Micron's Memory "Chip Shortage" to Continue Through Next Year! Earnings Call: Can Only Meet 50% of Customer Demand, Urgently Needs to Speed Up "Burning Cash" to Build Factories

On March 18th Eastern Time, Micron Technology released its Q2 FY2026 earnings report, delivering an impressive performance that set multiple historical records.

Micron executives stated during the earnings call that driven by sustained explosive demand for AI, tight supply of industry storage products, and successful mass production of advanced technologies, the company’s core financial indicators—revenue, gross margin, free cash flow—hit all-time highs. All product lines and business segments flourished simultaneously. Additionally, the company’s guidance for the third quarter significantly exceeded expectations again.

The executives also mentioned that the current storage chip industry faces a supply shortage that is expected to persist beyond 2026, providing solid support for short- and medium-term product pricing. Under strong demand, Micron signed its first five-year strategic customer agreement.

They revealed that HBM4, designed specifically for NVIDIA Vera Rubin, has been shipped in volume, and development of the next-generation HBM4E is progressing smoothly, with mass production expected in 2027. To meet robust customer demand, Micron plans to significantly increase capital expenditures over the next two fiscal years to build factories and purchase equipment, boosting capacity.

1. Core Product Line Performance

• DRAM: Revenue of $18.8 billion, up 207% YoY and 74% QoQ; prices increased 65-67% QoQ; shipment volume grew in the low single digits. 1γ DRAM mass production is the fastest in the company’s history, expected to become the main product by mid-2026.

• NAND: Revenue of $5 billion, up 169% YoY and 82% QoQ; prices increased 75-79% QoQ; shipment volume grew in the low single digits. G9 NAND is progressing as planned, expected to be the main product by mid-2026, with data center NAND revenue doubling QoQ to set records.

• High-end products: HBM4, designed for NVIDIA Vera Rubin, has been shipped in volume; HBM4E development is on track for 2027 mass production. PCIe Gen6 SSD based on G9 NAND achieves high-volume production, with 122TB SSD highly recognized in the market.

2. Industry and Market Outlook

• Supply and demand: DRAM/NAND supply shortages will continue until after 2026. In 2026, industry-wide DRAM bit shipments are expected to grow about 20%, NAND about 20%. Micron’s supply growth will be in line with the industry. Core customer demand fulfillment rate remains at only 50-2/3.

• AI impact: By 2026, AI will drive data center DRAM/NAND TAM to surpass 50% of the industry. Strong demand for AI servers, with server shipments growing low double digits, and increasing DRAM capacity per server.

• Terminal markets: PC and smartphone shipments may decline low double digits in 2026, but storage content driven by AI will grow significantly (flagship phones with 12GB+ DRAM rising from 20% to 80%; AI PCs recommended with 32GB+).

• Emerging sectors: Robotics industry entering a 20-year growth cycle, with humanoid robots’ storage needs comparable to L4 autonomous vehicles, becoming a new growth driver.

3. Strategic Layout and Capacity Expansion

• Customer cooperation: Signed the first 5-year Strategic Customer Agreement (SCA), replacing traditional 1-year LTA, with clear capacity/supply commitments. Negotiations ongoing with multiple customers to deepen R&D and roadmap collaboration.

• Technology R&D: Significantly increased R&D investment in FY2027; expanding EUV applications for 1δ DRAM, accelerating product launch through combined R&D and mass production.

• Global capacity: Acquired Lehi Fab and plans to build a second cleanroom (shipments in FY2028); capacity expansion in Idaho, New York, Hiroshima, and a new NAND fab in Singapore (mass production in 2028); packaging and testing plant in India now commercially shipped; advanced HBM packaging plant in Singapore expected to contribute in 2027.

Our HBM4E will be built on Micron’s industry-leading 1γ process DRAM technology, validated in mass production, with performance leaps to support next-generation AI computing platforms. Custom options for HBM4D will bring more differentiation and foster deeper R&D collaborations with customers.

Micron pioneered low-power DRAM (LPDRAM) for data centers, with power consumption only one-third of data center DDR server modules. Leveraging this lead, we launched the industry’s first 256GB low-power system-level chip (LP SoC-M2) based on 1γ process, supporting up to 2TB per CPU, four times larger than a year ago.

We expect low-power DRAM applications in data centers to continue expanding over the coming years, and the company will maintain its industry-leading innovation roadmap in this market. The rapid development of AI has driven new computing architectures optimized for specific workloads, including token economy, with products like high-bandwidth memory (HBM), low-power memory (LP), DDR, and SSD being core enablers. At NVIDIA’s GTC, the Grok 3 LPX model with up to 12TB DDR5 memory exemplifies this trend.

Driven by AI applications such as vector databases, key-value cache offloading, and increasing SSD capacity share, NAND flash demand in data centers is accelerating. Micron’s vertically integrated product portfolio covers everything from high performance to ultra-large capacity. Currently, mass production of PCIe Gen6 data center SSDs based on 9th-gen NAND is underway. Our 122TB SSD has received high market recognition, with continuous read throughput 16 times that of similarly sized HDDs. Our strategic deployment and operational execution continue to translate into strong results.

In 2025, Micron’s data center SSD market share has grown for four consecutive years, reaching a new high. In Q2 FY2026, revenue from data center NAND doubled QoQ to a record high, with further growth expected next quarter. Our data center SSD product lineup is industry-leading, with many design wins. Future demand for NAND is expected to far exceed current supply capacity.

In 2026, due to supply shortages of DRAM and NAND, PC and smartphone shipments may decline low double digits. However, long-term, AI’s commercial value will drive substantial growth in storage capacity in PCs and smartphones. For PCs, generative AI applications are exploding, e.g., OpenClaw’s AI agent can perform tasks locally or request cloud processing. PCs with terminal AI features are recommended with at least 32GB storage, double that of typical PCs. Emerging AI workstations like NVIDIA DGX Spark and AMD Ryzen AI Halo, equipped with 128GB storage, are ideal for running large language models locally.

In smartphones, flagship models with generative AI features, such as Samsung Galaxy S26 and Google Pixel 10, have been released, with AI functions integrated into the OS. By Q4 2026, nearly 80% of flagship phones will have 12GB+ DRAM, up from less than 20% a year earlier. Micron is well-positioned to seize this opportunity with its leading product portfolio.

In PCs, we have completed certification of low-power computing memory modules (LPCAM2) with a major OEM. We launched the industry’s first 9th-gen QLC SSD for client use.

Micron’s LPDDR5X (5X) has been successfully introduced into mainstream AI workstations, expanding our target markets with large-volume shipments to key customers. In smartphones, samples of 1γ process-based LPDDR6 are highly regarded by manufacturers and partners. Our 10.7 Gbps 1γ LPDDR5X 16Gb products have completed more certifications and are in mass production, with market momentum rising. Prices for automotive, industrial, and embedded markets continue to climb, with AEBU revenue hitting record highs, with automotive and industrial combined exceeding $2 billion this quarter.

In automotive, automakers are accelerating deployment of L2+ ADAS across all models. Typical cars have less than L2 automation with about 16GB DRAM; L4 autonomous vehicles require over 300GB DRAM. As advanced ADAS and smart cockpits become widespread, long-term automotive storage demand is expected to grow strongly. We introduced the industry’s first automotive-grade 1γ LPDDR5 DRAM sample; in NAND, we launched UFS 4.1, based on 9th-gen NAND, further strengthening our technological leadership.

Rapid AI technology iteration is greatly enhancing robot capabilities. We believe the robotics industry will enter a 20-year growth cycle, potentially becoming one of the largest product categories in tech. Humanoid robots equipped with AI systems will have computing platforms comparable to high-end L4 autonomous vehicles, requiring massive storage and memory. We expect this emerging growth area to further solidify the long-term positive trend in the storage industry. With leading technology, products, and close customer collaboration, Micron is well prepared to capitalize on robotics opportunities.

Next, I will share market outlook. By 2026, industry-wide DRAM and NAND bit demand will remain supply-constrained, with shortages expected to persist beyond 2026. In 2026, DRAM bit shipments are projected to grow about 20%, slightly above previous estimates. Factors include limited cleanroom capacity, long construction cycles, increased trade of high-bandwidth memory (HBM), accelerated demand growth, and wafer output slowdown due to process node upgrades. NAND bit shipments are expected to grow about 20% in 2026. Some suppliers are shifting cleanroom capacity from NAND to DRAM, constraining NAND supply growth. Micron’s DRAM and NAND bit growth will align with industry averages.

To address the significant supply-demand gap, Micron is actively expanding capacity globally, achieving key milestones this quarter. In DRAM, we announced early acquisition of Lehi Fab, with existing wafer capacity expected to ship in FY2028. We plan to start building a second cleanroom at this site before 2026. The Idaho fab is expected to begin wafer production mid-2027, with site prep underway. The New York fab has broken ground, with site prep progressing well. In Hiroshima, cleanroom expansion is on track, supporting future process node upgrades.

In NAND, driven by increased market demand and strategic decision to co-locate R&D cleanrooms with production fabs, we plan to build a new NAND fab in Singapore, with first wafers in 2028. Our India packaging and testing plant is now commercially operational, becoming one of the largest single-layer packaging cleanrooms globally. The high-bandwidth memory (HBM) advanced packaging plant in Singapore is progressing as planned, expected to support Micron’s HBM supply in 2027. Capital expenditures for FY2026 are projected to exceed $25 billion, mainly driven by cleanroom investments, especially the Lehi Fab acquisition and construction, followed by increased US wafer fab investments.

For FY2027, capital expenditures are expected to rise sharply, primarily for high-bandwidth memory (HBM) and DRAM investments. Construction-related capex will increase by over $10 billion YoY, mainly for global wafer fab builds to meet long-term demand. Equipment purchases are also expected to grow. While making these investments, we will continuously monitor market conditions and customer needs, adjusting supply plans accordingly. Next, Sajan will summarize the financial results for Q2 FY2026.

Sanjay Mehrotra: Thank you, Mark. Decades of sustained investment, innovation, and operational execution have made Micron a leader in storage technology and one of the biggest beneficiaries and enablers in the AI sector of the semiconductor industry. As the only advanced storage product manufacturer in the US, Micron has a unique advantage in seizing unprecedented future growth opportunities. I thank all our global employees for their hard work; it is their excellent execution that has delivered this outstanding quarter. Despite this strong performance, I am even more optimistic about Micron’s future development. Now, let’s open the floor for questions.

Moderator: Thank you. We will now begin the Q&A session. Each questioner may ask one main question and one follow-up. The first question comes from Kresh Sangal of RBC Capital Markets. Please go ahead.

Kresh Sangal: The company’s guidance of 81% gross margin is very impressive. I’d like to understand, as HBM4 production scales up, how sustainable is this gross margin? Could you analyze the outlook for Q4 and beyond? Also, I have a follow-up question for Sanjay.

Mark Murphy: Kresh, this is Mark. The company’s guidance for Q3 gross margin has increased significantly by 6 percentage points, showing strong momentum. Currently, we are not releasing specific guidance for Q4 gross margin, but as previously stated, supply tightness in the industry is expected to persist beyond 2026. Therefore, gross margins in Q4 and subsequent quarters will continue to benefit from multi-year AI-driven investment cycles, with most of the benefits still ahead. AI requires more and higher-performance storage products, which will directly impact gross margins. Additionally, as mentioned earlier, industry supply constraints will support margins through 2026 and beyond.

The 81% gross margin guidance already factors in growth from HBM4 mass production, and as I said, the market environment remains favorable. However, at such high gross margin levels, the marginal benefit of product pricing increases may diminish. Beyond that, we are not disclosing further details on Q4 gross margin at this time.

Kresh Sangal: Understood. Next, I’d like to ask Sanjay about the strategic customer agreement (SCA). Congratulations on signing the first five-year SCA. How does this differ fundamentally from previous long-term agreements (LTAs)? Does it include multi-year shipment volume and pricing lock-ins, or are prices still negotiated annually? Also, in a down cycle, what are the termination clauses?

Sanjay Mehrotra: Thank you. I appreciate your interest in our first SCA. As I mentioned earlier, the SCA is a multi-year agreement, replacing the typical one-year LTA. Given the foreseeable industry supply tightness, customers are eager to sign such structured long-term strategic agreements to better plan their business and improve predictability. These agreements provide stability for customers and enhance our business visibility.

Since we have only signed one SCA so far, we cannot disclose specific details, and I trust you understand the confidentiality involved. The core goal is to give customers clear supply commitments for planning, while also securing their explicit cooperation. The agreement’s term covers different phases of the industry cycle, whether tight supply or others, and the clauses are binding for both parties, with well-defined rights and obligations.

Kresh Sangal: Thank you very much.

Moderator: The next question is from Joseph Muir of Morgan Stanley. Please go ahead.

Joseph Muir: I’d like to ask about product allocation strategies across end markets. Clearly, AI demand is most urgent, but is there concern about demand decline in PCs and smartphones? How does the company balance the needs of large customers versus small and medium ones? What is the overall allocation approach?

Sanjay Mehrotra: Currently, supply across all end markets is extremely tight, with demand trends remaining strong. While some price-sensitive markets like consumer electronics may see demand affected by price increases, overall demand remains robust.

Our core strategy is to be a diversified supplier across all end markets, which is vital for our growth. Data centers are becoming an increasingly important part of the total addressable market (TAM), so our supply resources are shifting toward that segment, which is also the main growth driver.

However, markets like PCs, smartphones, automotive, and industrial are equally important. We aim to maintain a diversified product portfolio across these segments. It’s worth emphasizing that AI trends are continuously pushing for higher storage capacities in all these markets. Under tight supply conditions, customers are actively adjusting their product mixes. Overall, we are maintaining close communication and cooperation with customers across all end markets.

Joseph Muir: Got it, thank you. I recall the company previously mentioned that about 70% of customer demand is currently met; has this ratio changed? Do different customers have different fulfillment rates? How does this compare to three months ago?

Sanjay Mehrotra: As we stated in our last earnings call, in the medium term, for some core customers, we can only meet about 50% to two-thirds of their demand, and this situation remains unchanged.

Joseph Muir: Understood, thank you. Excellent quarterly performance, appreciate the insights.

Moderator: The next question is from Timothy Michael Ackery of Barclays. Please go ahead.

Timothy Michael Ackery: Thank you. Sanjay, I’d like to ask about the strategic customer agreement (SCA). Given industry downturns, we hope these agreements can include protections for gross margin decline. I understand confidentiality, but can you share whether the agreement includes mechanisms to limit margin erosion during downturns?

Sanjay Mehrotra: Due to confidentiality, we cannot disclose specific SCA terms. Currently, we have only one SCA signed and are negotiating with multiple customers. Once more agreements are signed, we will share details as appropriate. I want to emphasize that the SCA is a multi-year agreement with clear commitments, designed to improve visibility and stability. Beyond that, I cannot disclose further details at this time.

Timothy Michael Ackery: Thank you. Mark, I have a cash flow question. The company expects free cash flow of $35-40 billion this year, and by the end of 2026, cash reserves could exceed $50 billion. How does the company plan to use these funds? Will some be reserved for buybacks during downturns? Also, given the restrictions from the CHIPS Act, are there plans to seek adjustments to buyback terms? Thanks.

Mark Murphy: Timothy, our strong business performance and ongoing balance sheet optimization are very encouraging. In Q2, net cash and free cash flow hit record highs, with free cash flow up 77% from the previous record. Our Q3 guidance suggests free cash flow could nearly double QoQ, supported by capex plans. We will continue to strengthen our balance sheet, further increasing net cash, and reducing debt. Notably, our credit ratings have been upgraded twice this quarter, now at a stable BBB.

Our financial position remains strong, and as you noted, we plan to increase capex and R&D investments. Our capital allocation priority remains maintaining a solid balance sheet, with organic investments in technology upgrades and high-value capacity expansion, which are key market needs. Currently, our capital return rate exceeds 30% and could rise further, but we remain cautious. As announced today, we are increasing quarterly dividends by 30%, reflecting confidence in our outlook, stability, and future cash distribution plans.

Additionally, we plan to use ample funds for share repurchases to reward shareholders, hedge against equity dilution from stock-based compensation, and seize market opportunities.

Timothy Michael Ackery: Thank you, Mark.

Moderator: The next question is from Christopher James Mues of Barclays. Please go ahead.

Christopher James Mues: Good afternoon, thank you for the insights. I’d like to follow up on the SCA. The evolution from LTA to binding LTA and now to SCA—what is the scope of current SCA negotiations? Are they limited to mega cloud providers, or do other customer types participate? Also, do these agreements include preconditions related to company capital expenditures? Are pricing terms linked to ROIC on these investments? Thanks.

Sanjay Mehrotra: We have previously disclosed that our first SCA partner is a large customer. The core goal is to enable us to plan capacity investments with more confidence and to improve demand visibility through clear terms. As I mentioned earlier, this also enhances the stability of our business model. We are not disclosing more details now, but we are negotiating with multiple customers across various markets.

Christopher James Mues: Thank you. One more question about HBM. Last quarter, you guided a 40% CAGR for HBM market, estimating a $50 billion market in 2026. Has this outlook changed? Given higher margins for non-HBM products, are some players shifting capacity from HBM to DDR5? Thanks.

Sanjay Mehrotra: Yes, non-HBM products currently have higher margins than HBM, but demand for HBM remains strong. We have not revised our previous HBM TAM outlook. In data centers, demand for DDR5, low-power memory, and HBM continues robustly. As AI data center demand grows, we will continue optimizing our product mix. Besides data centers, we will focus on other core markets to maintain market share.

Demand from data centers to edge AI remains strong, and we will keep refining our product portfolio. Our product lineup—including HBM, LP, SoC-M, DDR5, and SSD—has seen significant growth in data center market share over recent years.

Moderator: The next question is from Vivek Arya of Bank of America Securities. Please go ahead.

Vivek Arya: Thank you. Sanjay, I’d like to ask about HBM market share targets—do you aim for 20-25% directly, or is it a gradual increase? How does Micron plan to grow HBM share supporting NVIDIA Vera Rubin?

Sanjay Mehrotra: We previously indicated that in Q3 2025, we achieved our HBM market share goal for 2026, reaching parity with the overall DRAM market. We also said that going forward, HBM will be managed as part of our overall product portfolio, and quarterly share figures will no longer be separately disclosed. I can tell you that we are confident in HBM’s market position and competitiveness.

In 2026, both HBM3E and HBM4 will be mainstream products, and we will supply both, confident in our overall market strategy and product mix management.

Vivek Arya: Thanks. Mark, regarding the 81% gross margin guidance—I understand you won’t provide detailed outlook, but historically, gross margins peaked just above 60%. How does current market environment differ? How do you see margins evolving over the next few quarters? When customers see such high margins, how might they react as a core AI chip supplier? Thanks.

Sanjay Mehrotra: Before Mark responds, I’d like to correct a slip-up. I previously misstated the year; the correct statement is that Micron aimed to reach HBM market share parity with the overall DRAM market in 2025, which we achieved ahead of schedule in Q3 2025. Since then, we no longer disclose HBM share separately. Sorry for the confusion.

Mark Murphy: Vivek, it’s important to note that the industry is in a supply shortage that will last beyond 2026, providing strong support for short- and medium-term pricing. We have been working closely with customers to meet demand through rational product allocation, capacity expansion, supply assurance, and new product development. The high gross margin guidance reflects the current market environment, especially the value of storage in AI. As Sanjay explained, AI needs more and higher-performance storage, which reduces token costs, energy per token, and increases token throughput, further boosting storage demand.

High margins demonstrate the significant value of storage products in AI, and this trend extends from data centers to edge applications. Over a year ago, we pointed out that multiple supply constraints—low inventory, process node upgrades, HBM trade ratios, and new capacity building—are long-term issues. Our strategies include capacity expansion and R&D investment to enhance storage value, supporting long-term margin stability. Customers recognize this, which is why they sign long-term agreements.

Moderator: This concludes the conference call. Thank you all for participating. The call is now concluded.