Q2-25 FY24 Earnings Release and Analyst Call

Stocks A to Z
1

3/20/25

Press Release and Investor Presentation Commentary

Here are the points from the press release, outside the financial statements, that I found notable:

  • HBM revenue for the company grew more than 50% sequentially, to over $1B in the quarter. That puts it at around 16% of their DRAM revenue, not far below their low-20% DRAM market share. Their HBM ramp is ahead of plan.
  • Last quarter, their HBM revenue doubled sequentially. Thus, HBM revenue in FQ4-24 was approximately $325M, then $650M in FQ1-25, then $1B in FQ2-25.
  • They see the ramp of HBM across the DRAM industry and demand for leading edge products “contributing to tightness at the leading edge and constraining non-HBM DRAM supply.” This statement is short of predicting that the non-AI DRAM market will be constrained, just that the ramp of HBM will take bits out of other segments. It sounds like a statement about improving market health, but read closely, management is only saying something obvious here; that more DRAM bits going to HBM takes bits away from non-HBM products.
  • They shipped their first products on their 1-gamma DRAM node last month, leading the industry. This is the first node using EUV. It has a >30% bit density improvement over their 1-beta node.
  • They are ramping their G9 NAND node, “mindful” of the currently oversupplied NAND market.
  • Construction began in January on their new advanced packaging facility in Singapore.
  • They reached a construction milestone in January on the new fab in Idaho that earned them their first disbursement of money from the CHIPS act. “Meaningful” DRAM output will come out of that fab beginning in fiscal year 2027, which begins in September of 2026.
  • The company has increased their view of HBM TAM to >$35B in calendar 2025. The company is on track to reach HBM share similar to their overall DRAM supply share, “on a run-rate basis,” in calendar Q4 of 2025. This is either in their fiscal Q1-26 or early in their FQ2-26. I read “supply share” to mean bit share rather than revenue share. But assuming the Big Three have similar HBM trade ratios, similar HBM pricing, and similar HBM process technologies, bit share and revenue share will be similar.
  • They are sold out of their HBM supply through calendar 2025 and are in discussions with customers about their 2026 supply.
  • The “vast majority” of their HBM shipments in the second half of calendar 2025 will be 12-high HBM3E.
  • They expect multiple billions of dollars of HBM revenue in FY-25. This is an obvious statement to make, given that they already said they have sold at least $1.625B of HBM products so far this fiscal year at the half way point. As they are still ramping, the Q2 run rate of >$1B of HBM products implies at least $2B more of HBM sales in Q3+Q4, bringing total HBM sales for FY25 to >$3.625B, at a minimum. Given they are ramping, HBM sales closer to $5B are more likely. I wonder why they didn’t say “several billion dollars of HBM sales” instead.
  • Data center NAND demand moderated in fiscal Q2 due to short-term customer impacts. They see a return to bit shipment growth in the months ahead. This is the time I remind readers that company executives don’t know much beyond one quarter out.
  • PC unit growth in calendar 2025 is expected to be mid-single digits, weighted to the second half of the year (same forecast as last quarter.) That actually means they think unit growth will be low single digits, and they hope it improves in the second half of the year to bring up the average.
  • Mobile phone unit growth will be low single digits percent in calendar 2025 (same as last quarter.)
  • Auto, industrial, and consumer segments are in the late stages of adjusting their inventory levels. This is a nice way of saying demand from those segments is still weak.
  • Calendar 2024 DRAM bit demand growth was in the high teens percent, consistent with their prior expectations.
  • Calendar 2024 NAND bit demand growth was approximately 10%, slightly below their prior forecast. Management did not forecast the weakness that has emerged in the NAND market.
  • They are forecasting calendar 2025 DRAM bit demand growth to be mid-to-high teens percent. That is up from the mid-teens percent growth they forecasted last quarter. Since they have increased their estimate for the HBM TAM size, this bit demand growth increase is from a higher view of HBM demand in calendar 2025.
  • Micron sees NAND bit demand in calendar 2025 in the low-double-digit range, which is slightly better than where 2024 came in. Their forecast for NAND bit demand growth in 2025 is the same this quarter as it was last quarter.
  • Over the medium term, their industry bit demand growth averages are mid-teens percent for both DRAM and NAND.
  • Their last bullet point on the “Demand outlook” slide is a plea to the rest of the NAND industry to increase the time between node transitions and for sustained reductions in industry wafer capacity and CapEx. They also suggested other companies underutilize their NAND fabs in the near term to improve the current oversupply situation.
  • The company expects to maintain their market share in DRAM and NAND. They also say their bit growth in both DRAM and NAND will be below demand growth. The company also plans to bring down their inventory in their FQ3 and FQ4, so they plan to absorb at least some of the difference between their bit growth and the higher market demand they are forecasting with this draw-down of inventory.
  • The wafer trade ratio for HBM3E is 3:1, something we already knew. They expect the trade ratio to exceed 4:1 for HBM4E, which is two product generations away.
  • The company continues to underutilize their NAND fab capacity. NAND wafer output is down mid-teens percent from prior levels. They plan to take this period of underutilization as a chance to make capital efficient conversions to leading edge nodes. This will structurally reduce their wafer capacity by 10% exiting fiscal 2025 compared to what it was exiting fiscal 2025. This sounds good, but what matters is bit output. They are going to transition NAND tech nodes faster than the prior plan, so bit output will increase per wafer. Yes, they are reducing wafers, but they are increasing bits per wafer. Depending on the node mix change, the increase in bits per wafer is likely more than 10% per wafer. Thus, while wafers will go down 10%, bit output will still increase over the course of FY-25.
  • Capital expenditures in FY-25 are unchanged at $14B.
  • A “very limited” volume of products will be subject to the new tariffs on imports from China, Mexico, and Canada. They plan to pass those costs on to customers. This doesn’t mean they will be able to do so.
  • DRAM: 76% of total revenue. Revenue decreased 4% Q/Q, bit shipments decreased high-single-digits range, and ASPs increased mid-single-digits percentage. I think the DRAM market is near balance, because bits declined in the quarter, yet prices rose. There is also the factor of growing HBM fraction in their product mix, which will increase ASP per bit. The second derivative of DRAM ASPs is shrinking. HBM revenue has risen about the same dollar value in this quarter as in FQ1, so the rate of ASP increase being half as much indicates the non-HBM DRAM pricing reduction is higher. This is consistent with management commentary that the product mix this quarter was weighted more heavily to consumer products.
  • NAND: 23% of total revenue. Revenue decreased 17% Q/Q, bit shipments were “modestly higher” quarter-over-quarter and ASPs decreased high-teens percentage range quarter-over-quarter. I said last quarter that I believed NAND has tipped into oversupply. That is now clear, as ASPs dropped 19% on essentially flat bit shipments.
  • CNBU, where HBM revenue is recognized, continues to carry the company. Their revenue was up 4% sequentially, a surprisingly small percent increase, considering HBM grew 50% in the quarter. HBM was at least $1B of revenue out of CNBU’s total sales of $4.6B, an increase of at least $350M sequentially. That means >8% of the rise in revenue was from HBM. Since overall revenue for this business unit rose only 4%, then we can conclude that the rest of their business declined somewhat. At least some of this decline was from the ~9% drop in DRAM bit shipments. Note that, because of the 3:1 HBM trade ratio, if all else stays the same, a shift in DRAM product mix to more HBM will reduce overall bit shipments, for the same level of revenue.
  • Storage Business Unit revenue was down 20%, because of the weak NAND market.
  • Mobile Business Unit revenue was down 30% sequentially. The company pinned this on customer inventory reductions. This is a reflection of weakness in both the non-AI DRAM market and the overall NAND market.
  • Embedded Business Unit revenue was down 3%, also because of customer inventory reductions.
  • Again this quarter, the company did not buy back any stock. That is the second quarter in a row they have not done so, despite ample opportunities to buy back their own stock at prices in the $90s. This answers my question from last quarter. My conclusion is that management is concerned about current weakness in non-AI memory markets, and they want to preserve cash.
  • Capital expenditures in FQ2 were $3.1B, net of government grants. This is the same as last quarter. They were less precise in forecasting how their CapEx will be spent the rest of the year than they were last quarter, saying FQ3 expenditures will be >$3B. Total CapEx for FY-25 will still be $14B, with $7.3B of that already spent. Thus, they will average about $3.35B in CapEx in each of the next two quarters. As I said last quarter, this total amount of capital expenditures in FY-25 is a large rise over the $8B spend in the prior fiscal year. If Hynix and Samsung also increase CapEx similarly, I think the DRAM market will be oversupplied by the end of calendar 2025.
  • Non-GAAP guidance for FQ2-25: revenue $8.6B to $8.8B, gross margin between 35.5% and 37.5%, operating expenses of $1.13B, +/-15M, diluted EPS of $1.47 to $1.67.

Financial Statements

Statements of Operations

Three months ago, management provided tepid guidance for FQ2 and the stock dropped twenty dollars. That guidance was for a 9.3% decline in revenue. The company beat that, posting revenue of $8.05B, a level halfway between the midpoint and the high end of the guided range. However, gross margin of 36.8% was near the bottom end of the guided range. They may have taken some lower pricing to stretch shipping bits in order to hit a higher revenue number. It could also just be the difficulty of predicting pricing. At the bottom line, where it matters, the company came in above the top end of their guided range. Earnings per share of $1.41, which is income of $1.58B for the quarter, was down sequentially from $1.67 in the prior period. I can’t see under the hood of the guidance that was provided last quarter, so I don’t know why gross margin was at the low end of guidance while EPS was above the high end. I can guess some of it was from a lower tax rate. They provisioned 13.1% last quarter and only 10.1% this quarter. The other candidates are operation expenses. These were about flat from FQ1 to FQ2. Even with steady OpEx, the drop in revenue caused operating margin to fall from 25.0% to 22.0%. Management made good, at least in guidance for FQ3, that the revenue drop seen in FQ2 was the result of short-term demand weakness. They are guiding for a 9.3% sequential increase in revenue at the midpoint. This is a return to the revenue level of two quarters ago. However, the company’s guidance on gross margin does not recover to that prior level. In FQ1-25, GM was 38.4% and the midpoint of guidance for FQ3-25 is 35.5%. EPS bounces back with revenue, though not to the level seen in FQ1, because of the weaker gross margin. In the last upturn, gross margin flattened out for a year before falling off sharply. There was a last gasp surge in revenue before this drop. That was the COVID upturn, a cycle with unusual characteristics. This does make me wonder about the decoupling of revenue and GM. The two competing forces in Micron’s financials are the growing fraction of HBM in their total revenue and the persistent weakness in the rest of DRAM, and in NAND. The guided recovery in revenue but weakness in gross margin suggests the non-HBM part of their business – which makes up most of it – is stronger right now.

Cash Flows

The weakness in memory markets hit Micron in the inventory line of their balance sheet. The value of WIP has been stable over the first six months of the year. Raw materials are up about $80M. It is in finished goods where the company is absorbing changes in demand. In the first quarter, when the market was stronger, finished goods inventory came down by about $100M. In the second quarter, FG rose by almost $150M. The company held back some finished products, either for lack of demand at any price, or, more likely, to support pricing. The company tightened other components of their working capital in the year so far. They have brought in $338M of receivables and paid down $714 of AP and other accrued expenses. Operating cash flow, which Micron’s management likes to focus on because it takes out how CapEx intensive their business is, was $7.19B in the first six months of the year. $4.11B of this cash flow, or 57%, was depreciation and amortization. In the first half of the year, Micron has spent $7.26B in capital expenditures, more than their total operating cash flow over the same period. And they are planning on another $6.8B of CapEx in the second half of the year. So far this year, they have received $1.03B of government incentives. The two biggest projects that brought this money in are the construction of the ID1 fab in Boise and the new AT facility in Singapore. Without these government grants, the company would be about cash flow neutral. Elsewhere in investing activities, they rolled about $800M of their available-for-sale securities. Down in the financing section, they also rolled over $2.6B of their long-term debt. There were two parts to this activity. Both were executed in January. One, they issued $1.0B of new unsecured 2035 notes with an interest rate of 5.80%. And two, they took a term loan for $1.68B maturing in January of 2029 and an adjustable rate ranging from 0.875% to 1.50% above SOFR. Also in the quarter, the company paid down $2.45B of other debt: notes due in 2026 and two term loans, one due in 2026 and the other due in 2027. All in, the debt activity in the first half of the year has been a wash. $261M of cash was paid out in dividends and another $121M out to “other.” For the first six months of the year, total cash has increased by $511M. They churned both working capital and debt in a net neutral way, paid out all their operating cash flow in capital expenditures, and have generated positive cash flow only because they received $1.03B of government incentives. The company really needs gross margin in the mid-40% range, at least, to generate some decent cash flow.

The Balance Sheet

The company’s balance sheet didn’t change much in the last three months. Current assets rose ~$200M, which is 1% of the $24.7B total value. They pulled in $919M worth of receivables, grew inventory by $302M, and saw other current assets rise by $186M. Cash balance stayed almost flat so lower short-term investments made up the rest of the difference. The consistency of the cash balance indicates management targets a specific level of cash, which ended the period at $7.552B. Total assets stand at $73.0B, up about $1.5B as the company invests more in CapEx than is exiting in depreciation. Total liabilities also changed by a small percentage during the quarter, ending at $24.42B. Long-term debt increased by $599M, to $13.85B. This is more than twice the level the company carried at the end of the last upturn. The rise in long-term debt was majority caused by rolling refinancing costs in. Book value on 1,123K diluted shares is $43.31. Micron trading at $94 per share puts its equity price at 2.17x book value, in the middle of the historical range.

Conference Call

Numbers given during the conference call are non-GAAP. As I provide summary and commentary on the investor presentation slides above, this section will only include new information not included in the slides, plus bullet points on the Q&A with analysts.

Sanjay Mehrotra (President and CEO)

  • · The CEO, Sanjay Mehrotra, spent the first seventeen minutes of the call reading the slides to the audience, with little additional commentary. This is a silly. Send out the slides ahead of the call, let analysts and investors read it themselves, and leave more time for Q&A. I think the management team does this because they don’t want more time answering questions.
  • · NAND shipments to the data center segment “moderated” in FQ2. They anticipate bit shipment growth to increase again in FQ3, indicating that bit shipments to data center customers were negative in the second quarter.
  • · They expect a “strong ramp” of HBM throughout calendar 2025.

Mark Murphy (CFO)

  • · DRAM ASPs rising mid-single-digits percent in the quarter was a result of improving portfolio mix. I think that means a higher fraction of HBM products.
  • · NAND bit shipments in Q2 were above management’s expectations, driven by higher consumer-oriented shipments. Those are the lowest-margin bits. Either they had more of this type of bits than expected and wanted to move them out, or they sought to boost sales.
  • · The theme across non-AI customer segments for why demand was weak in the quarter is “inventory actions” or “inventory adjustments.”
  • · The gross margin decline of 160 bps in the quarter was from pricing erosion in consumer-oriented markets, especially in NAND, and NAND mix shift to consumer markets. The language chosen suggests the price erosion is limited to consumer products, but “consumer oriented” includes the PC and mobile segments. Thus, the price erosion seen in the quarter was more widespread than just consumer devices like TVs.
  • · OpEx was flat sequentially. R&D was lower than expected because of timing of product qualifications.
  • · Their income tax rate was lower than guidance because of one-time items.
  • · Inventory finished the quarter at 158 days, an increase of 9 days from the prior quarter.
  • · The two debt actions taken in the quarter extended their maturities by about a decade.
  • · After the end of the quarter, they renewed and expanded their revolving credit facility by one billion dollars. They may have done this out of concern for a pending downturn, or they may simply have wanted higher liquidity when the time for renewal came up.
  • · For FQ3, they are forecasting higher bit shipments in both DRAM and NAND. Also, they foresee lower gross margin in the current quarter, from a higher mix of consumer-oriented products. My feeling from these comments is the weakness in the NAND market, especially weaker demand in data center SSDs, is the largest single factor behind their tepid margin profile.
  • · Fiscal 2025 operating expenses will increase more than 10%.
  • · They anticipate inventory to decrease in the third fiscal quarter. DRAM inventories are expected to be “tight” by the end the fiscal year. That is at the end of August, six months away. Regular readers know I am dubious on most predictions on the memory market beyond one quarter.
  • · For FQ3 and FQ4 of 2025, they anticipate their tax rate will be approximately 14%.

Mehrotra and Murphy spent more than half the call on prepared remarks. This is way too much. Sanjay loves to hear himself talk, which demonstrates a baffling lack of self-awareness as he is monotone and uninspiring to listen to. Write an investor letter and leave more time for Q&A.

Analyst Q&A

  • · Conditions in the market have improved since they made public comments in the Wolfe conference in February. They do expect gross margin to be “up somewhat” in FQ4-25. HBM growing as a percentage of total mix is a tailwind here. NAND underutilization is a headwind. They will see more of those costs hit them in the fourth fiscal quarter as inventory clears. Also in the fourth quarter, they will see higher construction costs as well as higher cost from DRAM node transitions.
  • · Improved business conditions are the result of lower customer inventories, as management predicted.
  • · They do expect bit growth in both DRAM and NAND in FQ3. Asked directly if management expects to see revenue growth in DRAM and NAND in FQ3, the CFO said the “bias to growth” would be in DRAM. He wasn’t willing to commit to revenue growth in NAND. Micron’s NAND ASPs have declined for the last two quarters, by 1% in FQ1 and then down 19% in FQ2. I think the NAND market will continue to deteriorate such that, even if they ship more NAND bits, price declines will offset, leading to a sequential decline in NAND revenue.
  • · For FY-25 overall, they expect DRAM cost to be flat and NAND costs to be down low double digits. The flat DRAM cost is because HBM is more expensive to make on a per-bit basis. I am surprised NAND will be down that much given the underloading charges they are taking on. That indicates they are moving quickly to a mix richer in advanced nodes.
  • · When asked if gross margins will return to high-30%-levels, management talked about cost, their products, their strong position, but nothing useful on pricing.
  • · When asked directly if the current improved pricing trends are sustainable, Mehrotra talked about stronger demand. He also said leading edge supply is tight. What he didn’t address is the supply of older nodes, where the market is worst.
  • · They are at 158 DIO today and have said they will be below the 120 DIO by the end of the fourth quarter. One analyst said this is a steep reduction. Are they sure? Management said it is enabled by HBM growth and the trade ratio plus the underloading of NAND capacity. They declined to provide a target for DIO exiting FQ3.
  • · NAND underutilization will flow through into results in FQ4 and into FY26. They believe growth elsewhere in the business and a better product mix will lead to higher gross margin in the fourth quarter.
  • · One analyst asked why gross margin is ten points lower at the same revenue level as was reached in the last cycle. He asked if they could forget about margin ever getting back into the fifties. This is an uncomfortable question because it asks if the memory industry has structurally become worse since the last cycle. Sanjay gave the kind of rambling non-answer that he is famous for.

Post Earnings Q&A call with Mark Murphy (CFO), Sumit Sadana (CBO) and Manish Bhatia (VP of Global Operations)

  • · Higher startup costs in FQ4 are in the range of a 30 to 40 bps drag on gross margin, depending on revenue. This will increase as activity in the construction of the Idaho fab and on the newest DRAM node grows.
  • · The TAM increase in the HBM market is driven in part by a faster shift to 12-high stacks. 12-high cubes have a higher per-bit revenue than 8-high.
  • · HBM revenue in FY-26 will be significantly higher than in FY-25.
  • · Server SSD demand will improve in the coming months and will continue to get better as calendar 2025 unfolds.
  • · China supply question! Some Chinese customers stayed on DDR4 longer in order to consume indigenous Chinese supply.
  • · Micron’s supply growth will be below market demand growth They will make up the difference with inventory.
  • · Management gave a long answer on why gross margins are so much lower at this point in this cycle than they have been in past cycles. To summarize, NAND is not a healthy market. DRAM is better and Micron is growing their share in the most profitable segments, while maintaining their market share.

Summary

This upturn is unusual. It has been marked by the emergence of HBM as a new product category and the insatiable (for now) demand from AI servers. This cycle is also the first one to be affected by new bit supply out of China, from CXMT in DRAM and YMTC in NAND. These new players muted the strength of the upturn. Back at the end of 2019, the normal cyclicality of a downturn was concluding. Signs of strength in memory markets were visible in the Fall and Winter of 2019. Things were improving for memory makers right before COVID struck in 2020. First, demand evaporated. Then, it surged back. But it did so in a way that didn’t match the supply profile of the industry. Consumers all bought new PCs and new mobile devices at the same time. This led to undersupply in some segments and oversupply in others. While misshapen, the memory upturn that coincided with COVID ran its course in the typical duration of two years. Micron’s stock price peaked in February of 2022. A brutal downturn followed, lasting until late summer of 2023. The lollapalooza of computer buying during COVID pulled forward multiple years of memory consumption, leading to demand weakness that continues today.

But large language models burst onto the scene in 2022, and HBM demand followed. Hynix was, and is, the market share leader in this segment. Their upturn began in March of 2023. The rest of the memory players saw things improve six months later, in the fall of 2023. Pricing in both DRAM and NAND rose sharply for a year. But something was out there, weighing on the memory markets. Even as growing HBM supply, with its 3:1 trade ratio, soaked up the supply of DRAM bits from other segments, pricing strength subsided in the second half of calendar 2024. NAND pricing weakened even faster. The weight on the memory markets was new supply from indigenous Chinese suppliers. The already unhealthy NAND market now has five players, which is two too many. NAND’s upturn only lasted a year. DRAM has five participants. Two of them, Nanya in Taiwan and CXMT in China, only have products in the low end of the market. Consequently, prices in those segments (DDR4 and LPDDR4) have been under severe downward pressure for more than half a year. The Big Three, Hynix, Samsung, and Micron, are simultaneously moving as fast as they can away from the low end of the DRAM market and into more HBM. In NAND, Micron is pleading publicly for other manufacturers to structurally lower supply. If they do that, they will be ceding market share to YMTC. It is either this, or keep their foot on the gas, hoping to suffocate the Chinese upstart with lower prices. Micron’s management team has chosen the former.

Where does that put us today? Micron’s financials dipped in FQ2 and the company guided for improvement in FQ3. While that higher FQ3 guidance is to a level of revenue similar to what they produced in FQ1, the gross margin will be lower. That is to say, the pricing mix in the non-HBM, non-leading-edge portion of Micron’s business has deteriorated more than the improvement seen from their ramp of HBM. And that is the question facing Micron investors today. Do you think the ramp of HBM, and the return of customer demand following a stretch of inventory adjustments, will be enough to counteract the current market weakness? I think NAND is in real trouble. Pricing peaked two quarters ago and they are in for at least another two quarters of pain. In DRAM, I am less convinced. If HBM demand falters, either because of oversupply or because the AI hyperscalers pause their torrid rate of investment, it is over. If HBM demand stays strong, then there is a chance the DRAM market will continue to improve for the rest of 2025 and into 2026. Micron’s gross margin’s seeming to flatten in the mid-to-high 30% range, if that does turn out to be the peak, are not enough to make up for all the losses suffered in 2022 and 2023. If this comes to pass, Micron will have doubled their debt load to stay in a business with economics that have deteriorated in two consecutive cycles.

– S. Hughes (short MU)

2 Likes