On July 3, 2026, the crypto market saw a broad rally. According to Gate market data, Bitcoin (BTC) was priced at $61,506, up 1.7% over the past 24 hours, while Ethereum (ETH) traded at $1,698, up 4.7% in the same period. Bitcoin rebounded from yesterday’s low of $59,776 to $61,507, a 2.86% increase. Ethereum surged from $1,605 to $1,706, jumping 6.26% in a single day. In the US, nonfarm payrolls for June increased by just 57,000 and the unemployment rate hit 4.2%, both below market expectations. This eased concerns over a near-term Fed rate hike. Bitcoin’s market cap stands at approximately $1.23 trillion, while the total crypto market capitalization is around $2.21 trillion.
In contrast, traditional financial markets painted a different picture. US stocks in the storage and semiconductor sectors faced heavy sell-offs, with the Philadelphia Semiconductor Index dropping 11% over two days. The three storage giants—SanDisk (SNDK), Micron (MU), and Seagate (STX)—all suffered significant losses, sharply diverging from the crypto market’s rebound.
What Happened in the Markets?
On July 3 (Beijing time), the three major US stock indices posted mixed results: the Dow rose 1.1% to a new closing high, the Nasdaq fell 0.8%, and the S&P 500 edged down slightly. The storage sector was the biggest drag of the day.
SanDisk (SNDK) led the declines, closing down 14.13% at $1,745. The stock had already shown weakness in the prior session, bringing its two-day cumulative drop close to 25%. Micron Technology (MU) fell 5.49% to $975.56, after plunging 10.57% the previous day. Seagate Technology (STX) dropped 10.38%.
Other storage-related stocks also came under pressure: Western Digital (WDC) fell about 10%, Intel (INTC) lost 5%, KLA Corporation dropped over 11%, and Arm declined more than 6%. In Asian markets, Samsung Electronics slid 6.52%, SK Hynix dropped 7.73%, and Kioxia Holdings fell 12.57%. The AI storage sector saw an almost across-the-board sell-off.
Why the Sell-Off?
This round of selling was triggered by a confluence of factors. However, the core driver wasn’t a deterioration in fundamentals, but rather a sharp shift in market sentiment and trading structure.
First, concerns about the return on AI infrastructure investments are rising. On July 1, news broke that Meta plans to sell excess AI computing power to external clients. Previously, Meta announced capital expenditures of $145 billion for 2026. The market interpreted this as even leading tech firms reassessing the ROI of their AI infrastructure spending. Industry analysts believe the issue isn’t a true surplus of AI computing power, but that cloud providers must make trade-offs to maintain positive cash flow when their investments in compute can’t be fully converted into revenue. If AI investments continue to underperform, future capex could be constrained.
The Bank for International Settlements (BIS) recently warned of overheating in AI investments. Tech giants like Alphabet (Google’s parent), Microsoft, and Amazon plan to invest over $1 trillion in AI-related businesses between 2025 and 2026. The BIS cautioned that while AI investment can boost productivity, excessive investment could trigger financial system turmoil if the trend reverses.
Second, the sector had already seen outsized gains and was technically overheated. Since the start of 2026, AI storage stocks have posted remarkable rallies. Micron is up about 240% year-to-date, setting 37 all-time closing highs in the first half. SanDisk soared 388.4% in all of 2025. Seagate skyrocketed 146.32% in Q2 alone. Such steep climbs mean that any marginally negative news can spark widespread profit-taking.
Third, short selling has acted as a catalyst for releasing bearish sentiment. Market reports indicate that Michael Burry, the real-life inspiration for "The Big Short," shorted Micron around $1,051.87 on July 1, triggering the latest wave of sharp declines in AI storage stocks and amplifying market panic.
Fourth, a DRAM antitrust lawsuit has added to uncertainty. On June 25, Samsung Electronics, SK Hynix, and Micron faced a class-action lawsuit in US federal court in California. They’re accused of colluding to restrict traditional DRAM capacity under the guise of HBM transformation, causing memory prices to surge about 700% over the past four years. The suit seeks triple damages and injunctive relief for antitrust violations. While the final outcome will take time, the lawsuit has had an immediate chilling effect on market sentiment.
Have the Fundamentals Deteriorated?
The answer: No, they haven’t.
Recent earnings reports show that all three companies remain fundamentally strong. For its fiscal Q3 2026 (ending May 28), Micron reported revenue of $41.46 billion, far exceeding last year’s $9.3 billion and analysts’ estimate of $35.91 billion. Adjusted EPS was $25.11, beating the expected $20.86. The company projects Q4 capex of about $10 billion and full-year capex of $27 billion. For fiscal 2027, quarterly capex will exceed Q4 2026 levels.
SanDisk’s fiscal Q3 2026 (ending April 30) revenue reached $5.95 billion, up 97% quarter-over-quarter and 251% year-over-year, far surpassing both company guidance and Wall Street’s $4.7 billion estimate. Adjusted EPS was $23.41, compared to analysts’ forecast of just $14.51. SanDisk expects Q4 revenue of $7.75–$8.25 billion and adjusted EPS of $30–$33.
Seagate reported Q3 2026 revenue of $3.11 billion, non-GAAP EPS of $4.10, and free cash flow of $953 million. Its Q4 revenue guidance is $3.45 billion, with adjusted EPS of $5.00—both above analyst expectations.
On the supply-demand side, shortages persist.
TrendForce data shows that in Q2 2026, traditional DRAM contract prices are expected to rise 58%–63% quarter-over-quarter, while NAND Flash contract prices are projected to increase 70%–75%. Entering Q3, the overall DRAM market remains extremely tight.
Goldman Sachs expects NAND’s average selling price to surge 4.5x year-over-year in 2026, with another 38% increase in 2027—higher than the previous 27% forecast. Channel checks indicate major storage manufacturers are prioritizing capex for DRAM, not significantly expanding new NAND capacity. With AI demand growing, new NAND supply may not ramp up meaningfully until 2028.
Morgan Stanley’s updated supply-demand model shows a 15% global NAND supply gap in 2026 and a 9% gap in 2027. AI-related NAND demand is expected to rise from 205 EB in 2025 to 400 EB in 2026, and further to 609 EB in 2027—a roughly 60% annual growth rate. AI’s share of the overall NAND market will climb from 18% in 2025 to 32% in 2026, and 41% in 2027.
UBS forecasts DRAM shortages will persist through Q2 2028, while NAND shortages will last through Q4 2027. Server SSD demand is projected to grow 56% in 2026 and 47% in 2027.
For HBM, TrendForce expects shipments to rise 60% year-over-year in 2026 and another 60% in 2027. HBM will remain in short supply through at least 2027. HBM’s share of total DRAM wafer input will increase from 18% in 2025 to about 30% in 2027.
The Industry’s Real Dilemma: Crowded Short-Term Trades vs. Long-Term Shortages
The market’s core issue isn’t whether AI storage demand is real, but whether the current short-term trading structure is sustainable.
From a long-term perspective, AI infrastructure construction is still in its early stages. According to Sigmaintell, from 2024 to 2028, global AI infrastructure and compute investments will grow at double-digit rates annually. In 2026, the year-over-year growth rate remains high at 51%, though slower than 104% in 2025. AI server storage requirements are 4–5 times those of general-purpose servers. A top-tier AI training server uses over 100 GB of HBM alone, with storage accounting for more than 40% of total costs—25 percentage points higher than the 15% seen in traditional servers.
However, the short-term trading structure does carry significant risks. Stock gains have far outpaced improvements in fundamentals. While Micron’s 240% year-to-date rally and SanDisk’s 251% annual revenue growth are impressive, their share prices have risen even more steeply. As the market questions the sustainability of hyperscale cloud providers’ trillion-dollar capex plans, any marginal change can trigger sharp corrections.
What Should Investors Focus On?
HBM (focus on Micron and SK Hynix): HBM is the most critical storage component for AI training and inference, and will remain in short supply through 2027. As one of the three main HBM suppliers, Micron stands to benefit directly from this structural trend. HBM consumes 4–5 times as many wafers as DDR5, naturally limiting capacity expansion.
NAND (focus on SanDisk): SanDisk holds a key position in NAND Flash, with enterprise SSDs for AI inference becoming the largest application market for NAND. Goldman Sachs expects NAND prices to continue rising sharply in 2026 and 2027. The global NAND supply shortage is likely to persist through 2027.
HDD Data Storage (focus on Seagate): The demand for cold data storage in AI data centers is reshaping the HDD market. Western Digital and Seagate’s 2026 hard drive capacity has essentially been booked by AI data centers. After Seagate’s 146% surge in Q2, the stock has pulled back, but its fundamentals remain solid.
Conclusion
The collective decline of the three AI storage giants on July 3, 2026, is a textbook example of "good fundamentals, bad prices." The fundamentals—supply-demand gaps in DRAM and NAND, structural shortages in HBM, and long-term capex in AI data centers—haven’t materially worsened. What’s really changed is the market’s repricing of AI investment returns and profit-taking after last year’s extraordinary rally.
On the same day, the crypto market rallied across the board, with BTC climbing above $61,500 and ETH breaking $1,700. While the two markets appear to be moving in opposite directions, they actually share the same macro backdrop: marginal improvements in liquidity expectations and relief from rate hike concerns following the nonfarm payroll data. However, the AI storage sector in traditional finance is still digesting its own unique valuation pressures.
For investors, the key is distinguishing between "short-term trading noise" and "long-term structural trends." The long-term demand story for AI storage remains intact—structural shortages in HBM, NAND, and enterprise SSDs are unlikely to be resolved before 2027. However, the process of deflating short-term valuation bubbles may not be over yet. Morningstar’s research director Lorraine Tan told Bloomberg TV that AI-related stocks may need to fall another 20%–30% before they become attractive buys again, citing new supply from Samsung and SK Hynix and a potential plateau in AI capex.
The storage industry stands at a crossroads in a supercycle. The direction of the cycle hasn’t changed, but volatility is intensifying.
FAQ
Q1: Is there really an oversupply of AI storage chips?
At present, there is no substantial oversupply. TrendForce data shows that DRAM will remain "extremely tight" in Q3 2026. Morgan Stanley’s model indicates a 15% global NAND supply gap in 2026. The current sell-off mainly reflects concerns about "potential future oversupply," not the current reality.
Q2: Why did SanDisk’s share price fall the most?
SanDisk’s stock soared 388.4% in 2025, making it one of the biggest gainers in the storage sector. This massive earlier rally means there’s now the greatest pressure for profit-taking. Additionally, the NAND market is seen as slightly less certain than DRAM by institutional investors, amplifying the sell-off.
Q3: Has there been a problem with Micron’s fundamentals?
No. In fiscal Q3 2026, Micron reported revenue of $41.46 billion and EPS of $25.11, both far exceeding expectations. The company projects $27 billion in capex for fiscal 2026, with further increases in 2027. The recent decline is mainly due to valuation adjustment after a 240% rally and short-selling activity.
Q4: Does the long-term investment thesis for AI storage still hold?
The long-term thesis remains unchanged. AI server storage requirements are 4–5 times those of general-purpose servers. Goldman Sachs expects NAND prices to rise 4.5x in 2026 and another 38% in 2027. UBS forecasts DRAM shortages through Q2 2028. The key issue is timing the alignment of valuations and fundamentals.
