---
title: "Micron conference call: AI will reshape storage into a \"strategic asset\"! To cope with shortages, it is necessary to spend money to build factories and sign the first 5-year long-term contracts, HBM4 directly supplies NVIDIA"
type: "News"
locale: "zh-CN"
url: "https://longbridge.com/zh-CN/news/279697207.md"
description: "In its Q2 fiscal year 2026 earnings report, Micron Technology reported a nearly twofold year-on-year revenue increase to $23.9 billion, with a record gross margin of 75%. The company plans to exceed $25 billion in capital expenditures for fiscal year 2026, with construction expenditures in fiscal year 2027 expected to increase by over $10 billion. Investors have expressed concerns about the high spending and future cash flow. Micron stated that the demand for high-bandwidth memory driven by AI has forced it to increase investments and change its business model, and it has signed a long-term supply agreement with NVIDIA"
datetime: "2026-03-19T02:18:01.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/279697207.md)
  - [en](https://longbridge.com/en/news/279697207.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/279697207.md)
---

> 支持的语言: [English](https://longbridge.com/en/news/279697207.md) | [繁體中文](https://longbridge.com/zh-HK/news/279697207.md)


# Micron conference call: AI will reshape storage into a "strategic asset"! To cope with shortages, it is necessary to spend money to build factories and sign the first 5-year long-term contracts, HBM4 directly supplies NVIDIA

On March 18, Micron Technology's Q2 fiscal year 2026 earnings report showed that the company's revenue nearly doubled year-on-year to approximately $23.9 billion, with gross margins soaring to a record 75%, and provided guidance for gross margins as high as 81% for Q3. At the same time, Micron announced that capital expenditures for fiscal year 2026 would exceed $25 billion, and construction-related expenditures for fiscal year 2027 would increase by more than $10 billion year-on-year. After hours, Micron's stock price initially rose before falling, with a maximum decline of nearly 5%.

The core point of divergence in the market lies in "money." During the subsequent conference call held by Micron, investors' most pressing questions focused on three areas: why initiate such a massive capital expenditure plan at this time? Is the guidance for gross margins as high as 81% sustainable? And what is Micron's true position in NVIDIA's AI chip supply chain?

According to the guidance, Micron's capital expenditures for fiscal year 2026 (ending this August) will exceed $25 billion, far exceeding analysts' expectations of $22.4 billion. Construction-related capital expenditures for fiscal year 2027 will surge by more than $10 billion on top of this. **This "burning money" pace has raised investors' concerns about future free cash flow and overcapacity, and the high costs overshadow strong sales expectations.**

**Sanjay explained, "Most of the increase in expenditures is driven by capital expenditures related to cleanroom facilities." This includes expansions of the factory in Tainan, Taiwan, and the U.S. factory.** The insatiable demand for high-bandwidth memory (HBM) is consuming existing capacity. Due to the larger die size and more complex processes of HBM chips, their consumption of wafers far exceeds that of traditional DRAM.

**At the same time, the physical limits of advanced processes are also approaching.** Sanjay pointed out, "The growth in bits per wafer due to node migration is declining." This means that simply relying on technological upgrades can no longer meet the market's thirst for capacity; substantial investments are needed to build new factories and purchase extreme ultraviolet lithography (EUV) machines. The construction cycle for new capacity takes several years, and Micron must pay now for AI demand in 2027 and even 2028.

To hedge against the risks of massive investments, Micron is changing its business model. Sanjay Mehrotra announced, **the company has signed its first five-year strategic customer agreement (SCA), which is distinctly different from the typically one-year long-term agreements (LTA) of the past.**

At the same time, the market is highly focused on Micron's progress in the HBM field and its partnership with NVIDIA. Sanjay Mehrotra confirmed during the call that Micron began mass production and shipment of the HBM4 36GB 12H product in Q1 2026, which is specifically designed for NVIDIA's Vera Rubin architecture.

"As HBM4 production progresses and mass shipments commence, we expect to reach mature yields faster than HBM3E," Sanjay Mehrotra emphasized. Additionally, Micron is developing the next generation HBM4E, which is expected to enter mass production in 2027, further solidifying its position in the high-end AI accelerator supply chainThis breaks the traditional storage cycle fate of "price surges due to supply-demand mismatch, followed by price plummets due to overcapacity." Sanjay added a striking statement: **"AI not only increases the demand for storage, it fundamentally reshapes storage into a decisive strategic asset of the AI era."**

Currently, there exists a "barrel effect" in the market: the computational bottleneck of large AI models is no longer in computing chips, but in the bandwidth and capacity of memory. Without faster and larger memory, AI cannot further reduce the computational cost of tokens, nor can it support more complex Agentic AI reasoning.

Not only HBM, but the demand for traditional data center SSDs is also surging. The conference call revealed that due to the increase in AI use cases such as vector databases and cache offloading, Micron's data center NAND revenue doubled quarter-over-quarter in the second fiscal quarter, with demand far exceeding available supply.

Regarding the sustainability of the cycle, Mehrotra bluntly stated: "Supply is extremely tight, and all end markets are tight." When analysts inquired about the actual procurement situation of customers, he confirmed that the previous quarter's statement still holds: **"For some key customers in the medium term, we can only meet 50% to two-thirds of their demand, and this situation still exists."**

The "hard constraints" on the supply side were repeatedly mentioned: cleanroom limitations, long new construction cycles, increased trade ratio of HBM occupying wafer capacity, slowed growth of bits per wafer due to advanced process migration, and low inventory levels. Murphy also stated more directly when answering about the sustainability of gross margins: "We have pointed out that market conditions are expected to remain very tight even after 2026."

Micron Technology Q2 Fiscal Year 2026 Earnings Call

Event Date: 03/18/2026 Company Name: Micron Event Description: Q2 2026 Earnings Call Source: Micron

Presentation

Operator: Ladies and gentlemen, thank you for joining us, and welcome to Micron Technology's Q2 Fiscal Year 2026 Financial Conference Call. After today's prepared remarks, we will host a Q&A session. Now, I will turn the call over to Satya Kumar from the Investor Relations team. Satya, please go ahead.

Satya Kumar, Vice President of Corporate and Investor Relations and Chief Financial Officer: Thank you, everyone, and welcome to Micron Technology's Q2 Fiscal Year 2026 Financial Conference Call. Joining me today are our Chairman, President, and CEO Sanjay Mehrotra, as well as our Chief Financial Officer Mark Murphy. Today's meeting is being webcast live through our Investor Relations website (investors.micron.com), including audio and slides. Additionally, a press release detailing our quarterly performance and the prepared remarks for this meeting have also been posted on the websiteToday's discussion contains forward-looking statements that are subject to risks and uncertainties. These forward-looking statements include statements regarding our future financial and operational performance, as well as trends and expectations related to our business, customers, markets, industries, products, regulations, and other matters. These statements are based on our current assumptions, and we do not undertake any obligation to update these statements. Please refer to the financial reports in our recent filings with the U.S. Securities and Exchange Commission (SEC), including Form 10-K, Form 10-Q, and other documents, for more information about the risks and uncertainties that could cause actual results to differ materially from those anticipated.

Unless otherwise noted, today's discussion of financial performance is based on Non-GAAP reporting. A reconciliation of GAAP to Non-GAAP financial metrics can be found on our website. Now I will turn the meeting over to Sanjay.

Sanjay Mehrotra, Chairman, President, and CEO: Thank you, Satya.

**Micron achieved extraordinary performance in the second fiscal quarter, setting remarkable records in revenue, gross margin, earnings per share (EPS), and free cash flow. Quarterly revenue nearly doubled year-over-year, with record revenues in DRAM, NAND, HBM, and across various business segments. Our revenue guidance for the third fiscal quarter exceeds the total revenue for any full year in the company's history up to fiscal year 2024. Looking ahead to the third fiscal quarter, we expect revenue, gross margin, EPS, and free cash flow to reach extraordinary historical highs.**

Based on our continued confidence in the strength of our business, I am pleased to announce that the Board of Directors has approved a 30% increase in our quarterly dividend. The improvement in our performance and outlook is driven by the growth in AI-driven storage demand, structural supply constraints, and Micron's comprehensive and robust execution. Our memory and storage solutions are at the core of this AI revolution. Memory makes AI smarter and more powerful, supporting longer context windows, deeper reasoning chains, and multi-agent orchestration.

As AI evolves, we expect computing architectures to become increasingly reliant on memory. This is why we firmly believe that Micron is one of the biggest beneficiaries and enablers of the AI wave. **AI not only increases the demand for memory but fundamentally reshapes memory as a decisive strategic asset in the AI era.**

**We continue to advance the signing of "Strategic Customer Agreements (SCA)" with our customers. Unlike past "Long-Term Agreements (LTA)," SCAs involve specific commitments spanning multiple years, providing better visibility and stability for our business model. These SCAs also offer greater certainty for customers in planning their own businesses while strengthening their long-term investment in our broad product portfolio. We are very pleased to have signed our first five-year SCA.**

We have made significant progress in advancing the industry-leading 1-gamma DRAM and G9 NAND technology node mass production. We expect 1-gamma to become the highest volume node in Micron's historyOur 1-gamma node has become the fastest node to achieve mature yield, with a ramp-up speed that exceeds all previous nodes in our history, and it is expected to account for the vast majority of our DRAM bit mix by mid-2026. We plan to increase the application of extreme ultraviolet lithography (EUV) in the 1-delta DRAM node, utilizing the latest generation of EUV equipment. These more advanced EUV devices will help us optimize cleanroom space efficiency and patterning processes as we evolve towards the 1-delta and more advanced nodes. In terms of NAND, our G9 node is also progressing as planned, and it is expected to account for the majority of bit output by mid-2026. This quarter, our QLC bit ratio also reached a historical high.

Looking ahead, we expect to concentrate our research and development (R&D) and mass manufacturing operations at our Boise and Singapore facilities, which will accelerate the speed at which we bring our cutting-edge products to market. We see unprecedented opportunities for memory and storage to empower the AI era across various segments, and therefore we anticipate significantly increasing our R&D investments in fiscal year 2027.

Micron's technological leadership, outstanding products, and manufacturing execution are being recognized in customer quality ratings. I am pleased to report that the vast majority of customers rank Micron first in quality.

Turning to our end markets. **AI demand is driving the total addressable market (TAM) for DRAM and NAND bits in data centers to exceed 50% of the entire industry TAM for the first time in 2026.** Traditional server demand remains strong, driven by workload demands from agentic AI and widespread server refresh cycles. AI server demand continues to be robust. Currently, both AI and traditional server demand are constrained by insufficient DRAM and NAND supply. With the growth of AI and traditional servers, we expect server shipments to achieve low-teens growth in 2026. With the introduction of new platforms, we expect the DRAM attachment rate in servers to continue to grow in 2026.

At NVIDIA's GTC conference, **we announced that Micron has begun volume shipments of the HBM4 36GB 12-layer (12H) product designed specifically for NVIDIA Vera Rubin in the first quarter of 2026. As HBM4 capacity ramps up and volume shipments commence, we expect its speed to reach mature yield to be faster than HBM3E. We also provided samples of the HBM4 16-layer (16H) product, which offers 48GB of capacity per HBM module, a 33% increase over HBM4 12H. Our next-generation HBM product, HBM4E, is progressing smoothly in development and is expected to achieve mass production in 2027.** Our HBM4E will utilize Micron's production-validated, industry-leading 1-gamma DRAM technology node and is expected to achieve a step change in performance, empowering a new generation of AI computing platforms across the industryIn addition, the customization options of HBM4E provide us with further opportunities for differentiation and enable deeper R&D collaboration with customers.

**Micron has pioneered LPDRAM specifically designed for data centers,** with power consumption only one-third that of DDR DRAM server modules. Based on this leading advantage, we have launched samples of the industry's first 256GB LP SOCAMM2 product. This product is manufactured using our 1-gamma node and can provide up to 2TB of massive capacity for each CPU, which is four times the capacity of a year ago. We see the application of LPDRAM in data centers expanding in the coming years, and we are pleased to maintain an industry-leading and innovative product roadmap in this market.

The rapid growth of AI inference is driving the emergence of new architectures optimized for specific workloads, known as "Token economics." Micron's extensive portfolio of HBM, LP, DDR, and SSD products is a key enabler of these architectures. In the latest release at the GTC conference, NVIDIA and Groq 3 LPX achieved up to 12TB of DDR5 memory in a rack-level architecture.

With the emergence of AI use cases such as vector databases and KV cache offload, along with the growing share of SSDs in the capacity storage layer, we see an acceleration in data center demand for NAND-based storage. Micron's data center SSD product portfolio benefits from our technological leadership and vertical integration capabilities, covering the full range from highest performance to maximum capacity. We are currently mass-producing PCIe Gen6 high-performance data center SSDs based on G9 NAND. Our 122TB high-capacity SSD is gaining widespread adoption, with its sequential read throughput per watt being 16 times that of equivalent capacity mechanical hard disk (HDD) configurations. Our strategy and execution are translating into tangible results. In 2025, our market share in the data center SSD market has grown for the fourth consecutive year, setting a new record. In the second fiscal quarter, data center NAND revenue more than doubled quarter-over-quarter, reaching a new historical high, and we expect further growth in the upcoming quarters. Micron's data center SSD product portfolio is industry-leading, and we have secured a significant number of solid design wins among our customer base.

In the foreseeable future, we see that the demand for NAND will greatly exceed our available supply. In 2026, various factors, including constrained DRAM and NAND supply, may lead to a slight low double-digit decline in PC and smartphone shipments. However, over time, we expect the value of on-device AI to drive strong growth in memory capacity in PCs and smartphones. In the PC space, there have been some exciting innovations recently, such as the OpenClaw application based on intelligent agent AI, which allows AI agents to perform tasks independently on host PCsAt the same time, cloud workloads can also be initiated. PCs equipped with edge AI capabilities have a recommended memory configuration of at least 32GB, which is twice that of ordinary PCs. In addition, the rapidly growing category of personal AI workstations (such as NVIDIA DGX Spark and AMD Ryzen AI Halo) is equipped with a memory configuration of 128GB, making them very suitable for running large language models on the edge.

Similarly, in the smartphone sector, major OEM manufacturers have recently released new flagship devices (such as Samsung Galaxy S26 and Google Pixel 10), all of which have integrated AI capabilities into their mobile operating systems. The shipment share of flagship smartphones with 12GB or higher DRAM capacity increased to nearly 80% in the fourth quarter of last year, compared to less than 20% a year ago. With an industry-leading product portfolio, Micron is well-prepared to accelerate its capture of opportunities in these markets.

**In the PC sector, Micron has completed the certification of LPCAMM2 by a major OEM manufacturer.** At CES, we launched the industry's first Gen5 QLC client SSD based on G9 NAND. Micron's LPDDR5X is now designed for leading personal AI workstations, further expanding our addressable market with significant shipments to key customers. In the smartphone sector, Micron continues to receive strong interest and positive feedback from OEMs and ecosystem partners regarding our 1-gamma node-based LPDDR6 samples. We have further built momentum by completing additional certification and mass production of the 10.7 Gbps 1-gamma LPDDR5X 16Gb product.

We are seeing continued improvement in pricing conditions in the automotive, industrial, and embedded markets. AEBU (Automotive and Embedded Business Unit) achieved record total revenue, with combined revenue from the automotive and industrial sectors exceeding $2 billion this quarter. In the automotive sector, OEM manufacturers are accelerating the deployment of L2+ level ADAS (Advanced Driver Assistance Systems). Currently, the ADAS capabilities of ordinary cars on the market are below L2 level, containing about 16GB of DRAM; while vehicles with L4 level autonomous driving capabilities require over 300GB. With the proliferation of higher-level ADAS and intelligent cockpits, we expect strong long-term growth in automotive memory demand. We have showcased the industry's first automotive-grade 1-gamma LPDDR5 DRAM samples; in NAND, we have taken the lead in launching UFS 4.1 automotive solutions based on the G9 node, further consolidating our technological leadership in this market.

**The rapid advancement of AI is greatly enhancing the capabilities of robots.** We believe we are at the starting point of a 20-year growth trajectory for robotics, with robots expected to become one of the largest product categories in the tech sector. Humanoid robots will be fully equipped with AI capabilities and driven by computing platforms comparable to high-end L4 level autonomous vehicles, thus requiring massive memory and storage capacityWe expect this exciting new emerging growth category to further solidify the long-term favorable factors shaping our industry environment. Micron is in a very advantageous position to leverage this opportunity through close collaboration with customers, thanks to its industry-leading technology, product solutions, and operational capabilities.

Now turning to our market outlook. We anticipate that the bit demand for both DRAM and NAND in the industry will be constrained by supply in 2026. **We continue to expect that the supply-demand situation for DRAM and NAND will remain tight after 2026. We expect the total industry DRAM bit shipments to grow in the low 20s percentage in 2026, slightly higher than our previous expectations. In terms of DRAM, limitations in cleanroom space, long construction cycles, a higher trade ratio of HBM wafer consumption, a higher growth rate of HBM, and a decline in single wafer bit output due to node migration have collectively constrained the growth of bit supply.** We expect the total industry NAND bit shipments to grow by approximately 20% in 2026. In terms of NAND, some industry suppliers are reallocating cleanroom space for DRAM production, coupled with the overall limited cleanroom space, which constrains the growth of bit supply. We expect Micron's supply growth for DRAM and NAND in 2026 to be roughly in line with the industry average.

Micron is working to address the unprecedented supply-demand gap, and last quarter we achieved several important milestones in expanding our global manufacturing footprint. In terms of DRAM, earlier this week, we announced the successful completion of the acquisition of Powerchip's Tongluo factory, ahead of schedule. We expect this factory to provide meaningful product shipment support for its existing facilities starting in fiscal year 2028. Based on the existing facilities, we plan to initiate the construction of a comparably sized second cleanroom at this site by the end of fiscal year 2026. We continue to expect our first wafer fab in Idaho to produce its first wafers in mid-2027, while ground preparation for the second Idaho wafer fab has also begun. Our first wafer fab in New York has broken ground, with initial ground preparation ahead of schedule. In Japan, ground preparation for the cleanroom expansion at our Hiroshima factory is progressing smoothly, which will support future technology transitions.

In terms of NAND, higher demand expectations and our decision to integrate R&D cleanrooms with manufacturing wafer fabs have prompted us to decide to construct a new NAND wafer fab at our Singapore site. We expect this fab to produce its first wafers in the second half of 2028.

In packaging and testing, our new factory in India has begun commercial shipments. This state-of-the-art facility will become one of the largest single-layer packaging and testing cleanrooms in the world. Our advanced HBM packaging facility in Singapore is progressing as planned and is expected to make significant contributions to Micron's HBM supply in 2027**We expect capital expenditures (CapEx) for the fiscal year 2026 to exceed $25 billion. The increase compared to our previous earnings call expectations is primarily driven by capital expenditures related to cleanroom facilities, with the largest component being the copper gong factory, followed by increased spending on our U.S. wafer fab project.** We anticipate a significant increase in capital expenditures for the fiscal year 2027 to support investments related to HBM and DRAM. As we need to expand our global manufacturing facilities to meet long-term demand opportunities, **we expect construction-related capital expenditures for the fiscal year 2027 to increase by more than $10 billion year-over-year.** Additionally, we expect equipment spending for the fiscal year 2027 to also increase year-over-year. While making these investments, we will continue to respond sensitively to market conditions and customer demand to adjust our supply plans accordingly.

Now, I will hand the meeting over to Mark, who will introduce our financial performance and outlook for the second fiscal quarter.

Mark Murphy, Executive Vice President and Chief Financial Officer: Thank you, Sanjay, and good afternoon, everyone.

Micron achieved strong financial performance in the second fiscal quarter, with revenue, gross margin, and earnings per share exceeding the upper end of our guidance. In the second fiscal quarter, we generated record free cash flow, reduced debt, and ended the quarter with the highest net cash position in the company's history.

Total revenue for the second fiscal quarter was $23.9 billion, a 75% increase quarter-over-quarter and a 196% increase year-over-year, marking the fourth consecutive quarter of record revenue. The $10.2 billion quarter-over-quarter increase is also the largest in our history. Second fiscal quarter DRAM revenue reached a record $18.8 billion, a 207% year-over-year increase, accounting for 79% of total revenue. Quarter-over-quarter, DRAM revenue grew by 74%. Bit shipments achieved mid-single-digit growth. Due to tight industry conditions and a favorable product mix, product prices increased by mid-60s percent. Second fiscal quarter NAND revenue reached a record $5 billion, a 169% year-over-year increase, accounting for 21% of Micron's total revenue. Quarter-over-quarter, NAND revenue grew by 82%. NAND bit shipments achieved low-single-digit growth. Similarly, due to tight NAND industry conditions and a favorable product mix, product prices increased by high-70s percent.

The consolidated gross margin for the second fiscal quarter was 75%, an 18 percentage point increase quarter-over-quarter. This improvement was primarily driven by price increases, along with a favorable product mix and cost control performance. The gross margin for the second fiscal quarter nearly doubled compared to the same period last year, setting a new company record.

Now, looking at the quarterly financial performance by business segment: The Cloud Memory Business Unit (CMBU) achieved record revenue of $7.7 billion, accounting for 32% of the company's total revenue. Driven by price increases and a favorable product mix, CMBU revenue grew by 47% quarter-over-quarter. CMBU's gross margin was 74%, driven by price increases and cost execution, improving by 9 percentage points quarter-over-quarterThe core data center business unit (CDBU) achieved record revenue of $5.7 billion, accounting for 24% of the company's total revenue. Driven by price increases and a favorable product mix, the CDBU gross margin was 74%, an increase of 23 percentage points quarter-over-quarter. The mobile and client business unit (MCBU) achieved record revenue of $7.7 billion, accounting for 32% of the company's total revenue. Driven by price increases (partially offset by lower bit shipments), MCBU revenue grew 81% quarter-over-quarter. The MCBU gross margin was 79%, primarily driven by price increases and a favorable product mix, which increased by 25 percentage points quarter-over-quarter. The automotive and embedded business unit (AEBU) achieved record revenue of $2.7 billion, accounting for 11% of the company's total revenue. Driven by price increases (partially offset by lower bit shipments), AEBU revenue grew 57% quarter-over-quarter. The AEBU gross margin was 68%, primarily driven by price increases, which increased by 23 percentage points quarter-over-quarter.

Operating expenses for the second quarter were $1.4 billion, an increase of $87 million quarter-over-quarter. The quarter-over-quarter growth was mainly due to increased R&D expenses. We achieved $16.5 billion in operating profit for the second quarter, with an operating profit margin of 69%, an increase of 22 percentage points quarter-over-quarter and 44 percentage points year-over-year. Taxes for the second quarter were $2.5 billion, with an effective tax rate of 15.1%. The second quarter non-GAAP diluted earnings per share (EPS) was $12.20, an increase of 155% quarter-over-quarter and 682% year-over-year.

Turning to cash flow and capital expenditures: Operating cash flow for the second quarter was $11.9 billion. Capital expenditures were $5 billion, generating $6.9 billion in free cash flow. The free cash flow for the second quarter set a company record for a single quarter, exceeding the previous record set in the first quarter of fiscal 2026 by 77%. Ending inventory for the second quarter was $8.3 billion, an increase of $62 million quarter-over-quarter, with inventory turnover days at 123 days. The inventory days for DRAM remain particularly tight, below 120 days. Cash and investments at the end of the quarter reached a record $16.7 billion; including our unused credit lines, liquidity exceeds $20 billion. In the second quarter, under the terms of the CHIPS Act, we repurchased $350 million of stock. This quarter, we also reduced debt by $1.6 billion, including the redemption of senior notes maturing in 2029 and 2030. Our weighted average maturity of outstanding debt is August 2034. At the end of the quarter, our debt balance was $10.1 billion, with a net cash balance of $6.5 billion.

Reinvesting in R&D, capital expenditures, and other strategic investments for profitable growth remains our top priority in capital allocation. We are committed to maintaining a strong balance sheet, having reduced total debt by over $5 billion in the past three quarters, and we are currently in the strongest net cash position in our historyBased on our continued advantages from our technological leadership and strong cash flow generation capabilities, as Sanjay mentioned, the board has approved a 30% increase in the quarterly dividend to $0.15 per share.

Now turning to our performance guidance: We expect third-quarter revenue to reach a record $33.5 billion, with a variance of plus or minus $750 million. **Gross margin is expected to be around 81%, with operating expenses of approximately $1.4 billion.** Based on approximately 1.15 billion shares outstanding, we expect EPS to reach a record $19.50, with a variance of plus or minus $0.40. We anticipate that higher prices, lower costs, and a favorable product mix will collectively drive gross margin expansion in the third quarter. As mentioned last quarter, Micron's operating expenses for the fourth quarter of fiscal 2026 will also reflect the impact of an additional week in this 53-week fiscal year. We expect operating expenses for fiscal 2027 to increase as we ramp up R&D investments to support unprecedented long-term opportunities in the memory and storage sectors. We anticipate a tax rate of approximately 15.1% for the third quarter and the full fiscal year 2026.

Micron continues to invest rigorously on a global scale. To meet customer demand, as previously mentioned, we now expect capital expenditures for fiscal 2026 to exceed $25 billion. In the third quarter, we expect capital expenditures of approximately $7 billion, while strong operating cash flow will lead to significantly increased free cash flow. Due to the demand for cleanroom capacity, we expect our construction spending growth rate to exceed the growth rate of equipment spending in fiscal 2026 and 2027. Any potential impacts from trade or geopolitical developments are not included in our guidance.

Now I will hand the meeting back to Sanjay for a summary.

Sanjay Mehrotra, Chairman, President, and CEO: Thank you, Mark. Decades of investment in innovation and execution have established Micron's technological leadership in the memory and storage sectors and made us one of the biggest beneficiaries and enablers of the AI wave in the semiconductor industry. As the only advanced memory product manufacturer headquartered in the United States, Micron is uniquely positioned to seize the unprecedented opportunities at hand. I want to thank the Micron team members around the globe; it is their outstanding execution that has made this exceptional quarter possible.

While these achievements are impressive, I am even more excited about Micron's future ahead. We will now begin the Q&A session.

Questions And Answers

Operator: Thank you. We will now begin the Q&A session. (Operator instructions) Your first question comes from Krish Sankar of TD Cowen. Please go ahead.

Krish Sankar: Thank you. **Mark, the guidance of an 81% gross margin is quite impressive. I am curious about how you view the sustainability of the gross margin, especially as you introduce more HBM4 into your product mix. If you could share your thoughts on the gross margin trends for the August quarter (i.e., the fourth quarter) and beyond, that would be very helpful.**Then I have a follow-up question for Sanjay.\*\*

Mark Murphy, Executive Vice President and Chief Financial Officer: Krish, this is Mark. We provided a strong guidance for a 600 basis point sequential increase in the third quarter. We will not provide specific gross margin guidance for the fourth quarter. However, we have indicated that we expect market tightness to persist beyond 2026. So, the strong momentum will clearly continue beyond the fourth quarter. What you see in our gross margin is the dividend of a multi-year investment cycle driven by AI, and this cycle has mostly just begun. AI requires more memory and higher performance memory. This is reflected in our margins. Additionally, we have also talked about supply-side factors that will continue to exist beyond 2026. The 81% guidance has already taken into account the growth of HBM4, but as I mentioned, we expect market conditions to remain strong. It is important to remember that at the current gross margin levels, further incremental price increases will have a diminishing marginal impact on gross margins. But aside from that, we will not provide gross margin guidance for the fourth quarter.

Krish Sankar: Understood. Thank you, Mark. Next, I have a quick question for Sanjay about the SCA (Strategic Customer Agreement). **Congratulations on signing your first five-year SCA.** How does it differ from the LTA (Long-Term Agreement)? Is it an agreement that includes multi-year quantity and price commitments, or do prices need to be renegotiated annually? Additionally, if there is a slowdown during the agreement period, how are the cancellation terms in the SCA defined? Thank you, Sanjay.

Sanjay Mehrotra, Chairman, President, and Chief Executive Officer: Thank you for recognizing our achievement in reaching our first SCA. As you pointed out, we also mentioned in our remarks that \*\*the SCA is a multi-year agreement. Previous LTAs were typically one-year agreements. Of course, in the foreseeable future, in an environment of extreme supply tightness, customers are highly motivated to establish these structural strategic agreements with us for their own business planning and better predictability. Of course, these agreements are also intended to bring stability and greater visibility to our business model. We have currently completed one SCA, so we will not delve into the specific details of it. I am sure you can understand that all these agreements are confidential by nature. But these SCAs are designed to achieve the goals of both parties: for customers, it allows them to plan their business and rely on the supply commitments in the agreement; for us, we can also rely on the specific commitments from customers in the agreement. These agreements are intended to span extremely tight periods in the industry and to respond to other industry environments. Therefore, they are long-term agreements and contain strong terms. These terms are guaranteed for both us and our customers.

Krish Sankar: Understood, thank you very muchOperator: Your next question comes from Morgan Stanley's Joseph Moore. Your line is open, please go ahead.

Joseph Moore: Regarding the allocation among various end markets. Clearly, AI is the most pressing area right now, **but are you concerned about demand destruction in markets like PCs and smartphones? How do you balance large customers and small customers? How do you think about this allocation process?**

Sanjay Mehrotra, Chairman, President, and CEO: Clearly, supply is extremely tight, and this tightness spans all end markets. Demand trends across various end markets are strong, although price-sensitive markets like consumer electronics, as you mentioned, may be affected by higher prices, the overall demand in these markets remains quite robust. Our consistent goal and strategy is to be a diversified supplier across all end markets. I think this is very important for us. Of course, data centers are increasingly occupying a larger share of the industry's total TAM. Therefore, the vast majority of supply will flow there, which is also the main driver of growth for the entire industry and for Micron itself. However, other parts of the market are equally important to us, such as PCs, smartphones, automotive, and industrial markets, and we want to maintain a good mix of diversification in the end markets. I would like to point out that overall, whether in data centers or in consumer markets like smartphones or PCs, the trend of AI continues to drive demand for higher memory capacities. Of course, in this tight supply environment, customers are also working to manage their own product mix. But overall, we are working closely with our customers across various end markets.

Joseph Moore: Okay, thank you. I believe you have previously indicated that **some customers can only receive 70% of their requested amounts. Is that still the case now? Has the customer fulfillment rate improved or declined compared to three months ago?**

Sanjay Mehrotra, Chairman, President, and CEO: We mentioned in our last earnings call that for some key customers, **in the medium term we can only meet 50% to two-thirds of their demand. The situation remains the same.**

Joseph Moore: Okay, thank you. Congratulations on a great quarter.

Sanjay Mehrotra, Chairman, President, and CEO: Thank you.

Operator: Your next question comes from Timothy Arcuri. Your line is open, please go ahead.

Timothy Arcuri: Thank you. Sanjay, I also want to ask about SCA. I think we are all trying to think about the other side of the cycle, hoping that these SCAs provide some mechanism to limit your gross margin floor to a certain number during a downturn. I know you don’t want to disclose too many details, but can it be assumed that there is a mechanism in the SCA that can limit the downside of your gross margin during a cycle downturn?Sanjay Mehrotra, Chairman, President, and CEO: For obvious confidentiality reasons, I will not go into the details of these SCAs. We have successfully completed one SCA. We are in discussions with several other clients, and if these agreements are completed in a timely manner, we will certainly share more details with everyone. But I want to emphasize that these SCAs span multiple years and contain specific commitments; moreover, these are very robust agreements. Of course, these agreements are absolutely intended to provide visibility and stability for our future business model. Beyond that, I cannot disclose any details at this time.

Timothy Arcuri: Okay, thank you. Mark, I have a question about cash. You may generate $35 billion to $40 billion in free cash flow this fiscal year. By the end of the calendar year, you may have over $50 billion in cash on your balance sheet. So, what do you plan to do with that money? Are you planning to reserve some funds for significant stock buybacks when the cycle turns? Regarding this, you received funding from the CHIPS Act, which has restrictions on buybacks. Is there any way to renegotiate those restrictions? Thank you.

Mark Murphy, Executive Vice President and Chief Financial Officer: Tim, we are very excited about the business performance and the improvement of our balance sheet. In the second quarter, we achieved a record net cash position and free cash flow, breaking the previous quarter's record and exceeding it by 77%. Based on our guidance for the third quarter, combining these numbers and considering our capital expenditure expectations, we anticipate that cash flow may double quarter-over-quarter. We will continue to strengthen our balance sheet and improve our net cash position. We are continuously deleveraging and paying down debt, and it is worth noting that we received credit upgrades from two rating agencies this quarter, now holding a solid BBB rating. So, as we become stronger, as you can see, we are also discussing increasing capital expenditures and boosting R&D investments. Regarding your specific question about balance sheet priorities or capital allocation: the balance sheet is always a priority; in addition, there is organic investment in the business to advance technology and increase capacity for high-value-added products (we are clearly seeing that demand now), and currently, our capital return rate exceeds 30% and is moving towards 50%, but we will remain disciplined in this regard. As you saw today, we are pleased to announce a 30% increase in the dividend, reflecting our confidence in the business, outlook, business stability, and future cash returns. As you mentioned, we believe there will be significant capacity to return cash to shareholders through buybacks, including offsetting dilution from stock compensation and conducting opportunistic buybacks.

Timothy Arcuri: Thank you, Mark.

Operator: Your next question comes from CJ Muse of Cantor Fitzgerald. Your line is open, please go aheadCJ Muse: Good afternoon, thank you for accepting my question. I would like to follow up on the SCA issue. You have gone through the evolution from LTA (Long-Term Agreement), binding agreements to now SCA. I am curious if you could talk about the breadth of the types of customers you are negotiating with? Is it only hyperscale cloud service providers, or are other types of customers also interested? I know you don't want to delve into the specific details of the contracts, but as a follow-up to the previous question, do these agreements have upfront capital expenditure requirements? Is the pricing linked to the return on invested capital (ROIC) of these investments? Any help you can provide would be greatly appreciated. Thank you.

Sanjay Mehrotra, Chairman, President, and CEO: We will share that the SCA we just signed is with a large customer. Of course, the focus of these agreements is to enable us to confidently invest in future supply plans, and they also include specific terms that give us better visibility into future demand, as I mentioned earlier, which brings overall stability to the business model. CJ, beyond that, we have no further comments on the SCA, but I will say, as I mentioned earlier, we are in discussions with multiple customers regarding these SCAs, and they indeed span multiple markets.

CJ Muse: Very helpful. Then I have a quick follow-up question about HBM. I believe last quarter you guided a compound annual growth rate (CAGR) of 40%, which implies that the market revenue this year is about $50 billion. Has that number changed? Considering the higher margins of DDR5 (D5) now, are you seeing industry participants more inclined to shift towards D5 rather than HBM? Thank you very much.

Sanjay Mehrotra, Chairman, President, and CEO: Currently, the margins for non-HBM products are indeed higher than those for HBM. Of course, the demand for HBM remains strong. We have not updated the data regarding the HBM TAM outlook that we provided last time. In the data center space, the demand for DDR5, LP (Low Power Memory), and HBM remains robust, and as the demand for AI in data centers continues to grow, we will continue to manage our product mix. As I mentioned earlier, in addition to data centers, we are also highly focused on ensuring we maintain relevant market share in other key segments. Therefore, in an environment where there is strong demand from data centers to edge AI, we are very focused on continuing to manage our product mix and see that Micron faces strong growth opportunities across its entire product portfolio in data centers. I would like to point out that this product portfolio includes HBM, LP, SOCAMM, DDR5, **as well as our data center SSDs. We have made significant progress in our market share in the data center SSD market over the past few years.**

CJ Muse: Thank you.

Operator: Your next question comes from Harlan Sur of JP Morgan. Your line is open, please go aheadHarlan Sur: Good afternoon, congratulations on achieving solid performance and strong quarterly execution. Sanjay, perhaps we can continue with your earlier comments on SSDs. **In the quarter last November, I estimated that your enterprise SSD (eSSD) business accounted for nearly half of your total flash business, growing approximately 60% quarter-over-quarter. From a margin perspective, this is clearly a very favorable product mix transition for the Micron team.** As you mentioned, you are still one of the top three eSSD suppliers globally. Based on this strong data, it seems that your eSSD business doubled quarter-over-quarter in the February quarter, maintaining a 50% share in the NAND mix. Looking ahead, with the continued ramp of the G9 node, your next-generation performance-optimized, capacity-optimized, and mainstream eSSD products will all be built on G9. Does this provide ample room for the team to continue driving quarter-over-quarter growth in the eSSD business for the remainder of this year and next year? I would also like to hear your thoughts on the new proposed storage tier of "High Bandwidth Flash (HBF)." Is this an area where the Micron team might begin to invest R&D resources?

Sanjay Mehrotra, Chairman, President, and CEO: Regarding your question about data center SSDs, this is certainly an area with strong growth potential. NAND supply is very tight, while demand for NAND remains robust. Data center SSDs are a major driver of NAND growth. Micron is in a very favorable position with our SSD product portfolio, which can meet the various capacity and performance requirements of different customers using TLC and QLC, and our data center portfolio is just that. So we have a significant advantage in this area. As part of our strategy to continuously shift our product portfolio and revenue structure toward high-profit pools and high-value segments in the industry, we will certainly continue to seize opportunities to expand our SSD business. We feel very good about the past trajectory and future plans for our data center SSD business.

As for your question about High Bandwidth Flash (HBF), of course, HBF has some positive attributes, such as capacity advantages, but it is also limited by the inherent constraints of NAND itself, such as write speed, power consumption, and data retention. Therefore, it may be a potential solution for certain specific workloads, but it is still in a very early stage. What is truly needed is engagement with customers in this area to really understand the commercial value proposition of HBF. We will certainly continue to research this.

Harlan Sur: I completely agree. Additionally, how much of these multi-year SCA agreements is due to the customization of next-generation HBM architectures for GPU and XPU chip customers (especially the customization of the base die), leading them to need earlier and longer-term engagement with you?Considering the 12 to 18 months design cycle for these customized base chips, as well as the IP sharing between you and your chip customers, and optimizing the base chips for your process flows, it indeed means that they must engage with you very early in the GPU and XPU design stages. Is this another factor driving the signing of these multi-year SCAs?

Sanjay Mehrotra, Chairman, President, and CEO: We still won't delve into specific details or involve specific customer types. But I can clearly tell you that, yes, these SCAs do bring us closer to our customers in terms of collaboration. This partnership naturally extends to a close integration of R&D collaboration and roadmap planning, including both our roadmap and the customers' roadmap. So this is definitely one of the benefits brought by these SCAs.

Harlan Sur: Yes, thank you.

Operator: Your next question comes from Tom O'Malley of Barclays. Your line is open, please go ahead.

Thomas O'Malley: Hello everyone, thank you for taking my question, the performance is indeed outstanding. There has been a lot of discussion in the industry this week at GTC and OCP regarding the LPU architecture and the increase in SRAM usage. How do you view the long-term memory market? Because you see more and more workloads starting to rely on types of memory other than HBM. Another slightly macro question is: since so much demand and these long-term agreements are related to data centers, and the number of customers who can truly acquire and build these products is very few, how do you conduct benchmark forecasting when increasing capacity? Do you have internal forecasting models for accelerators? Are you aggregating bottom-up forecasts with different customers? To ensure that in the 3rd, 4th, and 5th years, you provide sufficient supply for the industry without falling into the trap of overcapacity? Thank you very much.

Sanjay Mehrotra, Chairman, President, and CEO: First, regarding your question about SRAM and LPU architecture, I want to point out that this architecture makes AI infrastructure more efficient. Any architecture that makes AI infrastructure more efficient is good for the overall development of AI, as it allows this cake to grow faster. Please note that this LPU architecture works in conjunction with Vera Rubin, which itself uses a large amount of HBM and DRAM. Additionally, the LPU-based architecture that NVIDIA is collaborating on with Groq 3 LPX actually uses up to 12TB of DRAM per rack. Therefore, all of this is actually processing workloads in a more efficient manner. This helps optimize token economics, token speed, scale AI across inference stages, and helps reduce power consumption, with each beneficial improvement having a positive impact on further expanding and accelerating the deployment of AI demandIt is important to remember that the current proportion of AI deployment in enterprises is still very, very low. There are still huge opportunities in all verticals, all industries, and the entire economy in the future. Therefore, we are excited about the opportunities we face in meeting these future market demands with our full product portfolio in HBM, LP, DRAM, SOCAMM, and SSD. Ultimately, all of this proves how strategically significant memory is in the field of AI. Because without more and faster memory, AI cannot scale, and cannot deliver powerful capabilities in either training or inference. Just look at how the required DRAM capacity in advanced AI accelerators has doubled from last year to this year. And these factors are one of the reasons for the current supply shortage. Of course, the trend of deploying AI data also applies to edge devices, such as smartphones and PCs. Therefore, we are excited about future opportunities and absolutely look forward to the strong market opportunities for our full product portfolio in the future.

Operator: Your next question comes from Vivek Arya of Bank of America Securities. Please go ahead.

Vivek Arya: Thank you for taking my question. Sanjay, regarding HBM4, do you expect to reach your target range of 20% to 25% market share right from the start, or do you think it will be built up gradually over time? Conceptually, how do you view the pros and cons of expanding HBM share in the upcoming Vera Rubin generation?

Sanjay Mehrotra, Chairman, President, and CEO: As we shared before, in the third calendar quarter of last year (CQ3), we achieved our established goal of matching our HBM market share with our DRAM market share (which was our previously set target). We have also mentioned that looking ahead, we will manage HBM as part of our overall product portfolio and will no longer disclose market share on a quarterly basis. But what I can tell you is that we are very satisfied with our positioning of HBM products and the overall HBM product line. Of course, by 2026, both the HBM4 and HBM3E markets will be there. We will supply both products simultaneously and are confident in our overall positioning here and our ability to manage the business portfolio comprehensively.

Vivek Arya: I would like to follow up with a question, Mark, **I want to revisit the guidance of an 81% gross margin.** I understand you won't provide specific forward-looking forecasts, but looking back at when Micron's gross margin reached historical highs (I remember it was just over 60%), what are the differences between then and now? What insights do these historical precedents provide you in assessing the trajectory of gross margins over the next few quarters? When customers see such high levels of gross margin, which is a very, very important input cost in their AI chips, will they start to react differently? Thank youSanjay Mehrotra, Chairman, President, and CEO: Before Mark answers that question, please allow me to point out that I mistakenly said our goal is to reach the HBM share target in the third fiscal quarter of 2026. I misspoke. What I meant to say is that our original goal was to reach our HBM share by 2025, and we achieved parity between HBM share and DRAM share in the third quarter of fiscal year 2025. At that time, we also mentioned that after the third quarter of fiscal year 2025, we would not provide further details on HBM share. So I just wanted to correct my accidental slip of saying 2026 instead of 2025.

Vivek Arya: Okay.

Mark Murphy, Executive Vice President and Chief Financial Officer: Vivek, what I want to say is that the current industry is supply-constrained. And this tight situation will continue beyond 2026. So, this undoubtedly supports pricing in the near and medium term. We have also discussed how we are working with customers to allocate capacity for their businesses as much as we can, and we are closely collaborating with them on increasing capacity, ensuring supply, and new products. Regarding your question about reverting to historical averages, I think that may need to be re-evaluated. Our current situation is that AI is a transformative, long-term driver. As Sanjay mentioned, AI requires more and higher-performance memory, which helps reduce the cost of tokens and the energy consumption per token. It can increase the number of tokens and comprehensively enhance the intelligence level of AI, thereby addressing more complex problem sets and applications of agents, which in turn drives the generation of more tokens and requires more memory. Therefore, the current margins reflect that the market has recognized that memory has become more valuable and is an effective way to achieve AI commercialization, covering the entire process from data centers to the edge. On top of that, we clearly pointed out over a year ago that there are multiple aspects of supply constraints, and these constraints take time to resolve. Current inventory levels are very low. As nodes evolve and the trade ratio of HBM wafers increases, the bit output per wafer is declining. Any new capacity release actually requires new greenfield facilities, which is a physical limitation that takes a long time. Therefore, both of these factors are persistent: one is the increase in the value of memory, and the other is the structural challenges faced in increasing supply, and we are addressing both issues simultaneously. We are investing in capacity expansion while also increasing R&D spending to continue advancing technology and enhancing the value of memory. We believe these will help support margins in the long run, and I think customers recognize this, which is why they are signing these agreements.

Vivek Arya: Thank youOperator: The conference call ends here today. Thank you for your attendance, and you may now disconnect. The meeting is over.

The market has risks, and investments should be made with caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Invest at your own risk

### 相关股票

- [GraniteShares 2x Long NVDA Daily ETF (NVDL.US)](https://longbridge.com/zh-CN/quote/NVDL.US.md)
- [Invesco PHLX Semiconductor ETF (SOXQ.US)](https://longbridge.com/zh-CN/quote/SOXQ.US.md)
- [Direxion Semicon Bull 3X (SOXL.US)](https://longbridge.com/zh-CN/quote/SOXL.US.md)
- [VanEck Semiconductor ETF (SMH.US)](https://longbridge.com/zh-CN/quote/SMH.US.md)
- [NVIDIA (NVDA.US)](https://longbridge.com/zh-CN/quote/NVDA.US.md)
- [GraniteShares 2x Long MU Daily ETF (MULL.US)](https://longbridge.com/zh-CN/quote/MULL.US.md)
- [T-Rex 2X Long NVIDIA Daily Target ETF (NVDX.US)](https://longbridge.com/zh-CN/quote/NVDX.US.md)
- [First Trust Nasdaq Food & Semicon (FTXL.US)](https://longbridge.com/zh-CN/quote/FTXL.US.md)
- [Invesco Semiconductors ETF (PSI.US)](https://longbridge.com/zh-CN/quote/PSI.US.md)
- [YieldMax NVDA Option Income Strategy ETF (NVDY.US)](https://longbridge.com/zh-CN/quote/NVDY.US.md)
- [Direxion Daily NVDA Bull 2X Shares (NVDU.US)](https://longbridge.com/zh-CN/quote/NVDU.US.md)
- [SPDR S&P Semicon (XSD.US)](https://longbridge.com/zh-CN/quote/XSD.US.md)
- [iShares Semiconductor ETF (SOXX.US)](https://longbridge.com/zh-CN/quote/SOXX.US.md)
- [Micron Tech (MU.US)](https://longbridge.com/zh-CN/quote/MU.US.md)
- [Direxion Daily MU Bull 2X Shares (MUU.US)](https://longbridge.com/zh-CN/quote/MUU.US.md)

## 相关资讯与研究

- [Planet to build world’s first GPU-native AI engine for planetary intelligence with Nvidia](https://longbridge.com/zh-CN/news/279330399.md)
- [LPU: Nvidia popularizes a new AI acronym](https://longbridge.com/zh-CN/news/279662346.md)
- [Micron Options Signal Bullish Bias Ahead of Earnings](https://longbridge.com/zh-CN/news/279643198.md)
- [Micron Technology (MU) Is Up 13.5% After New AI-Focused DRAM and HBM Partnership With Applied Materials](https://longbridge.com/zh-CN/news/279414086.md)
- [EXL advances EXLerate.ai agentic AI platform to support enterprise-scale adoption with NVIDIA technologies | EXLS Stock News](https://longbridge.com/zh-CN/news/279335695.md)