Micron (MU) shares surge over 725% in a year: Is AI memory demand ushering in a new semiconductor supercycle?

On July 10, 2026, Micron Technology (Micron Technology, U.S. stock ticker MU) closed at $979.30. Although it pulled back 1.24% on the day, its year-to-date gain has reached 243.12%, and over the past 52 weeks it is up more than 700%. On June 25, Micron’s stock price hit an intraday all-time high of $1,255. Its market capitalization briefly exceeded $1.1 trillion.

In the same period, the market capitalizations of SK hynix and Samsung Electronics also broke through $1 trillion one after another. The combined market capitalization of the three memory makers reached $4.1 trillion—10 years ago, this figure was only $254 billion. For the first time in May 2026, Micron, SK hynix, and Samsung simultaneously achieved the $1 trillion market-cap milestone.

The core driver behind this surge is not the traditional PC or smartphone memory cycle, but a brand-new narrative: the AI memory supercycle. High-bandwidth memory (HBM) is shifting from a supporting role for GPUs to a strategic bottleneck for AI infrastructure. As the only major-scale HBM manufacturer in the United States, Micron is redefining its industry position in this process. From three dimensions—market logic, industrial structure, and the competitive landscape—this article analyzes the structural changes behind Micron’s rally and assesses the sustainability of this trend.

From the “memory cycle” to the “AI memory supercycle”: a fundamental shift in market logic

The dilemma of the traditional memory cycle

Over the past three decades, the memory industry has been plagued by a deeply rooted problem: high cyclicality.

On the demand side, it is driven by upgrades and replacements in PCs, smartphones, and general-purpose servers, resulting in sharp fluctuations. On the supply side, capacity expansion cycles can last for years, repeatedly causing misalignment between supply and demand—demand rises → manufacturers expand → supply overshoots → prices fall → manufacturers cut output → the next round of shortage. This “boom-and-bust” cycle makes memory makers’ profitability extremely unstable. For a long time, the market has treated memory stocks as a classic example of cyclical stocks.

How AI changes the demand structure

The arrival of the AI era is rewriting this logic. The scale of large language models grows by an order of magnitude every 12 to 18 months. Training these models requires GPU clusters to deliver massive compute power, and GPU operation in turn requires HBM to provide extremely high memory bandwidth and data throughput.

The supply-chain transmission path is as follows:

AI model scale expands → GPU cluster demand grows → HBM demand surges → high-performance memory becomes a scarce resource

2026 is widely seen by the market as the first year of the “HBM supercycle.” UBS forecasts that in 2026, HBM demand will grow 90% year over year, reaching about 33.1 billion Gb; in 2027, it will grow another 77% to about 58.7 billion Gb. Bernstein notes that compared with traditional memory, HBM has significantly higher profit margins, and demand is increasing by a substantial amount beyond supply.

From “bargaining” to “contracts”: a structural shift

Even more crucial is the change in the pricing model. In the past, memory was a standardized commodity, and prices moved with the spot market. Today, Micron has signed 16 “Strategic Customer Agreements” (Strategic Customer Agreements, SCA) with its major customers, covering about $100 billion in long-term supply commitments, plus about $22 billion in prepayments and letters of credit. These agreements not only lock in procurement volumes, but also establish a structured mechanism for price adjustments.

Micron has sold out all of its HBM capacity for 2026. Orders for HBM3E and HBM4 have already been scheduled through 2027, and demand even extends to 2028. The company expects the total addressable market (TAM) it can serve for HBM to exceed $100 billion in 2027—one year earlier than its prior forecast for 2028.

Why has HBM become the new core of AI competition?

Without HBM, there is no large-scale AI

In the past, the market focus was on “who can produce AI chips the fastest?” Now the question is shifting to “who can provide enough high-speed memory to support AI chip operation?”

HBM’s core value lies in its extremely high memory bandwidth and low-latency characteristics. Training large models requires massive data transfer between GPUs and memory, and traditional DRAM bandwidth cannot meet this requirement. HBM uses 3D stacking technology to vertically integrate multiple DRAM chips, delivering far more bandwidth per unit area than conventional solutions.

Industry sources point out that HBM4 has a production cycle of 4 to 6 months, its initial yields are significantly lower, and the wafer capacity consumed for production is about three times that of standard DDR5 DRAM. This means that even if manufacturers expand aggressively, the pace of supply release is still far slower than that of traditional memory.

Only three companies worldwide can mass-produce HBM

The supply side of the HBM market is highly concentrated. Globally, only three companies can mass-produce HBM: SK hynix, Samsung Electronics, and Micron Technology.

The competitive landscape for the first quarter of 2026 is as follows:

• SK hynix holds a leading position in the HBM market with about 58% market share
• Micron has about 21% to 23% market share
• Samsung has about 21% market share in the HBM market

UBS expects that by 2027, Samsung and SK hynix will each account for about 40% of HBM bit share, and Micron will account for the remaining approximately 20%.

HBM’s high technological barriers, long customer certification cycles, and capacity bottlenecks jointly build an extremely high industry moat. For a new HBM supplier, it typically takes years from technical validation to mass production, further strengthening the competitive advantages of the current three giants.

How is Micron benefiting from the AI memory boom?

A leap in the scale of financial figures

Micron’s results for fiscal third quarter 2026 (ended May 28) fully reflect the AI-driven structural changes:

| Metric | Data | Year-over-year change | | --- | --- | --- | | Total revenue | $41.46 billion | +345.7% | | GAAP net profit | $28.24 billion | +1,398.3% | | GAAP gross margin | 84.6% | +46.9 percentage points |

Among them, cloud storage business revenue (including HBM) was $13.77 billion, with a gross margin of 83%; core data center business revenue was $11.52 billion, with a gross margin of 87%. Micron’s shipped HBM4 has generated more than $1 billion in revenue.

The company’s guidance for next quarter calls for revenue of $50 billion (plus or minus $1 billion) and a GAAP gross margin of 86%. During the earnings call, Micron executives said that the tightness in storage supply will continue beyond 2027.

Revenue visibility locked in by strategic customer agreements

Micron converts about 40% of its revenue into “contract-like” revenue through 16 SCAs. The core value of this model lies in:

• Revenue predictability: Multi-year contracts lock in orders and prices for years ahead
• Stronger pricing power: In an environment of tight supply and demand, Micron has greater leverage in negotiations
• Confidence in capital expenditures: Stable revenue expectations support large-scale capacity expansion

Micron’s capital expenditures for fiscal year 2026 are about $27 billion. For fiscal year 2027, capital expenditures in each quarter will exceed the fourth quarter of fiscal year 2026.

$250 billion in U.S. domestic investment: a dual layout for both capacity and policy

On July 9, 2026, Micron announced that it will raise its U.S. domestic investment plan from $200 billion to $250 billion. The overall investment will continue through 2035. The additional $50 billion will be used to build fabs in New York, Idaho, and Virginia.

Micron’s goal is to achieve about 40% of DRAM produced in the United States domestically in the future, up from the current level of about 10%. In Clay, New York, multiple advanced DRAM wafer fabs are planned, with total investment as high as $100 billion. Groundbreaking took place at the start of 2026, and the first concrete pour was completed one quarter earlier than the original schedule.

The dual significance of this layout is: on the one hand, it directly expands HBM and advanced DRAM capacity; on the other hand, it aligns with the direction of U.S. industrial policies that encourage the return of semiconductor manufacturing, giving the company leverage for policy preferences and approval priority.

Can Micron keep rising? Five dimensions the market is watching

Bullish factors

AI capital expenditures are still in the expansion phase

AI capital expenditures by global hyperscale cloud service providers are expected to exceed $800 billion in 2026, up 67% year over year. They are expected to surpass $1 trillion in 2027. Alphabet’s expected 2026 capital expenditures have been raised to $195 billion and 2027 has been raised to $290 billion; Meta’s 2026 figure has been raised to $145 billion. A substantial portion of these expenditures will directly translate into demand for HBM.

HBM supply constraints are difficult to break in the short term

HBM manufacturing involves complex processes such as advanced packaging, silicon-through vias (TSV), and 3D stacking. From planning to mass production, new capacity typically takes more than two years. Even if the three major vendors expand at full throttle, industry sources expect the structural shortage of HBM to persist through mid-2028.

HBM prices still have upside room

Industry sources say the price of next-generation HBM4 could rise from about $2 per gigabit in the second half of 2026 to $4 to $5 or even higher. On July 13, 2026, the Bank of Korea also said that because customized products such as HBM are becoming mainstream, the pace of semiconductor supply expansion is more constrained than in the past. The global semiconductor market is expected to maintain an expansion trend for a considerable period.

Analysts are broadly bullish

As of mid-July 2026, analysts maintain a “strong buy” consensus rating for Micron, with an average target price of about $1,328. TD Cowen maintains a $1,600 target price. Melius’ analyst sets the highest target at $2,200. Morgan Stanley raised its target price to $1,200, and Bank of America Securities raised it to $1,550.

Risk factors

Is AI valuation getting overheated?

Micron’s current trailing price-to-earnings (P/E) ratio is about 22x, and its price-to-book (P/B) ratio is 2.35. Despite strong earnings growth support, the market has begun to ask whether AI investment can translate into long-term, sustainable profitability. Bernstein expects Micron’s earnings to peak in 2027 and then start declining.

Could the memory cycle reverse again?

Although SCAs lock in about 40% of revenue, the remaining about 60% still fluctuates with spot prices. Micron is raising capital expenditures from $27 billion in fiscal year 2026 to more than $45 billion in fiscal year 2027—adding capacity at the peak of the cycle. If demand growth falls short of expectations, it could repeat historical cycle patterns.

Intensifying HBM competition

Samsung plans to increase its HBM capacity by 50% in 2026, targeting 250,000 wafers per month. Industry watchers believe Samsung could eventually catch up—and even overtake—in the HBM4 space, potentially returning to the leading position in the HBM market in 2027. SK hynix raised about $29.4 billion through a Nasdaq listing to fund capacity expansion. Whether Micron’s share in the HBM market can keep rising remains highly uncertain.

Conclusion

Micron Technology’s stock price surged by more than 243% in 2026, reflecting not only a performance breakout for a single company, but also a snapshot of structural changes across the entire semiconductor industry.

The traditional memory cycle is built on demand from PCs, phones, and general-purpose servers, with standardized products, volatile pricing, and hard-to-predict profitability. In the AI era, HBM demand is reshaping memory from a “cyclical product” into a “strategic resource”—with features such as high technological barriers, long customer certification cycles, and multi-year supply contracts, these characteristics have brought about a fundamental change in the memory industry’s business model.

However, the sustainability of this shift still faces challenges. SCAs lock in part of the revenue, but cannot fully eliminate cyclicality. Large-scale capacity expansion may lead to supply oversupply a few years later. The competitive landscape in the HBM market is also far from finalized. Whether Micron can fully transform from a “cyclical stock” into an “AI infrastructure growth stock” depends on whether HBM demand can continue to outpace the speed of supply expansion, and whether the company can maintain competitiveness through technological iterations.

For investors, understanding Micron’s logic shift matters more than predicting short-term stock price movements. Whether an AI memory supercycle is reshaping the semiconductor industry may be true. But the direction and depth of that reshaping still need to be validated over time.

FAQ

Q1: How much did Micron Technology (MU) stock rise in 2026?

As of July 10, 2026 (Beijing time), Micron Technology’s stock closed at $979.30, up 243.12% year-to-date in 2026. Over the past 52 weeks, it is up more than 700%. On June 25, it hit an intraday all-time high of $1,255.

Q2: What is HBM? Why is it so important for AI?

HBM (high-bandwidth memory) is a high-performance memory that vertically integrates multiple DRAM chips using 3D stacking technology, providing data transfer bandwidth far beyond that of traditional memory. Large AI model training requires massive data exchange between GPUs and memory. Without the high-bandwidth support of HBM, large-scale AI training cannot run efficiently.

Q3: What is Micron’s competitive position in the HBM market?

In the first quarter of 2026, SK hynix led the HBM market with about 58% share, Micron accounted for about 21% to 23%, and Samsung accounted for about 21%. Globally, only these three companies can mass-produce HBM. UBS expects that by 2027 Samsung and SK hynix will each account for about 40%, and Micron will account for about 20%.

Q4: What is Micron’s $250 billion investment plan?

On July 9, 2026, Micron announced that it will raise its U.S. domestic investment from $200 billion to $250 billion, continuing through 2035. The target is to achieve about 40% of DRAM produced domestically in the United States in the future, currently about 10%. Investments will be used for building fabs in New York, Idaho, and Virginia.

Q5: What are Micron’s main risks?

The main risks include: AI valuation could become overheated; the memory cycle could reverse again after capacity expansion; competition in the HBM market could intensify (both Samsung and SK hynix are expanding significantly); and about 60% of revenue still fluctuates with spot prices. Bernstein expects Micron’s earnings to reach a peak in 2027.