Q1-25 FY24 Earnings Release and Analyst Call

[SmoothHughes]

12/18/24

Press Release and Investor Presentation Commentary

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

• Fiscal Q1 was the highest quarterly revenue the company has ever had. Revenue, earnings, and margins were slightly above the midpoint of guidance. Certainly management did everything they could to be above the midpoint, if just barely.
• Data center revenue was more than 50% of total revenue, a first.
• HBM shipments were ahead of plan and revenue for this product doubled sequentially. The company estimates the total HBM market for 2024 to be $16B.
• Management previously commented that customer inventory reductions and seasonal weakness would reduce Q2 bit shipments. The impact of customer inventory reductions has been higher than expected, leading to weaker than expected Q2 bit shipments. This is the reason behind the massive miss in Q2 revenue guidance. They anticipate this adjustment period to be “relatively brief” and believe bit shipments will be stronger in FQ3 and for the rest of 2025.
• The 1-beta DRAM node is ramping now. The 1-gamma node will ramp in calendar 2025. In NAND, they are ramping the G8 and G9 nodes consistent with demand. Unsaid but implied is that they are slowing their NAND technology transitions because the NAND market continues to be weak.
• On front-end cost reductions, DRAM (ex-HBM) will be mid-to-high single digits percent for FY25. NAND will be low-teens percent in FY25. This is the same forecast as last quarter for DRAM. For NAND last quarter, they forecasted “low-to-mid teens percent reduction,” thus they have reduced their cost reduction plan, likely because they have slowed their node migration pace in response to continued market weakness.
• Earlier in December, they finalized their agreement with the U.S. Department of Commerce for $6.1B of CHIPS and science act funding. They are in a preliminary memorandum of terms with Commerce for an additional $275M of funding for their Virginia fab to make long-lifecycle chips.
• Their view of server unit growth percentage in calendar 2024 is higher than prior, up to low teens percent.
• In FQ1, HBM gross margins were “significantly accretive” to overall DRAM gross margins. They have increased their HBM TAM for 2025. They now believe that market will exceed $30B in total. They are sold out of HBM for calendar 2025, with pricing already determined. Previously, the company said the HBM market was $4B in size in calendar 2023. They also said last quarter that, as a percent of overall industry DRAM bits, HBM is expected to grow from 1.5% in calendar 2023 to 6% in calendar 2025.
• For Micron’s fiscal 2025, HBM revenue will be “multiple billions of dollars.”
• Their market share in data center SSDs is multiple billions of dollars today and they plan to grow their share again in 2025.
• The PC refresh cycle is unfolding more gradually than they forecasted. Their “flattish” expectation for units in 2024 is slightly below their prior forecast. They believe calendar 2025 PC unit growth will be mid-single-digit percentage with growth weighted to the back half of the year.
• Smartphone unit volume in 2024 will be mid-single-digits percent up in 2024, consistent with prior expectation. The company is forecasting low-single-digit percent growth in calendar 2025. They are forecasting unit growth to slow in 2025 compared to 2024, which is contradictory to other comments the company has made about AI on mobile devices being a driver of growth. They expect higher bit shipments to mobile customers in FQ3 and FQ4 of this year, as customer inventories are reduced.
• Automotive demand is lower than expected. This, and inventory adjustments, continue to mute demand. Industrial is in a similar situation.
• They expect industry DRAM bit demand growth to be high-teens percent range for 2024 and mid-teens percent range for 2025. They see 2025 DRAM industry bit supply growth “roughly in line with bit demand.” They see tightness in the leading edge because of the HBM ramp. If you put these two statements together, the implication is the market for older nodes will continue to be in its current weak state throughout 2025. If this is true, that is an okay situation, because Micron’s gross margins are pretty good now. However, the memory market isn’t typically static. It is either swinging one way or the other.
• Their industry outlook for NAND bit demand growth for both calendar 2024 and 2025 is low-double-digits percentage range. This is lower than their prior forecast, which was for mid-teens percent growth in bit demand in calendar 2025. If this forecast comes to pass for 2025 and the industry slows their supply growth, the NAND market will remain in an okay state. Not healthy but not terrible. I don’t think this will happen, because I believe bit supply from YMTC in China will continue to grow, putting downward pressure on the market.
• Lower than expected NAND bit demand in 2024 and their reduced forecast for demand in 2025 implies that “supply actions” will be necessary to get the market out of oversupply. “Supply actions” means some of the NAND capacity running today needs to be idled and tech node transitions will need to be slowed. Micron said here they have reduced capital expenditures on node transitions in their plan. They have also reduced current wafer starts by mid-teens percentage compared to prior levels.
• Three months ago, Micron’s management forecasted the NAND market to be healthy in 2025. They did not see this weakness/downturn in the NAND market coming.
• Citing analyst reports, for calendar year 2024, China-based supply will represent mid-single-digit percentage of industry bit supply in DRAM and high-single-digit percentage of supply in NAND. They sought to minimize the weight of this comment by saying that these bits are focused on China market demand in less desirable products. While it is a slow process to shift bits from one type to another, NAND and DRAM are both common, global pools of bits. This much supply being added to the pool, even if it is at the low end, is disruptive to the overall supply-demand balance. I believe this indigenous Chinese supply that has come online in the last two years is the reason the NAND market has not bounced back the way it should have, given the low levels of investment. I also think new supply of DRAM bits from China are behind the persistent sluggishness in demand from the PC, mobile, and consumer markets.
• DRAM: 73% of total revenue. Revenue increased 20% Q/Q, bit shipments increased low-double-digits percentage, and ASPs increased high-single-digits percentage. NAND: 26% of total revenue. Revenue decreased 5% Q/Q, bit shipments decreased by low-single-digits percent quarter-over-quarter and ASPs decreased low-single digits %. Bit shipments rising this much in DRAM, paired with a healthy increase in ASPs, is a good indication that the overall DRAM market is good. However, the drop in both bit shipments and ASPs indicates to me that the NAND market has tipped into oversupply and has entered a downturn. It may be a soft decline because of the lack of investment and supply action, but we have passed the inflection point.
• Their Compute and Networking Business Unit is carrying the company right now. Revenue was up 46% sequentially in the quarter. Their mobile business was down 19% and the embedded business (auto, industrial, consumer) were down 10%. The storage BU, where most of Micron’s NAND bits are sold, was up 3% sequentially. This growth, I believe, is the strength of their gains in enterprise SSDs offsetting the weakness in the overall market. SBU will see a decline in revenue in FQ2, which is not going out on a limb given the weak guidance.
• The company did not buy back any stock in FQ1-25. This is a reversal from last quarter even though their stock price traded in a similar range. There could be two reasons for this. One, they think the market weakness will persist and they want to preserve cash, which would contradict management’s statements about the FQ2 guidance being a blip. The second reason is management anticipated the sharp decline in the stock price from the guide down and held back purchases to take advantage of it. Since they commented on this weakness in the last call, they saw it coming three months ago, early in FQ1 at least. I believe the latter is the reason for the lack of buybacks. We will find out for sure next quarter, if share repurchases resumed in Q2, following the sharp drop after FQ1 results were released.
• Capital expenditures in FQ1 were $3.1B, net of government grants. $3.2B was their gross spending. FQ2 net CapEx will be ~$3B and total CapEx for FY25 will be $14B, +/- $500M. The company spent $8.1B in CapEx in FY-24, so this is a huge increase in spending. The “overwhelming majority” of this spending in the fiscal year will be to support HBM along with cleanroom buildings, back-end manufacturing, and R&D. They want to send the message that this large increase in CapEx is not going to add a lot of wafer capacity. But “support HBM” includes some amount of wafer starts and tech node transitions, because HBM is made on the most advanced wafer processes. Thus, it is the ratios within this spending that matter. $14B is a concerning amount of capital expenditures. Around $7.6B of this will be in the second half of the fiscal year, with $6.4B in the first half.
• Non-GAAP guidance for FQ2-25: revenue $7.7B to $8.1B, gross margin between 37.5% and 39.5%, operating expenses of $1.1B, +/-15M, diluted EPS of $1.33 to $1.53. This is a major downward guide vs. what analysts and investors were expecting. It is an almost 10% reduction in revenue, though gross margin only comes down 100 bps, indicating the sales decline is from lower bit demand, while blended pricing holds up. This is not what the top typically looks like for the overall memory market. I think memory is now acting as three different markets. AI DRAM is strong, non-AI DRAM is TBD, and NAND has entered a downturn.

Financial Statements

Statements of Operations

Revenue increased 12.4% sequentially, to $8.709B. This was just above the midpoint of the guided range. Unlike the last three quarters, when most or all of the increase in revenue came from rising prices, about half of the higher sales came from more bit shipments and the rest from higher ASPs. This, combined with revenue narrowly clearing the guidance, leads me to conclude management stretched out this quarter to hit their numbers. We know from the investor presentation that the rate of DRAM price increases has slowed. We also know that NAND has peaked and came down slightly this quarter. Gross margin rose 310 basis points, to 38.4%. This is well below the highs of the last two upturns. Guidance for FQ2 has gross margin decreasing by 90 bps. Management had previously guided that operating expenses would be up mid-single-digits percent, and they were down somewhat instead. I’m not reading anything into this as it is probably a combination of the timing of product qualifications, accrual for variable compensation, and employee stock sales. Net income rose more than 100%, an unrepresentatively large increase because of the high income tax last quarter (41.3%.) This quarter’s rate of 13.2% is in the guided long term range of mid-teens. Net income was $1.87B and EPS was $1.64. If it weren’t for the guidance down, this would have been a promising quarter. Annualizing these quarterly financials gives an EPS of $7.60 and $7.5B in income. The S&P500 is trading at a 29x multiple of trailing earnings, so with this multiple on annualized income you get a share price of $220 each. Even before the guidance whiff, Micron’s stock price was trading at half that. The reason, of course, is Micron’s business is so volatile. The fact that this quarter marks the first time in more than two years that Micron has had four consecutive quarters of positive earnings is an illustration of this volatility. Those last four quarters have produced $3.10 in total EPS. That makes the current share price of $90 a close match to the wider market multiple. Share count rose by 12 million, giving a clear look at the rate of employee equity grant sales during an upturn, because no shares were repurchased.

Cash Flows

Inventories produced $170M in cash this quarter with finished goods coming down by about $100M. This will be a key number next quarter when we find out how severe the weakness in the NAND market is. Operating cash flow was $3.24B even with receivables and accounts payable rising by more than one billion dollars combined. D&A was $2.0B and capital expenditures were $3.2B, so the company invested 50% more in capital assets than the rate of depreciation in this period. That is upturn level spending. CapEx consumed almost all of Micron’s operating cash flow in the quarter. Guidance for FY25 capital expenditures is even higher than this, and is concentrated in the second half of the year. They will average $3.8B per quarter in CapEx in FQ3 and FQ4 of this year. Even if an atypically large fraction of this is spending on facilities and backend, I think this is enough investment, if it is matched by Micron’s competitors, to oversupply the industry. The company paid down a small amount ($84M) of their debt in the quarters. The cash flow statement can be summed up by saying that the company spent all of their operating cash flow on capital expenditures, then $400M was paid to holders of equity and debt. Total cash declined by $355M, to $6.70B.

The Balance Sheet

Current assets were about flat from three months ago. Cash and short-term investments went down while receivables went up by an offsetting amount. PP&E rose by almost $2B. Total assets increased by the same amount, to $71.5B. Long-term debt rose by about $300M, to $13.2B. If this is the end of the upturn, Micron didn’t make nearly enough cash flow in the last year for this business to make sense. The severity of the last downturn caused the company to double its debt load. The upturn that follows such a severe downcycle is supposed to pay this back. With the rate of Micron’s capital expenditures rising as much as it is, and at least one of their markets softening, it looks unlikely the company will be able make enough cash flow from this upcycle to get back to where they were at the end of the last one, in terms of cash position. Total equity was $46.8B at the end of the first fiscal quarters. At $90 per share, Micron is trading at 2.14x book value. This is in the middle of the historical range for P/B.

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 HBM4E architecture will include an option for customers to customize the logic base die in the cube. Sanjay calls out that this die would be foundried by TSMC. Sanjay is hoping analysts will interpret this to mean HBM will not be commoditized the way most other memory is. Also, he wanted to name-drop TSMC.
• · Every call, Sanjay wastes a lot of our time going through different products, qual milestones, etc. It’s mostly commoditized products and achieving these milestones are just what is required for Micron to stay competitive. They are not a reason to invest in Micron’s shares, so stop taking up time on the call with this sales language.
• · Smartphone customer inventory dynamics are following Micron’s forecast. They believe more bits will be shipped in the second half of the year to this segment. Beyond one quarter, I don’t think Micron knows much more than the rest of us about what is going to happen.
• · Data center purchases of SSDs have slowed down.
• · In the last call, they said they believe PC customer memory inventories will reach healthy levels in the spring. “The spring” gives a wide window, but they did stick with this comment in FQ1.

Mark Murphy (CFO)

• · Auto, industrial, and consumer market customers are continuing to manage their inventory down.
• · Gross margin expansion in the quarter was driven by higher DRAM prices, shifts in product mix to data center in both DRAM and NAND, partially offset by lower NAND pricing.
• · In the second quarter, they expect DRAM bit shipments to decline sequentially and for a “meaningful” sequential decline in NAND bit shipments. They believe bit shipment growth will recover in the third quarter and that the second half of FY-25 will see more bit shipments than the first half. This is a good time to remember that memory company executives don’t know any more than the rest of us beyond one quarter out.
• · They are planning to start taking NAND underloading charges in the third fiscal quarter, as the idling of some of their NAND capacity, which is happening now, flows through.
• · FY-25 operating expenses will rise low to mid-teens percent, below their prior plan of a mid-teens rise.
• · Inventory dollars and days of inventory will increase sequentially in Q2 on lower volumes. Following that, they believe DIO will improve in the second half of FY-25. They expect to finish this fiscal year with DRAM inventories below their target level.
• · For FQ2 and the remainder of FY-25, their tax rate will be mid-teens percent. In FY-26, their tax rate will rise to high-teens percent as Singapore adopts a global minimum tax rate.

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

• · Bit shipments in PC, mobile, and consumer products are “okay,” and are above the levels of their purchases. These customers are drawing down inventory. This may be because they see the memory market softening and anticipate lower prices in the future. Even if this isn’t the full reason, customers must not be anticipating a shortage of memory if they are consumer inventory rather than building it.
• · Underloading their NAND fabs will constrain their ability to expand gross margins in the third quarter. Management sees the conditions for gross margin expansion after the third quarter. That is five months away, beyond the horizon for management to have reliable predictive power.
• · With the weakness in the NAND market, they are revising up their CapEx for FY25 as a percent of revenue from mid-30s to higher-30s percent of sales. So $14B of CapEx in FY25 at this rate, call it 37.5%, gives a total revenue in the current fiscal year of $37.3B. If that happens, it would be huge growth in the third and fourth fiscal quarters. Even interpreting “higher 30s percent” to mean 39%, this is $35.9B in revenue for the full year. That would be 14% sequential revenue growth in both FQ3 and FQ4, leading to around $2.65 in EPS in the fourth quarter at a gross margin of 47%. That would be a sharp recovery from the decline in FQ2.
• · Mehrotra has an overtone in some of his answers that he believes he’s doing the analysts a favor with the information he is providing. He comes across as condescending. This is consistent with the views of people who have worked with and for him.
• · The company now sees their share of the HBM market matching their DRAM bit share (low twenties percent) in the second half of 2025. This is a revision from their previous view that it would be by the end of 2025.
• · An analyst asked if HBM revenue for the company exiting 2025 would be $5.5B to $6.0B previously and it is now will be $6.5B to $7.0B sometime in the second half of 2025. While Mehrotra said he won’t give revenue, it gives insight into at least one analyst’s model.
• · Chris Danely asked how, with excess inventory in the market right now, and the company saying supply and demand will be in balance in 2025, can the overall DRAM market be healthy? Sanjay gave a long statement that rehashed his talking points without answering the question. He then jumped into the follow-up response and somehow sounded less clear than his first answer. The exchange left me with the impression that the CEO of the company knows that the non-HBM part of the market, which will still be 80% of Micron’s business even at the end of 2025, is not healthy now and won’t recover in the foreseeable future. He seemed uncomfortable answering and didn’t have a clear message in his response. He just reiterated a mash-up of his previous talking points.
• · LP4 and DDR4 will be 10% of Micron’s business for the rest of FY25.

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

• · Data center SSD bit shipments are expected to grow again in FQ3 after the decline in FQ2. I am dubious on the claim that the NAND market weakness will dip for one quarter and then recover. It typically doesn’t work that way in memory. When the market turns down, it takes more than a year for it to correct and turn back up.
• · Overall DRAM inventory at Micron is “fine” and will continue to improve after the second quarter, before falling below their target by the end of FY25.
• · I noticed again that the Micron executives almost never thank analysts for their questions. This is highly unusual among all the analyst calls I’ve listened to from many companies. I think it is a reflection of the culture among this management team.
• · Slower than expected PC unit growth is part of the reason for the downgrade of the near term outlook. They said again here that most of the impact is limited to the consumer segment. I don’t think it is a coincidence that this is where Chinese suppliers are selling most of their bits.
• · PC, mobile, and consumer customers are shipping memory faster than they are purchasing it from suppliers.
• · The majority of the revenue decline in FQ2 will be from NAND. DRAM is part of it, but that is mostly in the consumer oriented segments.
• · Micron’s DRAM bit growth numbers for 2025 will be in line with what they cited for total bit demand growth for the overall market.
• · An analyst said they had heard of one other supplier in NAND reducing wafer starts.
• · “The majority” of the revenue decline from FQ1 to FQ2 will be in NAND.
• · Overall DIO is now down below 150 days. DIO and absolute inventory dollars will both go up in FQ2, then they forecast this will reverse in FQ3. The CFO said this is because data center SSD sales will rise again, along with other demand such as the PC refresh cycle, will be the reason for the rise in demand in FQ3.
• · For calendar 2025 overall, they expect both DRAM and NAND bit shipments to be in line with the market. This implies they are anticipating supply response from all the major NAND players that matches what Micron does, minus some amount of inventory reduction they can manage.
• · On the HBM supply agreements, the commitment point for customers is the start of the “relatively long” lead time for the product. The CBO didn’t specify the duration of that lead time. My guess is it is the full cycle time of making one cube from wafer start, which is probably around four months. If I’m right about this, then demand is agreed to right now out to mid-April of 2025. But the use of the term “lead time” implies it is the wait time for the next available product. If this is what they mean, then the lead time is at least one year, as they have said they are sold out through calendar 2025. I don’t interpret “lead time” this way because Sadana didn’t sound as confident in his answer as he would have were lead times really out into 2026.
• · One analyst said he estimates Micron’s HBM revenue in FQ1-25 to have been in the range of $800M to $900M. The CBO said this range is “a little high.”

Summary

I believe the memory market is now trifurcated. There is the leading edge AI DRAM market, the mid-to-lagging node DRAM market, and the NAND market. The AI DRAM market is strong and growing rapidly, but it is a fraction of Micron’s overall business. Even when it reaches their goal of low-twenties-percent of their business sometime in the second half of calendar 2025, the majority of Micron’s DRAM will still be non-AI. That segment, what I called the “mid-to-lagging node DRAM market” is experiencing persistent weakness, because of sluggish demand in PC, mobile, and consumer segments. The third market is the NAND market. In my estimation, last quarter marked the peak for NAND, and it is now entering a downturn. Micron has responded quickly by reducing their NAND wafer starts by around 15%. That is downturn behavior, despite what management said about the soft FQ2 guidance being a temporary slowdown. This upturn is defined by two features. First, the emergence of an entirely new and powerful source of demand, centered around a new type of memory. That source of demand is artificial intelligence, and that product is high bandwidth memory. The second feature is the persistent softness in non-HBM DRAM, and in NAND. Yes, pricing in both rose in four consecutive quarters, but now it has stalled. It feels as if there is an unseen force out there weighing on these markets. I’ve suspected for the last few months that that force is new indigenous Chinese supply. Micron said it estimates China’s DRAM bit share in 2024 was mid-single-digits percent and NAND was high-single-digits percent. These new bits are in the low end of the market. ~5% to 9% new supply is plenty large enough to stifle the undersupply that was emerging in 2024. If the weak FQ2-25 guidance marks the first quarter of the downturn, then this upcycle will be the shortest one in at least thirty years. On top of that, it followed a destructively deep downturn. Historically, the deeper the downturn the stronger the upturn that follows. That hasn’t happened this cycle, at least not yet. I think the emergence of China is the reason. Micron doubled their debt load over the last three years in response to the dramatic collapse in memory prices and the $8.2B in collective losses that resulted for the company. Now they have a capital expenditures plan underway that contains an upturn level of spending, but they have only begun to recoup all the cash they lost in 2022 and 2023. If this really does market the end of this upcycle, Micron’s business is running to stand still.

-S. Hughes (cyclical long MU)