larger shareholder-return programme that some investors had expected.

Nevertheless, the company’s fundamentals did not materially weaken. SK hynix has completed approximately 10 supply agreements, typically lasting around five years, and customers continue to request additional supply.

Sandisk (SNDK): Results Beat Expectations, but Guidance Varied Across Consensus Sources

Sandisk reported fiscal fourth-quarter revenue of $8.97 billion, above the market expectation of $8.39 billion. Adjusted EPS reached $39.25, also exceeding the expected $34.45.

Data-center revenue reached $2.98 billion and more than doubled sequentially, reflecting continued strength in enterprise SSD and AI-storage demand.

For the following quarter, the company guided for revenue of $10.3 billion to $10.8 billion and adjusted EPS of $44 to $46. This guidance was above the consensus estimate compiled by LSEG but below the higher forecasts published by some other data providers. It would therefore be inaccurate to describe the guidance as universally below expectations.

However, with SNDK shares already up nearly 470% year to date, the market chose to judge the company against the highest expectations. The shares consequently fell nearly 8% in after-hours trading before declining further during the following regular trading session.

On the positive side, Sandisk has signed eight long-term agreements with six customers, with a total contract value of at least $93.9 billion. More than 50% of its fiscal 2027 supply and approximately two-thirds of its fiscal 2028 supply are already covered.

Western Digital (WDC): Strong Data-Center Demand, but Consumer Weakness Could Slow Growth

Western Digital reported fiscal fourth-quarter revenue of $3.75 billion, up 44% year over year and above the market expectation of $3.69 billion. Adjusted EPS came in at $3.56, exceeding the expected $3.30.

For the next quarter, the midpoint of WDC’s revenue guidance was $4.1 billion, above LSEG’s consensus estimate of $4.04 billion. The midpoint of its EPS guidance was $4.00, also above the expected $3.81.

Despite this, WDC shares fell more than 9% in after-hours trading and declined by nearly 19% at one point during the following session.

The market was concerned that:

* Growth was not expected to accelerate significantly in the next quarter;

* Higher storage prices could weaken replacement demand for PCs and consumer electronics;

* Weakness in consumer and client HDDs could offset some of the growth in data-center products;

* After WDC shares had approximately tripled year to date, investors demanded more than a modest earnings beat.

Seagate Technology (STX): Rose After Earnings, Then Was Dragged Down by Sector Repricing

Seagate reported fiscal fourth-quarter revenue of $3.63 billion, up 48.5% year over year. Adjusted EPS reached $5.71, above the market expectation of $5.09.

The company provided the following guidance for the next quarter:

* Revenue: $4.1 billion;

* Adjusted EPS: $7.30;

* Market EPS expectation: approximately $5.80.

STX shares initially rose around 5% in after-hours trading, indicating that investors viewed its earnings and outlook positively.

The subsequent pullback was primarily caused by the broader industry valuation reset triggered by the declines in WDC, SNDK, and SK hynix—not by deterioration in Seagate’s own earnings.

4. SK hynix’s Capacity Expansion: No Immediate Supply Increase, but Longer-Term Cycle Risk

SK hynix approved investments of KRW54.3 trillion—approximately $38.3 billion—to expand semiconductor production capacity in South Korea through 2031:

* KRW35.2 trillion will be invested in the Yongin Y2 fabrication facility, which will produce advanced DRAM and HBM;

* KRW19.1 trillion will be invested in the Cheongju M17 facility, which will mainly produce NAND;

* Construction of M17 is expected to begin in 2027, with its first cleanroom scheduled to become operational by December 2028;

* The first cleanroom at Y2 is expected to be completed in June 2029.

This investment increased investor concerns about future supply growth and a potential reversal of the memory cycle, placing pressure on sector valuations.

However, the new capacity will not enter the market immediately. It is unlikely to have a meaningful effect on supply until at least 2028 or 2029.

Therefore:

* Short term: AI demand may continue to grow faster than supply, supporting memory prices;

* Long term: Investors will need to monitor whether new capacity eventually exceeds demand from AI and traditional devices.

This represents a future risk—not evidence that current industry fundamentals have already reversed.

5. South Korean Regulation: A Volatility Amplifier, Not the Original Trigger

South Korea previously saw a rapid increase in single-stock leveraged ETFs linked to companies such as SK hynix and Samsung Electronics. These products amplified buying pressure when the market was rising. When share prices declined, however, rebalancing, redemptions, and margin calls amplified the selling pressure.

The KOSPI fell approximately 40% from its June peak at one point, erasing more than $2 trillion in market capitalisation within just two days. Regulators subsequently announced measures including:

* Limiting the amount retail investors could invest in single-stock leveraged funds;

* Increasing the trading costs associated with such ETFs;

* Tightening access requirements and risk-management standards for leveraged products;

* Considering additional measures to stabilise the market.

These regulations do not force existing investors to sell their holdings immediately. They also cannot directly restrict similar products listed in the United States or Hong Kong. Therefore, South Korean regulation was not the original cause of the global memory-stock selloff.

Its real impact has two sides:

* Short-term negative: Reduced inflows into leveraged products may remove part of the aggressive momentum-buying that previously supported rebounds;

* Medium- to long-term positive: Lower leverage could reduce forced liquidations and mechanical ETF selling, allowing share prices to be driven more by earnings and cash flow.

J.P. Morgan estimates that the unwinding of South Korean leveraged ETFs may be largely complete, while approximately 90% of related hedge-fund deleveraging may also have concluded. This suggests that the most intense phase of forced selling could be approaching its end.

Conclusion: More Likely a Healthy Pullback Than the End of the AI Memory Cycle

We view this selloff as a healthy reset of expectations and valuations rather than a reversal of AI-storage demand.

Evidence supporting this view includes:

* Micron expects supply-demand conditions to remain tight beyond 2027;

* Micron has secured $22 billion in customer commitments;

* Sandisk has at least $93.9 billion in long-term supply agreements;

* SK hynix customers continue to request additional supply;

* Seagate’s next-quarter EPS guidance was significantly above market expectations;

* Most of SK hynix’s new capacity will not be released until 2028 or 2029;

* The unwinding of leveraged positions in South Korea may already be entering its final stage.

Of course, a large decline does not necessarily mean that the market has reached a bottom. The current price range can only develop into a new medium-term support level if volatility begins to narrow, earnings forecasts stop falling, memory prices remain firm, and future guidance continues to confirm strong demand.

However, if AI capital expenditure and demand for enterprise SSDs, HBM, and high-capacity HDDs continue to grow, this valuation reset could remove excessive leverage and short-term speculative capital—creating a healthier foundation for the sector’s next upward move.

In one sentence:

Memory stocks did not fall because AI demand disappeared. They fell because the market shifted from pricing in “unlimited acceleration” to recognising “strong growth that still needs to be continuously validated.” Near-term volatility remains high, but there is currently no clear evidence of a structural reversal in long-term demand, order visibility, or supply constraints.

This material is provided solely for market research and educational purposes and does not constitute investment advice.