Micron (MU.US) Q4 Earnings Call: Management Declares "No Point of Supply-Demand Balance in Sight," 75% of Next Year's Shipments Already Locked In, 2028 Tighter Than 2027

Title context: Micron (MU.US) Q4 Earnings Call: Management Declares "No Point of Supply-Demand Balance in Sight," 75% of Next Year's Shipments Already Locked In, 2028 Tighter Than 2027 Text: After U.S. market close on September 30, Eastern Time, Micron Technology, Inc. (MU.US) released its financial results for the fourth quarter of fiscal year 2026. On the earnings call, management sent optimistic signals, saying that AI-driven memory demand remains strong, supply and demand will remain tight in 2027 and 2028, and the company's confidence in long-term growth has strengthened. Management said that more than 75% of fiscal 2027 shipments are already locked in, SCAs cover about 35% of sales through 2030; HBM is growing faster than conventional DRAM, and 2027 pricing has already been significantly raised. On the supply side, constraints include cleanroom construction, diminishing returns from technology transitions, and a rising HBM trade ratio, and the company does not yet see a point when supply and demand will balance. Management's core view is this: AI-driven memory demand is far from reaching supply-demand balance, the market will remain tight in 2027 and 2028, and 2028 may even be tighter than previously expected. President and Chief Operating Officer Manish Bhatia said server units will continue to grow into 2027, and the rapid rise of agentic AI is also creating CPU-driven memory demand; the company has already locked in more than 75% of fiscal 2027 shipments and has extended customer allocation negotiations into 2028. The supply side is constrained by structural factors such as diminishing returns from technology transitions, HBM growing faster than conventional DRAM, a rising HBM capacity consumption ratio (trade ratio), and long cleanroom construction cycles, so the company "cannot see when supply and demand will balance." HBM remains the focus. Management said HBM shipment growth will outpace conventional DRAM, and its share of industry capacity will continue to rise through 2028; calendar 2027 HBM pricing has already been significantly raised and will reset at the beginning of the year to narrow the profitability gap with conventional DRAM. On HBM market share, the company is not pursuing a single number, but rather roughly maintaining a share comparable to its overall DRAM share. On HBM4E, Scott DeBoer said the company has been co-designing NV HBM with NVIDIA Corporation for more than a year, and this will be the first important custom HBM product, with differentiation in power consumption, speed, and product quality, and is expected to bring high value and strong ROI. On Strategic Customer Agreements (SCAs), the company has signed 26 SCAs covering about 35% of sales through 2030, spanning DRAM and NAND; DRAM volume is slightly below 35%, while NAND bits are slightly higher. The 10 newly signed customers range from small to large, with coverage across all business units. About three-quarters of SCA revenue has a clear pricing framework, while one-quarter is open to periodic negotiation or market-based pricing, with most having floors and ceilings, but new agreements have been repriced according to current market conditions and expectations of future tightness. On capital expenditure, CFO Mark Murphy said the increase in fiscal 2027 construction capital expenditure is mainly directed toward cleanrooms coming online at the end of 2028 and beyond, in order to accelerate capacity space; however, the spending will not immediately convert into bits, and the company will flexibly equip capacity based on demand trends and equipment supply agreements. On China, management said China exposure continues to decline and is expected to fall to the single-digit range in fiscal 2027; the company is at least 2 nodes ahead of Chinese competitors technologically, 1-gamma DRAM already accounts for most bits, 1-delta will enter mass production in the second half of next year, and EUV is a key differentiator. For mobile and client, management acknowledged that bit shipments declined quarter over quarter, but revenue grew due to pricing and mix improvement, with strong demand in high-end client and flagship smartphones. On NAND, the industry is expected to grow at around the mid-20% range in 2027-2028, with the market remaining tight; Singapore cleanroom space is being used for advanced R&D and HBM pilot production, causing this year's supply growth to be below the industry, but the G9 ramp will provide high-ROI supply. Overall, management believes AI hardware demand, SCA lock-ins, and supply constraints will support strong long-term financial performance, although startup costs and slower price increases may partially offset margin expansion. The following is the summary of Micron Technology, Inc.'s fourth quarter fiscal year 2026 earnings call: Satya Kumar Corporate Vice President, Investor Relations and Treasurer Welcome to Micron Technology, Inc.'s fourth quarter fiscal year 2026 post-earnings analyst call. Joining me today are President and Chief Operating Officer Manish Bhatia, President and Chief Technology and Product Officer Dr. Scott DeBoer, and Chief Financial Officer Mark Murphy. As a reminder, what we discuss today contains forward-looking statements regarding market supply and demand, market trends and drivers, expected performance and guidance, and other matters. These forward-looking statements are subject to risks and uncertainties that could cause actual results to differ materially from today's statements. Please refer to our filings with the SEC, including our latest Form 10-K and upcoming Form 10-Q, for risks that could affect results. Although we believe the expectations reflected in the forward-looking statements are reasonable, we cannot guarantee future results, levels of activity, performance, or achievements. We undertake no obligation to update any forward-looking statements to conform them to actual results. We will now open the line for questions. Q&A Session Benjamin Reitzes Melius Research LLC Congratulations on the new role, Scott. I want to talk about 2028. You added commentary on 2028 and said supply and demand will be tighter in 2028 than this year and 2027. I want to understand what has changed and what it means for margins. You gave more margin commentary around 2027 than before, which is good, but I don't think 2028 margins will deviate from the level you implied for 2027. Please elaborate further on 2028. Manish Bhatia President and Chief Operating Officer Sure, Ben. I'll start, and Mark can add if needed. Thank you also for the congratulations. As for 2027 and 2028, we are seeing stronger demand drivers than before. We mentioned that server units will continue to grow into 2027. At the same time, agentic AI is growing rapidly and is creating a CPU-driven demand stream. Therefore, when formulating our fiscal 2027 plan, we have already locked in more than 75% of full-year shipments, which demonstrates stronger demand and allows us to extend allocation negotiations with customers into 2028. This is one source of confidence: the demand picture for 2027 is stronger than previously expected. At the same time, as we conduct more long-term agreement (SCA) negotiations with customers, including extensions already reached, we are also more confident in the long-term outlook. Combined, these give us a very strong demand outlook through 2028. On the supply side, the structural constraints we discussed before remain: diminishing returns from technology transitions; HBM growing faster than conventional DRAM through 2028, meaning HBM's share of industry output capability is rising; and not only the capacity consumption ratio (trade ratio) for current HBM, but also higher capacity consumption ratios for more complex future HBM will constrain supply. In addition, new cleanrooms across the industry take a long time to build, qualify, and equip, and even once in production, it takes several quarters to generate meaningful shipments. Therefore, combining the demand and supply outlook, we currently do not see a point when supply and demand will balance. Mark Murphy Executive Vice President and Chief Financial Officer Ben, let me add one point. As you noted, for 2027 we have said that margins will expand over the course of the year relative to the first quarter, because prices continue to rise, but at a more moderate pace. We have also said many times that price increases will eventually slow. At that point, in addition to slower price increases, product mix will improve, and we will optimize mix based on technology and product leadership. As we said, market conditions are expected to remain tight and will provide support through 2028. Partially offsetting price and favorable mix are startup costs. But these are things we are managing, and we expect to be able to sustain strong financial performance. Benjamin Reitzes Melius Research LLC One follow-up. Sanjay had a prominent seat at Trump's lunch. Is he optimistic about industry growth and the ability to self-regulate? Does this support the optimistic guidance he gave? Was there more information from the meeting, and was memory mentioned frequently? Manish Bhatia President and Chief Operating Officer Ben, Sanjay is not on the line, so I will only speak to the conversation as I understand it. We were of course pleased to participate in that forum. Micron being invited alongside model companies and accelerator companies shows the importance of memory. The white paper framework released at the meeting has been signed by several model companies, and that framework is constructive for continuing to advance AI infrastructure, especially AI hardware infrastructure. From what I understand, one concept discussed a fair amount is that one way to manage some security issues is to adopt security solutions, and that will require more advanced hardware, including higher-performance, lower-latency memory, as well as higher-bandwidth memory. Because the responsiveness of gateways that may be set up in the future to manage security issues will depend to a large extent on the availability of high-performance and low-latency memory. Melissa Weathers Deutsche Bank Aktiengesellschaft Research Division Congratulations also to both newly promoted executives. Over the past few quarters, you have given views on total HBM TAM for 2028 and 2030. Clearly, the pricing environment has changed. Are you willing to give a new official number for HBM TAM? Directionally, can you help us estimate how much of that comes from bits and how much from pricing? And your latest view on market size? Manish Bhatia President and Chief Operating Officer Melissa, thank you. We are not updating that TAM outlook at this time. But we have said that HBM shipments are expected to grow faster than conventional DRAM, which means HBM's share of industry capacity will continue to rise through 2028. On pricing, at least for us, we have significantly raised calendar 2027 HBM pricing and will reset it at the beginning of the calendar year to narrow the profitability gap with conventional DRAM. Beyond that, we are not commenting on specific HBM TAM prospects. This market continues to grow and remains an important driver, and HBM deployment continues to be key to unlocking the broader potential of AI. So it is an important part of the market. Melissa Weathers Deutsche Bank Aktiengesellschaft Research Division Understood. On market share, do you have any update to your HBM share target? In the past you said you wanted to raise it to around the company average, meaning a share in the low 20s. Is that still the case? Also, I noticed you mentioned HBM4E and the collaboration with NVIDIA Corporation in your prepared remarks. Please provide more information on 2027 and HBM4E customer progress. Manish Bhatia President and Chief Operating Officer I'll answer the first one, and Scott will answer on HBM4E since he leads that product development. On HBM market share, about a year ago we reached a milestone where HBM market share matched our overall DRAM market share. At that time we said the target would move based on a variety of factors. Other than saying HBM share is expected to be roughly comparable to overall DRAM share, we have not updated a specific share target, nor do we treat a particular number as an absolute goal. Share will fluctuate due to many factors. But as mentioned in the previous question, HBM is a very important part of the market, allowing us to stay at the forefront of accelerator platforms designed and deployed close to customers, and it is also a key enabler for other AI promises and potential. Scott DeBoer President and Chief Technology and Product Officer Let me add on the collaboration with NVIDIA Corporation. This will be the first important custom HBM product on the market. We have worked with NVIDIA Corporation on HBM4E, the so-called NV HBM, for more than a year. We believe there is a significant opportunity because co-designing the product with a key customer makes its value significantly exceed standard HBM4E. I think this will have a real impact on the industry, showing how future systems can optimize products. Atif Malik Citigroup Global Markets Research Division My first question is about the 26 SCAs covering 35% of sales through 2030. Do these cover both DRAM and NAND? Can you break it down? Manish Bhatia President and Chief Operating Officer Atif, yes. We do not break it down, but the SCA agreements do cover both DRAM and NAND and extend through 2030. Specifically, DRAM volume is slightly below 35%, while NAND bit volume is slightly higher. Looking ahead, we have more available volume. As we said, as we continue negotiating SCAs, that number could be higher in the future. Atif Malik Citigroup Global Markets Research Division Understood. On the impact of Chinese competition, first can you confirm that China sales exposure is relatively small? Can Scott comment on how Chinese competitors are narrowing the technology gap? Manish Bhatia President and Chief Operating Officer On the first question, yes. Over the past several years and the past several quarters, our China exposure has continued to decline, and we expect it to fall to the single-digit range in fiscal 2027. Scott DeBoer President and Chief Technology and Product Officer On technology, we are currently at least 2 nodes ahead of Chinese competitors. It is important to emphasize that we are focused on maintaining technology leadership and achieving real differentiation in our products and how we compete. As Manish previously said and as mentioned on the call, our 1-gamma DRAM technology already accounts for most of our bits and is expected to become the largest node in the company's history. This technology relies on EUV. The next generation, 1-delta, is progressing well, and we are focused on mass production in the second half of next year. EUV technology will be critical for all advanced DRAM nodes going forward. Our expertise in this area, including collaboration with technology suppliers, mask technology, and more, continues to be a key differentiator for Micron. Karl Ackerman BNP Paribas Research Division You are seeing strong demand across most of your portfolio, but mobile and client bit shipments appear to have declined for a second consecutive quarter this quarter. Are higher memory prices weakening demand in that area? Also, SCA adoption is slower in that area; is your SCA growth coming from this customer group? Manish Bhatia President and Chief Operating Officer Thank you, Karl. We did see mobile and client business unit bits decline quarter over quarter, but revenue grew due to higher pricing and improved mix. It is worth noting that the high-end segments of client and flagship smartphones have strong demand for higher content and higher-performance solutions, and that is an area we are focused on. Driven by this, we expect PC and mobile industry revenue to grow even as unit sales decline. On SCAs, we have SCAs in all business units, including mobile and client. We do not break it out specifically, but maintaining SCAs helps us sustain diversified supply across all end markets. Karl Ackerman BNP Paribas Research Division Understood. A question for Scott. How do you view the competitiveness of your internally optimized base die on HBM4E relative to peers? Some customers are adopting custom solutions. Do the complexity and economic value flow mainly to compute customers or to HBM suppliers? Scott DeBoer President and Chief Technology and Product Officer Let me clarify one point first. On HBM4E, we are co-designing with NVIDIA Corporation, but it does not use an internal base die like HBM4. The HBM4E co-design is based on a foundry process, used both for custom products and JEDEC standard products. On differentiation, as we have demonstrated in past HBM products, it ultimately shows up in power consumption, speed performance, and the margin in product and customer collaboration. In previous generations, different suppliers performed differently, and we believe this will continue to be a Micron advantage in product quality and capability. Manish Bhatia President and Chief Operating Officer Karl, on economic value, HBM is a high-end product. As Scott said, NV HBM with customization is also expected to be a high-value product. We believe HBM will continue to contribute and be a high-ROI product. James Schneider Goldman Sachs Group, Inc. Research Division Congratulations to Manish and Scott. I want to understand the 10 new customer SCAs signed this quarter. What are customers asking for? Clearly they want supply and longer terms. But has the pricing structure changed? Given expected tightness in calendar 2027 and 2028, if more upside can be captured in future years, are you less inclined to require fixed ceiling and floor pricing? Manish Bhatia President and Chief Operating Officer Thank you, Jim. The SCA framework is similar, but the difference is that negotiations reflect current market conditions and future pricing prospects. Therefore, the pricing direction has been upward. These factors have been reflected in discussions with customers, whereas previous agreements were set under CQ2 market conditions. We have said that about three-quarters of SCA revenue has a clear pricing framework, and about one-quarter is open to periodic negotiation or market-based pricing. The overall framework is similar, and most pricing frameworks have floor and ceiling ranges, but new agreements already take into account current market conditions and the future tightness we see at the time of negotiation. James Schneider Goldman Sachs Group, Inc. Research Division Understood. Does that mean all price floors and ceilings have been reset to current higher market conditions? Or is the nature of the pricing terms also different? Also, have you disclosed signing the initial hyperscaler customer? Are these included in TAM? Manish Bhatia President and Chief Operating Officer There are multiple frameworks. What I am saying is that most frameworks with pricing have floors and ceilings, but we continue to use multiple frameworks. Since the last call, the 10 newly signed SCAs range from small to large customers. We do not break it out specifically, only saying that all business units currently have SCAs, and customer sizes range from small to large, including the latest 10 signed. That makes 26 in total. Christopher Caso Wolfe Research, LLC First question, please talk about capex and fab construction capex versus tool purchases. From the commentary, I heard that construction capex is growing faster, and I am not sure if I understood that correctly. Cleanroom space constraints are limiting the ability to bring in tools this year. But the increase in construction capex is notable because it will not bring bit production until at least 2029, and possibly longer. Manish Bhatia President and Chief Operating Officer Chris, I'll start, and Mark can add. That is correct. The main industry constraint is cleanroom space, because the strong growth in AI demand appeared relatively recently, and building cleanrooms takes a long time. So we are focusing on that area. You correctly pointed out that fiscal 2027 construction capex is increasing versus the prior plan, mostly for cleanrooms coming online at the end of 2028 and beyond. This both illustrates how long the construction cycle is and why construction investment must begin now, while also showing our confidence in long-term demand, both from recent market trends and from the SCA structure and discussions with customers extending commitments beyond 2030. These SCAs are transformative for us, allowing supply to match future demand and supporting investment with confidence. Mark Murphy Executive Vice President and Chief Financial Officer Chris, one addition. Most of the increase is indeed construction capex, mainly to accelerate cleanroom space availability in 2028 and beyond. We expect this trend to continue beyond 2027. One important point you mentioned is that this spending will not immediately convert into bits. These fabs will be equipped and produce wafers when needed, based on our view of the market and SCAs. These SCAs help us keep a pulse on the market and ensure returns on capital expenditure. Manish Bhatia President and Chief Operating Officer Another point, Chris, we will equip cleanrooms and build capacity according to demand trends. We have been executing long-term supply agreements with equipment suppliers to ensure we can obtain equipment when needed. But of course, we will still equip cleanrooms and build capacity based on demand trends at that time. Christopher Caso Wolfe Research, LLC Understood. A follow-up on the impact of CPU intensity on overall bit demand and supply-demand balance. This may be the biggest incremental surprise since the beginning of the year. CPUs do not have the same trade ratio effect as HBM. How much does this contribute to the supply-demand imbalance? Manish Bhatia President and Chief Operating Officer The recognition that agentic workloads run on CPUs is an important driver. LP, DDR memory, and SSDs all have very high attach rates to support these agentic workflows. These workflows have begun to be implemented and are driving real value in enterprise and consumer settings. This is also one reason server units are growing strongly, as we said, at a high-teens rate. This also shows that logic chip growth is another vector for capturing AI trends. Therefore, this drives logic chip demand and enters total AI compute demand, while also creating more constraints on DRAM, showing that DRAM is the main constraint, not logic or data center power. I mentioned that there are multiple software implementations for enterprise agentic workloads. Progress after Meta's Muse release has been rapid in just a few weeks, showing how quickly agentic workloads are delivering real value to consumers. Joseph Moore Morgan Stanley Research Division Next year's supply growth slowing is somewhat surprising in the context of capex. You mentioned the dampening effect of HBM, but the difference should not change that much. Please explain why supply is slowing despite high capex. What are the factors? Manish Bhatia President and Chief Operating Officer Joe, are you asking about both DRAM and NAND? Joseph Moore Morgan Stanley Research Division Yes, but mainly DRAM. Manish Bhatia President and Chief Operating Officer We have said that HBM is growing faster than conventional DRAM. As the industry moves from HBM3 to HBM4 and later to HBM4E in 2027, the trade ratio keeps rising. These two happening at the same time will constrain bit growth capability. At the same time, bit growth for us and other industry participants comes from new technology nodes, and the timing of transitions affects bit growth, while returns from new nodes are diminishing. These are all factors limiting supply. But the main factor remains cleanroom space across all manufacturers. Although our Idaho fab will produce wafers for the first time in the middle of calendar 2027, and other industry players may also open cleanrooms, meaningful supply growth after that will still take several quarters. These are the reasons we believe DRAM industry supply shipments will decline next year. Joseph Moore Morgan Stanley Research Division Understood. If you are conservative on industry supply, and there is more supply next year, there seems to be a lot of pent-up demand. Some AI racks are reducing specifications out of necessity; if supply increases, will specifications be raised again? Am I being too optimistic? Manish Bhatia President and Chief Operating Officer That is exactly the point. Sanjay mentioned on the main call that we do believe customers are choosing to maximize compute chip shipments given available memory supply. This creates potential demand for more memory to attach to those compute chips, thereby improving system performance and end-customer experience. Therefore, if more memory is available, it can easily be used for higher content growth in AI workloads, whether attached to accelerators or CPUs. Mehdi Hosseini Susquehanna Financial Group, LLLP, Research Division A few follow-ups. You emphasized that 2026 NAND bit shipments are below the industry average. How should we think about 2027 to 2028? Can NAND bit demand or shipments grow at the industry average of 25%? Manish Bhatia President and Chief Operating Officer Mehdi, frankly, we are not commenting on DRAM or NAND that far out. We do expect overall NAND conditions to remain tight, even though the industry is expected to grow at a mid-20% rate in calendar 2027 and 2028. Our supply growth is affected by several factors: we are currently using some Singapore cleanroom space for an advanced R&D line for future NAND growth; at the same time, we are preparing for the Singapore HBM fab ramp next year, with some existing cleanroom space used for pilot HBM operations. This is causing this year's supply growth to be below the industry. But we are confident in our technology and expect continued G9 ramp to provide high-ROI, cost-effective supply. In addition, the new cleanroom that broke ground earlier this year will come online in the second half of calendar 2028. Mark Murphy Executive Vice President and Chief Financial Officer Mehdi, given time, let me interject. Please ask your second question first, and I will comment after. Mehdi Hosseini Susquehanna Financial Group, LLLP, Research Division Sorry, Mark. I'll ask quickly. I want to understand the team's view. Looking at DRAM, especially at the wafer level, the equipment is the same. I think wafer-level DRAM is fungible, with differentiation at the back end. This wafer-level fungibility has not been seen before because past cycles were driven by a single product and customers were highly concentrated. Is my assumption correct that wafer-level DRAM fungibility allows you to manage DRAM costs better? Scott DeBoer President and Chief Technology and Product Officer Let me address one point. I think that is partly correct, but I may take the opposite view. If you look at front-end variability, or the different things we do in the DRAM process to optimize HBM, high-performance SOCAMM, LPDRAM, and DDR6, the differences today may be greater than at any time in DRAM history. You need to build different products on the same node, with more differentiation built in. Manish, would you like to add? Manish Bhatia President and Chief Operating Officer In the near term, running different products on the same manufacturing line does help us adjust mix without having to run them in different fabs; we can run them on the same line. But as Scott said, each product has its own optimization vector. HBM optimizes bandwidth and TSV and requires unique process steps; DDR and LP also each have unique process steps. The most important part of fungibility is that we can flexibly adjust to where demand is, and even adjust mix to meet customer requirements. This is all within the same fab, and that is the most important part. But I do not think it necessarily helps with cost. Mark Murphy Executive Vice President and Chief Financial Officer Mehdi, I think the call is nearing the end. I am pleased that the questions focused on the long-term solid fundamentals and performance of the technology and business. I had expected some routine questions earlier, so let me add: the first quarter guidance includes quarter-over-quarter unit growth in the single digits for DRAM and NAND bits, and cost growth in the double digits, so please keep that in mind when modeling. In addition, excluding the impact of incentive compensation, R&D growth year over year will exceed the $1 billion mentioned last quarter, and will exceed $1 billion in 2027, as we have increased R&D activity. Again, please make sure to use this information for modeling.