$MU Micron Technology (Micron TechnologyMU) In-Depth Investment Research Report

Micron Technology, Nasdaq ticker MU, is headquartered in Boise, Idaho, USA. It is one of the world's three major memory chip giants (alongside Samsung and SK hynix) and the only leading US memory chipmaker. The company's CEO is Sanjay Mehrotra.

Business: Micron operates the "temporary and permanent warehouses for data"—DRAM memory (temporary storage used when computers and phones run programs; data disappears when power is cut) and NAND flash memory (SSD solid-state drives and memory cards; data remains after power is cut).
The hottest product in recent years has been HBM, or high-bandwidth memory (high-speed memory stacked beside AI GPUs and one of the core bottlenecks in AI training and inference). The current share price is $950+, with a total market capitalization of approximately $1.08 trillion, making it one of the highest-market-cap semiconductor companies in the US. Its fiscal year ends in late August each year.

Financial analysis: Epic growth metrics amid the AI memory supercycle FY2025 FY2026 Q1 FY2026 Q2 FY2026 Q3 Q4 FY26 guidance
Revenue ($ billions) 37.378 13.643 23.860 41.456 ~50
Year-over-year growth — — — +345.7% +342% Non-GAAP EPS (dollars) — — 12.20 25.11 ~31.00
Gross margin (Non-GAAP) — — — 84.9% —

Key observations:
1. Performance is growing exponentially: Revenue for the first three quarters of FY2026 totaled $78.959 billion, already more than twice FY2025 full-year revenue of $37.378 billion. Q3 revenue came in at $41.456 billion, surging 345.7% year over year and 73.8% quarter over quarter, substantially exceeding market expectations of $35.8 billion. Q4 guidance is approximately $50 billion, and full-year FY2026 revenue could reach $129 billion.
2. Gross margin of 84.9%, surpassing NVIDIA: Q3 Non-GAAP gross margin was 84.9%, while GAAP gross margin was 84.6%, setting a historic record for the memory industry and even exceeding NVIDIA's approximately 75% gross margin during the same period. This is unprecedented in the capital-intensive memory manufacturing industry. The core reason is that HBM and data center DRAM are in severe undersupply, driving prices sharply higher while manufacturing costs remain relatively fixed.
3. Data center business is the core engine: Q3 data center revenue exceeded $25 billion, surpassing $100 billion on an annualized basis. Demand for HBM and high-performance DRAM from AI servers is growing explosively.
4. Extremely strong free cash flow: Q3 operating cash flow was $25.39 billion, and adjusted free cash flow was $18.3 billion. The profits are backed by real cash.

Core thesis: Why has AI transformed the memory industry?
1. HBM—the "holy grail of memory" in the AI era, explained simply: Conventional memory is laid flat on the motherboard, so data must travel a long distance from memory to the GPU, resulting in slower speeds and higher power consumption. HBM, or high-bandwidth memory, stacks several layers of memory chips like a hamburger and connects them vertically through "through-silicon vias" (TSVs), placing them right next to the GPU. Data transfer speeds are several times faster than conventional memory, while power consumption is lower. When training large AI models, GPUs must read and write data at an enormous rate. Without enough HBM, even the most powerful GPU cannot be "fed" adequately—so HBM has become a core bottleneck for AI computing power. HBM is extremely difficult to manufacture: it requires precisely stacking 12 or even 16 layers of chips, has low yields, and consumes substantial wafer capacity (producing the same capacity of HBM requires more than 3 times the wafer capacity of DDR5). This is why supply expansion is slow and prices remain high.
2. Micron's HBM progress
• HBM4: Mass shipments have begun to key customers (including NVIDIA), and qualification samples have been sent to multiple end customers
• HBM4E: Being developed based on 1-gamma DRAM, with mass production expected in 2027
• Micron is rapidly narrowing the gap with SK hynix and Samsung in the HBM sector
3. How is this cycle different from previous ones?
Historically, the memory industry has followed a typical "hog cycle": price increases → capacity expansion → oversupply → price collapse → losses → production cuts → another price increase, with one cycle every 2-3 years.
However, some believe this cycle represents a structural change:
• AI demand for HBM is long-term and exponential, rather than a short-term pulse
• HBM capacity expands extremely slowly due to high technological barriers and slow yield ramp-up; it cannot simply be expanded at will
• Memory manufacturers are signing long-term supply agreements (LTAs) with customers, locking in capacity and prices for the next 2-3 years and greatly improving revenue visibility
• Micron management has explicitly forecast that tight memory supply will continue beyond 2027Of course, some cautious observers believe this is ultimately still a cycle, just with a longer peak and higher profits.

Earnings forecasts (FY2027-FY2029) based on company guidance and industry trends, with reasonable estimates: Metric FY2027E FY2028E FY2029E Revenue ($ billions) 160-180 150-170 140-160 Year-over-year growth +24%~+40% -6%~-6% -7%~-6% Non-GAAP EPS (dollars) 130-160 90-120 70-100 Gross margin 75%-82% 60%-70% 50%-60%
Forecast logic: In FY2027 (the current fiscal year), HBM demand will remain strong, while long-term agreements support pricing. Revenue and profits will continue to grow, but at a slower pace. If the memory industry enters a downcycle in FY2028-FY2029, in line with historical patterns, prices will decline and profits will fall. However, if AI demand continues to exceed expectations and HBM supply remains tight, the downcycle could be delayed or weakened. Institutions expect FY2027 EPS of approximately $130-$160, with considerable divergence.

Valuation and investment recommendations
Current valuation:
• Market capitalization of approximately $1.08 trillion
• Based on estimated FY2026 net profit of approximately $75 billion, the PE ratio is about 14x
• Based on projected FY2027 EPS of $145, the forward PE ratio is approximately 6.6x
• Price-to-sales ratio (PS) of approximately 8.4x (based on FY2026 revenue of $129 billion)
• Consensus institutional rating: Strong Buy
• Average institutional price target: approximately $1502 (implying approximately 57% upside)
• Price target range: $361-$2200 (extremely wide divergence)

Views from major institutions:
Barclays: $2000 price target, Buy (raised by 70%)
• UBS: $1625 price target
• Susquehanna: $2000
• New Street Research: $1250 price target, Buy (believes AI has broken the cycle)
• Nova Capital: $1300.5 price target, Buy
• Of 39 analysts, 31 rate it Buy and 5 rate it Strong Buy

Investment highlights:
1. AI memory supercycle: HBM demand is surging, and data center revenue exceeds $100 billion on an annualized basis
2. Epic earnings explosion: FY2026 revenue is expected to reach $129 billion, while Q3 gross margin of 84.9% surpassed NVIDIA's
3. The only US memory giant: Supported by the US government, with clear supply chain security advantages
4. Rapidly catching up in HBM technology: HBM4 has entered mass shipment, and HBM4E is expected to enter mass production in 2027
5. Long-term agreements locked in: LTAs with customers provide high revenue visibility, and management forecasts tight supply to continue beyond 2027
6. Valuation appears inexpensive: FY2026 PE is only 14x, with an even lower forward PE

Core risks:
1. Memory cycle risk (the biggest risk): The memory industry has historically been highly cyclical, and the current 84.9% gross margin is unsustainable. If AI demand slows or capacity expansion catches up, prices and profits could collapse. The low PE of 14x based on peak-cycle profits is misleading
2. Competitive risk: SK hynix remains the leader in HBM, while Samsung has enormous capacity; a price war could erupt
3. AI demand falling short of expectations: If commercialization of large models progresses slowly, cloud providers' capital expenditures could slow
4. Geopolitical risk: China is an important market, and export controls and trade frictions could affect the business
5. Huge market capitalization: With a market cap of $1 trillion, further substantial gains require continued earnings that exceed expectations
6. Capacity expansion risk: If all three major manufacturers significantly expand HBM capacity simultaneously, oversupply could emerge in 2-3 years

Overall assessment: Micron Technology is one of the biggest beneficiaries of this AI memory supercycle. Explosive growth in HBM and data center businesses has enabled historic breakthroughs in both the company's performance and its share price. Unlike previous memory cycles, this cycle is supported by structural factors including high HBM technological barriers, long-term agreements, and long-term AI demand growth. Micron management also forecasts tight supply to continue beyond 2027. However, it is essential to recognize that the memory industry remains cyclical. An 84.9% gross margin cannot be maintained forever, and the current $1 trillion market capitalization has already priced in a large amount of optimistic expectations.

Recommendations:
• Aggressive investors: May continue holding or add small positions on pullbacks (such as below $800), betting on FY2027 earnings continuing to exceed expectations and HBM demand remaining tight, but should set strict stop-losses
• Conservative investors: Consider staying on the sidelines or gradually taking profits, and wait for clear signs of a memory downcycle and a substantial share price correction before considering an investment
• Key indicators to monitor: HBM spot prices and order visibility, DDR5/NAND price trends, capital expenditure plans of the three major manufacturers, AI server shipments from customers such as NVIDIA, changes in Micron's quarterly gross margin, and the signing of long-term agreements (LTAs)$MU ‌