Micron Conference Call: 75% of 2027 Capacity Already Sold, No Supply-Demand Inflection Point in Sight, Physical AI Will Be the Next Breakout Point

Micron Technology closed its fiscal year 2026 with a record-breaking quarterly financial report and offered guidance that far exceeded market expectations, suggesting a longer-lasting momentum for the AI-driven memory supercycle than previously estimated.

As Wallstreetcn reported, after the close of trading on September 30th, Eastern Time, Micron Technology released its performance report for the fourth quarter of the fiscal year 2026, which ended on September 3. According to the data, Micron’s Q4 single-quarter revenue hit a record $54.2 billion, up 31% quarter-on-quarter and surged 379% year-on-year; gross margin for the quarter soared to 87%.

For the entire fiscal year 2026, Micron’s total revenue reached $133.2 billion, 3.5 times the record level of the previous fiscal year, with data center revenue quadrupling.

During the earnings call, Micron Chairman and CEO Sanjay Mehrotra defined the ongoing industry transformation as:

Superintelligence is creating the most compelling opportunity in Micron’s history.

Mehrotra stated that Micron has already signed 26 long-term agreements securing about $150 billion in long-term orders. He emphasized that the supply and demand for memory in 2027 and 2028 will be even tighter than in 2026, and “there’s no end in sight to the supply-demand imbalance.”

Chief Financial Officer Mark Murphy said Q1 revenue for fiscal 2027 will reach $61.5 billion, with EPS at $38.15, showing continued quarter-on-quarter growth. He clearly pointed out that this quarter will be the low point for gross margin in FY27, with margins expected to gradually recover in subsequent quarters.

Furthermore, Micron outlined a new story for “Physical AI” to the market. Mehrotra pointed out:

Autonomous vehicles are the first significant deployment of Physical AI... L4 and above autonomous vehicles typically require more than 200GB of memory and several terabytes of storage, an order of magnitude greater than today’s L2+/L3 models.

He emphasized that humanoid robots are expected to have similar demand. By the end of this century, Physical AI will become a key driver of storage demand.

Supported by strong cash flow, Micron announced first-half capital expenditure of $25 billion for fiscal 2027 and promised to return 100% of excess cash to shareholders in the future.

“We do not see the inflection point when supply and demand will rebalance”

Behind the consistently better-than-expected performance lies an unprecedented structural shortage in the memory industry. The market is highly attentive to whether this high growth is sustainable, and Micron has given a strong and confident answer.

Mehrotra made it clear during the call that since the last earnings report, industry demand has further strengthened. He stated:

We expect the supply-demand situation for memory and storage in fiscal years 2027 and 2028 to be tighter than in 2026.

He attributed this trend to the evolution of AI architectures:

AI is becoming superintelligent, and storage enhances this intelligence and the competitiveness of client platforms. Whether running open or closed source models, AI applications running on platforms with stronger memory capabilities can achieve more scalable growth.

Regarding the supply-side bottlenecks, Mehrotra offered the most striking insight of the meeting:

The structural gap between supply and demand growth rates is causing ongoing supply tightness... Even considering industry plans for new DRAM cleanroom space, we don’t see an inflection point when supply and demand will rebalance under strong demand—including new incremental requests from clients.

In the much-watched HBM (High Bandwidth Memory) field, Micron said industry HBM demand growth would continue to outpace traditional DRAM through 2028.

Mehrotra revealed that most of Micron’s HBM supply for 2027 has already been contracted, “with prices rising sharply year-on-year, narrowing the profit margin gap between (HBM) and traditional DRAM.”

Locked in $150 Billion of “Sky-High” Long-Term Orders, Over 75% of 2027 Capacity Already Sold

To address extreme supply shortages, downstream customers are spending heavily to secure Micron’s long-term capacity, providing a solid foundation for Micron’s future predictable results.

Micron CFO Mark Murphy disclosed a set of stunning order figures during the call:

To date, we have signed a total of 26 Strategic Customer Agreements (SCA), with our Remaining Performance Obligations (RPO) around $150 billion.

SCA is a long-term, “take-or-pay” type contract. Mehrotra further added:

Within these 26 SCAs and extension agreements, customer financial commitments have increased to $32 billion, most of which are cash deposits. Customers are seeking supply assurance past 2030, and we have now signed SCAs extending to 2031.

This long-term contract model has fundamentally changed Micron’s business visibility. According to management, these agreements are expected to account for over 35% (possibly up to 50%) of Micron’s total revenue through 2030.

Currently, over 75% of Micron’s output for 2027 is already committed to clients, and current negotiations with clients are largely focused on 2028.

Shareholder Return Commitments and Surging Capital Expenditure

With extremely strong fundamentals, Micron’s cash flow and capital expenditure planning have become another major focus for investors.

In Q4, Micron generated an astonishing $44 billion in operating cash flow and $33.2 billion in free cash flow.

Facing this “mountain of cash,” Murphy offered a clear shareholder return timeline:

Over time, we expect to return 100% of excess cash to shareholders. We plan to increase capital return starting December 9, 2026—the two-year anniversary of signing our final CHIPS Act agreement—primarily through share repurchases.

Meanwhile, to ease capacity hunger for 2028 and beyond, Micron is being forced to sharply boost capital spending.

The company expects CapEx for the first half of fiscal 2027 to be about $25 billion, with the second half higher, and most of the increase devoted to new wafer fab construction (buildings) rather than purely equipment procurement.

The CFO explained that this is all because long-term agreements provide enough demand visibility, and greenfield plants require very long lead times.

The Next Decade: Physical AI and Humanoid Robots

Beyond today’s red-hot data center market, Micron outlined the next significant incremental market: Physical AI.

Mehrotra believes:

Autonomous vehicles are the first major deployment of Physical AI, and we believe over time it will extend to humanoid robots and other intelligent autonomous systems.

He offered specific data forecasts:

L4 and higher-level autonomous vehicles typically require more than 200GB of memory and several TB of storage, an order of magnitude greater than today’s L2+/L3 semi-autonomous vehicles. Humanoid robots are expected to have similar storage demand.

Micron believes that with both the number of devices and per-device storage content increasing, by the end of the century, Physical AI will become an extremely important driver of storage demand. As Mehrotra said when wrapping up the meeting:

Superintelligence is creating the most compelling opportunity in Micron’s history.

Full Transcript of Micron Technology's Q4 Earnings Call (AI-Assisted Translation):

Host:

Hello everyone, thank you for attending and welcome to Micron Technology’s Q4 Fiscal 2026 Earnings Conference Call. After today’s remarks, we will open the floor to questions.

I’ll now hand over to our Corporate Vice President, Head of Investor Relations and Finance, Satya Kumar. Satya, please begin.

Satya Kumar (Corporate VP, IR & Finance):

Thank you all for joining Micron Technology’s Q4 2026 earnings call. Joining me today are Chairman and CEO Sanjay Mehrotra, and Chief Financial Officer Mark Murphy.

This meeting is being webcast live on our IR website at investors.micron.com with audio and slides. In addition, this quarter’s earnings press release was issued on our website and the remarks for this call have also been uploaded.

This discussion contains forward-looking statements involving risks and uncertainties, including statements concerning the company’s future financial and operating performance, business models, and trends and expectations regarding our business, customers, markets, industries, products, regulations, and more. All statements are based on our current assumptions, and we do not undertake to update these statements.

Please refer to our most recent 10-K, 10-Q and other filings with the Securities and Exchange Commission for information on factors that may cause actual results to differ materially from expectations.

Today’s financial discussion is based on non-GAAP measures unless otherwise noted. A reconciliation between GAAP and non-GAAP is available on the company website.

Now, I’ll turn the call over to Sanjay.

Sanjay Mehrotra (Chairman, President and CEO):

Thank you, Satya. Micron delivered outstanding results in Q4 FY2026, setting new records in revenue, gross margin, and EPS—all above the top end of our guidance.

FY2026 was a remarkable year:

Revenue was 3.5 times last year’s record, data center revenue quadrupled;
Micron DRAM revenue surpassed $100 billion in FY2026.
I am grateful for the extraordinary efforts of our global team in achieving these results. To recognize these contributions and highlight Micron’s strong execution, we increased FY2026 performance bonuses for all employees, reinforcing a performance-driven culture and linking employee interests with long-term shareholder value creation.

While FY2026 performance was exceptional, we anticipate an even better FY2027.

Industry Demand Remains Strong

Since the last earnings call, industry demand has strengthened further. We expect supply and demand for memory and storage to be even tighter in FY2027 and 2028.

AI-Driven Memory Advantages

AI is evolving toward superintelligence, and memory is key to boosting that intelligence and enhancing the competitiveness of customer platforms. Whether based on open- or closed-source models, AI across end markets runs on a variety of competing client platforms, all sharing one important feature: their value proposition is enhanced by memory and storage performance and capacity. Running AI applications on platforms with stronger memory allows for greater scalable growth, improved user experiences, and more value capture from AI.

Micron’s Differentiated Opportunity

The strategic importance of memory for customers provides Micron with unprecedented differentiation opportunities. As we address more complex, AI-driven memory demands, we can stand out in differentiated performance and quality, go-to-market timing, and geographic diversification including U.S.-based DRAM manufacturing. Compared to competitors, we also have a richer product portfolio focusing on higher-value solutions.

Micron’s technology leadership, robust portfolio, strategic customer agreements (SCA), and exceptional manufacturing enable us to capture these opportunities.

Technology Leadership

Micron leads the industry in technology. Our 1-Gamma DRAM node and G9 NAND node are the industry’s largest-volume production nodes and are expected to be the highest-shipment technology nodes in our history.

Development of next-gen DRAM and NAND nodes is on track, with mass production expected to begin in the second half of 2027. Leveraging these nodes and advanced packaging, we’re delivering industry-leading products from HBM to large-scale SOCAMM, high-capacity high-performance DDR modules, and data center SSDs.

Global Capacity Expansion

We’re expanding global manufacturing to meet client demand growth through this decade and beyond. Micron continues to invest in wafer fabs in Virginia, Idaho, and New York, providing long-term supply security for DDR, LPDRAM, and HBM products for customers in data center, PC, mobile, automotive, aerospace and defense, medical, humanoid robotics and other industrial and consumer markets.

Last quarter, we celebrated a key milestone in New York’s first wafer fab, with initial output expected in 2030. ID1’s wafer fab should begin output mid-2027; ID2 by the end of 2028. In Q4 FY2026, we broke ground on a DRAM fab expansion in Japan, targeting initial output by end-2028 to support node transitions. In Taiwan, the Tongluo site is expected to see meaningful product shipments by mid-2027. In Singapore, HBM advanced packaging cleanroom buildout is ahead of schedule, targeting first output in early 2027; the new Singapore NAND fab is on schedule for output in the second half of 2028.

Volume ramp of new DRAM/NAND fabs takes time, with significant contributions occurring several quarters after initial output.

Strategic Customer Agreements (SCA)

Our SCAs are accelerating Micron’s business transformation. These multi-year take-or-pay deals let us plan long-term supply more clearly while strengthening durability and predictability of robust financial performance. These agreements also offer supply security for customers and deepen collaboration on technology roadmaps, boosting their ability to invest confidently and deliver value to end-users.

To date, we have signed 26 SCAs, expected to cover over 35% of our revenue through 2030. Three-quarters of projected SCA revenue features clear pricing frameworks, mostly with price floors and ceilings; the rest is market-based with periodic adjustments.

Customers want SCA supply beyond 2030; we have signed SCAs extending to 2031, with one-year renewals on two agreements. All new SCA pricing is negotiated at higher current-market levels.

To date, financial commitments on the 26 signed SCAs and extensions have increased to $32 billion, mostly cash deposits—demonstrating customer confidence in long-term memory demand.

End Market Highlights

See our earnings release for detailed product highlights covering high-capacity DDR and LP server DRAM, data center SSDs, PC, smartphone, and Physical AI products.

We expect server shipments in calendar 2026 and 2027 to each grow by high-teens percentages. This robust shipment growth in a tight supply environment is supported by slightly lower content growth rates than previously forecast.

Model parameter growth, longer context windows, and higher concurrency keep driving increases in memory content needed for efficient AI workloads. We’re leveraging our technology leadership and manufacturing strength to deliver across the storage hierarchy for data center customers.

For HBM, as we expand shipments to more customers, Q4 HBM revenue growth outpaced overall company revenue growth. Most 2027 calendar year HBM supply is already under contract, with pricing up significantly, narrowing HBM’s margin gap versus legacy DRAM. Our HBM4 ramp continues well, and our forward HBM roadmap is strong. We’re proud to collaborate with NVIDIA on the industry’s first custom HBM4e solution, NVHBM, for next-gen GPUs and NVLink Fusion platforms.

In NAND, AI context storage for KV Cache offload and HDD replacement are expanding the SSD addressable market. In Q4, data center SSD revenue approached $10 billion—over 10x year-over-year—accounting for more than two-thirds of total NAND revenue. We expect to set a data center SSD market share record for the fifth year in a row in calendar 2026. This success is due to our NAND technology leadership, data center product range, and deep customer design wins for critical deployments.

For the PC and mobile markets, even though overall shipments may decline in double digits, strong high-end demand means industry revenue growth this year. OEMs continue to introduce new AI features in flagship PCs and smartphones, driving demand for high-performance devices with greater DRAM and NAND content. Micron is focused on this high end, and with our leading product portfolio, we’re well positioned as customers expand edge AI. In Q4, nearly half of MCBU revenue came from 1-Gamma products, reflecting accelerated customer qualification and adoption of our latest lower-power, higher-performance technology.

For Physical AI, autonomous vehicles are the first major large-scale deployment. We believe this trend will extend to humanoid robots and other intelligent systems requiring more powerful, energy-efficient memory and storage to operate in real time. L4 and above vehicles generally need over 200GB of memory and several terabytes of storage—more than an order of magnitude above current L2+/L3 semi-autonomous cars. Humanoid robots are expected to require similar storage.

With anticipated increases in volume and content, Physical AI could become a major memory and storage demand driver by the end of this decade. Multiple customers are sampling our next-gen products, and we continue investing in our roadmap to capitalize on this opportunity.

Market Outlook

We expect memory supply-demand in calendar years 2027 and 2028 to be even tighter than in 2026.

For NAND: In calendar 2026, we forecast industry bit shipments to grow by low 20%-plus, above prior expectations; Micron’s NAND supply growth will lag the industry. In 2027 and 2028, we project industry NAND bit shipment growth in the mid-20% range, with tight supply throughout both years.

For DRAM: In calendar 2026, we expect 20%-plus bit shipment growth, with Micron growing roughly in line with the industry. For 2027 and 2028, we forecast industry DRAM bit growth in the low 20s percentage, with tight supply in both years.

We expect HBM bit growth to outpace legacy DRAM through 2028.

The structural DRAM supply-demand gap is causing persistent tightness, requiring more cleanroom space to close the shortfall. Even with planned new DRAM cleanrooms and strong incremental client requests, we still see no clear path to supply-demand rebalance.

CapEx Increased

Given the urgent need for DRAM cleanroom space and SCA-driven strong late-decade demand visibility, we plan to increase CapEx in FY2027, mainly for construction to accelerate post-2028 cleanroom readiness. We’re also optimizing existing capacity, bringing forward some equipment purchases. Mark will share more CapEx details.

As we invest in cleanrooms, we maintain strict capital discipline and will adjust equipment capacity expansion prudently as market conditions dictate.

Management Appointments

To further increase execution and innovation, last month we announced management changes: Manish Bhatia became President & COO, and Scott DeBoer became President & Chief Technical and Product Officer.

Manish leads worldwide business units and operations with P&L responsibility, and as COO oversees needs-to-supply execution, faster and integrated decision making, and organization synergies to meet evolving customer needs.

Scott leads innovation, technology, and product. As Chief Technical & Product Officer, he advances Micron’s memory leadership and innovation to meet rapidly evolving customer needs, and oversees Micron Research—a global flagship center focused on memory/computing breakthroughs.

I’ll now turn it over to Mark for a financial and outlook summary.

Mark Murphy (EVP & CFO):

Thank you, Sanjay, and good afternoon everyone.

Micron closed the year with excellent results—Q4 FY2026 revenue, gross margin, and EPS all above guidance range highs.

Annual Performance Review

Full-year record revenue of $133.2 billion, up 256% YoY. FY2026 gross margin expanded to 81.1%, up 40 points YoY; EPS up 811% YoY to $75.52.

As noted, we’ve signed 26 SCAs with RPOs around $150 billion. All SCAs include take-or-pay minimums; RPOs reflect only SCA value with fixed or bounded pricing, calculated at minimums—making them inherently conservative. As stated last call, even at price floors, we expect gross margin well above prior cycle peaks. Actual SCA-period revenue should far exceed stated RPOs.

Q4 Results

Q4 FY2026 total revenue was $54.2 billion, up 31% QoQ and 379% YoY—the sixth straight record.

DRAM: Record $39.8 billion, up 343% YoY, 73% total, up 27% QoQ; bit shipments up mid-single-digits %, ASP up high-teens %, benefiting from tight DRAM supply.
NAND: Record $14.1 billion, up 526% YoY, 26% total, up 42% QoQ; bit shipments up ~10%, prices up ~30%, in a tight market.
Gross margin 87%, up 210bps QoQ, driven by higher prices and strong execution, partly offset by product mix.

Business Segment Quarterly Results

Cloud Storage (CMBU): Record $16.3 billion, 30% of total, up 18% QoQ on higher prices and shipments; gross margin 83%, flat QoQ as higher HBM mix offset price gains.

Core Data Center (CDBU): Record $18 billion, 33% total, up 56% QoQ on higher prices and shipments; gross margin 90%, up 290bps QoQ on pricing/product mix.

Mobile & Client (MCBU): Record $13.1 billion, 24% total, up 14% QoQ on higher prices offset by lower shipments; margin 90%, up 260bps QoQ driven by price/mix.

Auto & Embedded (AEBU): Record $6.8 billion, 13% total, up 47% QoQ on prices and shipments; margin 84%, up 470bps, price-driven.

OpEx and Profitability

Q4 OpEx was $2.6 billion, up $1.1 billion (mainly higher performance bonuses for all staff and $300m community initiatives).

Op income $44.6 billion, op margin 82.3%, up 110bps QoQ and 47pts YoY.

Tax expense $6.8 billion, actual rate 15%. Non-GAAP diluted EPS $33.42, up 33% QoQ.

Cash Flow and CapEx

Q4 operating cash flow $44 billion, CapEx $10.8 billion, free cash flow $33.2 billion.

Note: SCA customer deposits are in financing, not free cash flow. Q4 SCA cash deposits $12.3 billion.

Q4-end inventory $10.4 billion, DIO 129 days, up 9 days QoQ. The increase partly reflects accrued manufacturing incentives counted in Q4 inventory; no EOL stocking included. Inventory and supply remain extremely tight, DIO is expected to decrease over coming quarters.

Ending cash and investments record $73.5 billion. Customer deposits on balance sheet: $12.7 billion. SCA deposits are unrestricted and returned at contract-end if minimums are met.

Q4, we repaid ~$500 million debt (including $300 million notes). Weighted avg debt maturity ~9 years. Q4-end debt: $5.2 billion, net cash: $68.3 billion.

Two credit agencies upgraded ratings this quarter; all main agencies now at BBB+ or equivalent. Our balance sheet has never been stronger and will stay so even as we invest more in technology and capacity.

As noted, from December 9, 2026 (the two-year anniversary of our CHIPS act agreement), we will accelerate capital returns. Over the long run, we aim to return all excess cash to shareholders.

Outlook

We forecast a record-high Q1 FY2027 revenue of $61.5 billion (±$1.5 billion), gross margin 86.25%, OpEx about $2.06 billion. Based on roughly 1.5 billion shares, EPS $38.15 (±$1).

We expect FY2027 will be another record year with sequential quarterly growth. As with strong prior guidance and record results, we anticipate FY2027 incentive pay to be higher than FY2026.

Q1 is the low point for FY27 gross margin, with steady increases in later quarters as price gains moderate.

As Sanjay noted, most of the higher FY2026 bonus accruals (mainly manufacturing) hit inventory in Q4 and will mainly flow through gross margin in Q1. Q2 will benefit as that cost falls away, but the benefit is offset by higher FY2027 incentive accruals.

We expect FY2027 OpEx to increase by ~$2.5 billion, mainly from larger R&D and compensation to pursue unprecedented memory opportunities.

We expect Q1 and FY2027 tax rate of about 15.5%.

CapEx Plan

Micron will maintain strict capital discipline globally to meet demand. CapEx numbers exclude expected government incentives.

We expect Q1 CapEx at ~$11.5 billion; FY27 H1 at ~$25 billion; H2 will be higher, and construction CapEx will grow faster than equipment CapEx.

Finally, I want to thank all Micron employees worldwide for their focus on technology, product innovation, and outstanding execution that has enabled today’s strong results and outlook.

I now return the call to Sanjay to wrap up.

Sanjay Mehrotra (Chairman, President and CEO):

Thank you, Mark. Superintelligence is creating the most compelling opportunity in Micron’s history. FY2026 was exceptional, and we expect FY2027 to be even better.

As Micron reaches its 48th anniversary, I want to pay tribute to the generations of employees whose innovation, unwavering execution, and resilience laid the foundations for this pivotal moment in our history.

We now begin the Q&A session.

Q&A Session

Host: Now we’ll start the Q&A session. The first question comes from Timothy Arcuri at UBS Securities. Please go ahead.

Timothy Arcuri (UBS Securities):

Mark, I’d like to ask about capital returns. I know you don’t want to give away too much too soon, but can you provide some benchmarks? How do you view minimum cash balances? You have about twice the cash of Apple or NVIDIA, so what’s enough? Are you planning to keep $100 billion and return the rest? Any framework you can share?

Sanjay Mehrotra:

Tim, happy to share. The current market, combined with Micron’s technology, product, and operational execution, has delivered very strong free cash flow—$33 billion in Q4. We expect these trends to continue even as we ramp up R&D and CapEx, given demand drivers, supply-side constraints, and our long-term contract base.

In the near term, combining Q1 guidance and CapEx, free cash flow in Q1 will be well above Q4’s $33 billion.

On target cash, we aim to reach our desired balance around the end of Q1. Once there, we will return the surplus mainly via stock buybacks. We’ve announced capital return acceleration from December 9. As a final note, buyback authorization is $2.2 billion—we will seek higher authorization soon.

Timothy Arcuri:

Thanks. One more—on CapEx, you haven’t issued a full-year number, but it could exceed $55 billion for the year, about high teens as a percent of FY27 revenue. With a larger revenue base, CapEx will catch up over time, but what’s the long-term capital intensity framework? Previously you mentioned mid-30s percent, which seems high; is 20-25% the new normal?

Sanjay Mehrotra:

Tim, you’re right—we guided for H1 CapEx and expect H2 to be even higher. Importantly, CapEx is shifting toward construction rather than equipment, a trend set to continue for several years.

On capital intensity, current capex intensity is historically low, reflecting storage as a strategic asset. The industry has structurally reset. We’ll continue to expand capacity with strong discipline, ensuring proper returns on each addition.

CJ Muse (Cantor Fitzgerald):

First, on gross margin. For the November quarter, can you quantify the impact of higher-cost inventory? Any other structural factors to watch?

Mark Murphy:

CJ, here’s some context on Q4 and Q1 margin drivers.

In Q4, we raised incentive compensation, mainly reflected in OpEx when comparing Q3 and Q4. On the cost side, most of that accrual flowed into Q4 inventory, so Q4 gross margin impact was minimal.

The higher-cost inventory—from higher incentive comp—will run through COGS in Q1, adding to previously discussed start-up and other costs. Incentive comp is the main drag on Q1 margin. In total, about a $1 billion cost hit to Q1, which gives you the gross margin impact magnitude.

Note, we also raised FY2027 incentive comp. Q1’s OpEx guide reflects this, and manufacturing costs will appear in gross margin from Q2 onward. In FY27, other than higher output and depreciation, incentive comp and start-up and other costs will add about $1 billion of persistent pressure. While Q1 faces a one-off headwind, company profitability has structurally changed—we are operating on a much stronger foundation and expect FY27 and FY28 to be better than FY26, with long-term agreements giving revenue visibility beyond contract durations.

Some costs are variable/episodic. Q1 is the margin trough for the year; as prices rise and execution remains strong, gross margin will increase in following quarters.

CJ Muse:

Very helpful. On HBM—you mentioned prices converging with standard DRAM, does the new price kick in from Jan 1? What’s the aggregate HBM growth outlook?

Sanjay Mehrotra:

2026 HBM prices were set in last year’s negotiations. As noted, most 2027 HBM supply is contracted at sharply higher prices, helping narrow HBM/non-HBM margin differentials.

Overall, HBM business is strong. Industry-wide, HBM demand growth should outpace overall DRAM growth. Our HBM3E, HBM4 and coming HBM4E are well-positioned. HBM’s robust outlook, premium portfolio, and higher 2027 prices will keep shrinking the HBM/DRAM margin gap.

Vivek Arya (Bank of America Securities):

Mark, back to cash returns. At this pace, from Q2 to Q4, you could generate over $100 billion in cash. Even if Q1 is for reaching your balance goal, thereafter you could support at least $100 billion in cash returns. Is this a reasonable view or am I missing anything?

Mark Murphy:

Vivek, I can only reiterate we have both ability and intent to increase returns. Watch for an acceleration after Dec 9 once new buyback authorization is in place.

Vivek Arya:

A second question for Sanjay. With memory stocks’ current valuation, the market might believe next year is the cycle’s peak for pricing/earnings, due to new supply or data center clients lowering specs due to shortages or costs. Without talking specific pricing, from a conceptual standpoint, with incremental supply and possible customer down-spec, how likely is it that industry pricing remains firm or even rises in 2028?

Sanjay Mehrotra:

As stated, for calendar 2027 and 2028, we see demand exceeding supply, with supply even tighter than 2026.

Even as we expand, as noted, even when new cleanrooms come online in 2028, supply remains tight: 1) Construction takes a long time; 2) ramping output post first wafers is gradual; 3) the shift from HBM3E to HBM4/HBM4E involves higher conversion ratios, slowing supply growth; 4) future node transitions yield smaller wafer productivity gains. All these create supply-side constraints, ensuring ongoing tightness even as new cleanrooms come online.

Meanwhile, demand stays robust. Even if some server platforms slow content growth, server shipments are rising: we forecast high-teens percentage shipment growth for 2026/2027, supporting data center DRAM growth in 2028.

Customer down-spec is about shipping more units and capturing scaled AI opportunities. It doesn’t alter the need for more—not less—memory, and optimization leeway decreases over time.

With larger AI models, bigger context windows, higher concurrency, and mass deployment of enterprise/consumer AI agents, demand—for both capacity and performance—will continue to go up. We’re very positive on supply-demand and pricing prospects.

Lastly, our SCAs provide very strong demand visibility. SCA customers want more supply and non-SCA customers are already placing 2027 orders, with most current negotiations focused on 2028. Demand visibility extends beyond 2030, and some clients have signed or renewed for 2031. We see no clear signs when the market will rebalance.

Krish Sankar (TD Cowen):

Sanjay, there’s talk a large customer is downscaling HBM specs; with HBM’s higher conversion ratio, if that capacity shifts to DDR, does this mean DDR supply will spike?

Sanjay Mehrotra:

In fact, we expect overall HBM demand in 2027 and 2028 to outpace total DRAM, keeping supply tight for those two years.

AI platforms need ever more, faster memory for expanding context, concurrency, and model sizes. The underlying need remains robust. Customer optimization is about balancing shipment growth and AI market opportunities via resource allocation—it doesn’t change the underlying need. Further optimization space declines over time. The value proposition and growth story for memory in AI is much more enduring than short-term tweaks. Overall, we’re more positive on 2027/2028 than 2026—and HBM demand growth will outpace legacy DRAM through 2028.

Krish Sankar:

Mark, a quick follow-up. FY2027 CapEx is clearly above $50 billion and FCF will go over $100 billion. With construction spending rising faster than equipment, do you see any equipment procurement constraints? Or is that not a bottleneck?

Mark Murphy:

Krish, this is mainly due to strong supply-demand imbalances and the cycle time for new fab capacity. We’ve detailed multiple new and planned fabs—ID1, ID2, Japan, Singapore, Tongluo, etc. SCA-fueled demand visibility gives us the confidence to move these forward, and equipment will be deployed at the appropriate times.

Harlan Sur (JP Morgan):

Earlier, with 16 SCAs, you said future revenue coverage could exceed 50%. Now you have 26, and the market is even tighter. If all SCAs in negotiation materialize, how much future revenue coverage could SCAs reach? You mentioned over 75% of 2027 output is already committed. Does this mean SCA revenue coverage could reach 60-70% in coming years for both SCA and non-SCA customers?

Sanjay Mehrotra:

First, to clarify—my “over 75% committed” remark for 2027 output covers both SCA and non-SCA customers, not just SCA. We work annually with some top non-SCA clients, and are already getting 2027 commit orders from non-SCA accounts.

On SCA revenue coverage, our previous guidance of around 50% coverage by 2030 remains valid, though the ratio could vary with overall sales base and may end up slightly below 50%.

We strive for structural balance—flexibility across accounts, end markets, and new entrants—a must in a fast-changing, innovative space like ours.

The core value of SCAs is providing long-term demand visibility and enabling earlier, data-backed capacity planning—this is a fundamental change for industry health. Unlike before, we can now manage supply-demand and investment planning on longer timeframes, which will reduce some cyclical volatility.

Harlan Sur:

On NAND, though Micron is fourth/fifth globally for NAND bit supply, the team is near second place for data center/enterprise SSDs. Is this just G9 node leadership, or also controller/firmware/customization, and involvement in NVIDIA's SCADA project? As KV Cache offload to storage becomes common, and flash-based architectures emerge, will this change your overall NAND R&D and CapEx stance?

Sanjay Mehrotra:

Across HBM-DRAM-SSD tiers, as context windows grow and AI evolves, we’re very pleased with our product positioning.

Our data center SSD lineup is our strongest ever for the huge AI-driven opportunity. As reported, DC SSD market share has risen for five years—Q4 DC SSD revenue nearly $10 billion, over two thirds of NAND sales, vindicating our team’s execution.

We are continuing to invest heavily in R&D and manufacturing, with NAND CapEx a significant part of the plan. We’re building a new NAND fab in Singapore, scheduled for 2028. CapEx today is going to support G9 node transition, equipment optimization, NAND R&D, and the rest of Singapore operations.

Host: That concludes the Q&A and today’s meeting. Thank you for participating. Please disconnect now. The meeting is adjourned.