03/20/24
Press Release and Investor Presentation Commentary
Here are the points from the press release, outside the financial statements, that I found notable:
- Revenue, gross margin, and earnings were all above the high end of the guided range.
- In Q2, capital expenditures were $1.25B, resulting in adjusted free cash flow of negative $29M.
- Improved market conditions resulted from strong AI server demand, heathier demand in most end markets, and industry supply reductions. AI server demand is driving rapid growth in HBM, DDR5, and data center SSDs, causing a positive ripple effect across all end markets.
- The company expects DRAM and NAND pricing levels to increase further throughout calendar 2024, leading to record revenue and “much improved profitability” in fiscal 2025.
- Over 75% of DRAM bits are on 1-alpha and 1-beta nodes and over 90% of NAND bits are on 176-layer or 232-layer nodes.
- FY-24 cost reductions (excluding HBM) will be in line with long-term targets of mid-to-high single digits percent in DRAM and low teens percent in NAND.
- Yields have been matched between EUV and non-EUV flows on 1-alpha and 1-beta DRAM nodes.
- Memory and storage inventories in data center (implies at data center customers) are expected to normalize in the first half of 2024. This is the same forecast as last quarter, following a negative revision (from early 2024) in the prior quarter.
- Inventories are “near normal levels” for auto, industrial, and other end markets.
- Data center unit shipments will grow mid-to-high single digits in calendar 2024.
- Their HBM is sold out in calendar 2024 and the “overwhelming majority” of their 2025 supply has already been allocated. They will sell “several hundred million dollars” of HBM in FY-24 and expect HBM revenues to be accretive to DRAM and overall gross margin beginning in FQ3-24.
- PC unit volumes will grow low single digits range for calendar 2024. This is a down-revision from low-to-mid single digits unit growth, which is what they said last quarter. Part of that change may be from the upward revision at the end of 2023.
- Smartphone unit volume in calendar 2024 will grow low-to-mid single digits.
- In calendar 2023, DRAM bit demand growth was in the low double-digit percent range. NAND was in the low-20s percent range. Both of these are “a few percentage points higher than previous expectations.” The year finished stronger than they thought it would. In the FQ1-24 call, they believed 2023 would end with high-single digits DRAM bit growth and high-teens percent NAND bit growth.
- Calendar 2024 bit demand growth will be near the long-term CAGR for DRAM (that is mid-teens percent) and around mid-teens for NAND. The company revised upward their expectations for DRAM and NAND bit demand for all of calendar 2024.
- Their medium term expectations for bit demand growth CAGRs are now mid-teens for DRAM and low-20s percent for NAND.
- Calendar 2024 bit supply growth for both DRAM and NAND is expected to be below demand for both DRAM and NAND, for the industry and for Micron.
- The company’s “high-volume manufacturing nodes” are now fully utilized, at a structurally lower capacity. At the end of fiscal 2024, they will have low double digits percentage lower wafer capacity in both DRAM and NAND than the peak levels in fiscal 2022.
- HBM3E consumes 3x the wafer supply as DDR5 to produce the same number of DRAM bits.
- Fiscal 2024 CapEx for the company is unchanged at $7.5B to $8.0B. WFE spending will be down year on year in FY-24.
- Projects in China, India, and Japan are continuing as planned. CHIPS money and other incentives are needed to execute projects in Idaho and New York. This is signaling to the government that those expansions aren’t guaranteed. At least in Idaho, this is only partially true as there are several cranes operating, and construction is proceeding at the new fab site there.
- DRAM in the quarter: 71% of total revenue, up 21% sequentially, bit shipments up low-single digit percent Q/Q. ASPs increased by high-teens Q/Q.
- NAND in the quarter: 27% of total revenue, up 72% sequentially, bit shipments decreased by low-single digit percentage Q/Q, ASPs increased over 30% Q/Q.
- The Storage Business Unit saw the largest revenue gain, both sequentially and year over year. With HBM/AI getting all the press, the strength of NAND in the quarter is a big positive for the company. With that said, NAND is coming off a lower bottom that DRAM. SBU still posted an operating loss of $217M in the quarter compared to around break-even for EBU and MBU, and a $28M operating profit for CNBU. Pricing is improved, but the NAND market is still in rough shape.
- Non-GAAP guidance for FQ3-24 is revenue of $6.60B +/-$200M, gross margin of 26.5% +/- 1.5%, operating expenses of $990B +/-$15M, and diluted EPS of $0.45 +/- $0.07.
Financial Statements
Statements of Operations
Revenue for the quarter was $5.82B, almost a quarter of a billion dollars above the high end of guidance. That is a 23% sequential increase in revenue, almost all of which came from better pricing. DRAM bits were only up low-single digits percent, and NAND bits were down by a similar percentage. The company is guiding for a 13.3% increase in revenue in Q3 over Q2. I think they are under-calling this again, because pricing will be stronger than they are modeling. The memory companies usually follow the same thinking at each phase in the cycle. In this phase – the steep part of the recovery – they are too conservative about how fast the market will improve. My prediction is they will pull out all the stops to get revenue over $7.0B in the third fiscal quarter. COGS was lowered by $382M in the quarter because of prior write-downs of inventory value. Adjusting for this, gross margin in the quarter was 12.0%, vs. the 18.5% reported. With both Q1 and Q2 adjusted to take inventory write-downs back out, the sequential increase in gross margin from Q1 to Q2 was 2550 basis points. Operating expenses were about flat, with the company commenting that employee compensation was higher in the period. Operating income was reported to be $191M positive, but this includes a $200M payment for a patent cross-licensing agreement, so an operating loss of $9M, effectively flat, better represents reality. There was also a tax benefit of $622M in the quarter, the result of changing their tax treatment because of improving market conditions. Backing out both of these adjustments, the company swings from $411M of net income to a net loss of $411M, or negative $0.37 per share. The share count increased 14M, to 1.114B. Yes, pricing improved a lot in the quarter, but the company is still losing money. But, the company lost $1.8B in the prior quarter, so losing only $411M this quarter is a massive improvement. If my revenue forecast for FQ3-24 of $7B is accurate, the company will post a profit of nearly $1B in the current quarter.
Cash Flows
In the first six months of the year, the company generated $2.62B in operating cash flow. Because of capital expenditures, which were well below depreciation and amortization at this point, free cash flow was negative $560M. Stock-based compensation is going at a run rate of $200M per quarter. The company has invested $3.18B of capital expenditures so far this year, so there is at least $4.32B to go in the second half, based on their (unchanged) guidance for the year. There have been $234M of government incentives received during the year. I think most of this is from Singapore, but the filing doesn’t specify. The company rolled over about $1B worth of debt, paying off $102M in the process. Micron’s cash position in the first half of the fiscal year is lower by $555M. Payment of dividends, cash consumed by receivables, and paying off a little debt are the predominant uses of this cash, in my estimation. I am standing by my prediction, made ninety days ago, that Micron will be free cash flow positive in the third fiscal quarter of 2024.
The Balance Sheet
Total assets declined by about half a billion dollars in the last six months, to $63.8B. Most important in this is the $500M reduction in cash. Also worth noting is receivables grew by $500M, a 20% increase. Micron’s debt load remains the most important feature of their balance sheet coming out of the downturn. The company has almost $13.5B of total debt, more than double the load they carried before the downturn started two years ago. Paying this down should be their top priority as free cash flow increases, because these new liabilities carry higher interest rates than the older paper. This management team has been poor at capital allocation with respect to share buybacks. They are in the most cyclical of industries, yet they do the opposite of what they should in buying back shares. During the downturn they suspended buybacks when they should have been using their borrowing capacity to buy back more shares. Now that the upturn is underway, I expect to see them reinstate their share repurchase plan, buying shares when they are highly valued. On inventory, the total level came down slightly, to $8.27B from $8.39B. With revenue increasing, this is actually a reduction in DIO. Within this inventory, finished goods have declined by 25% in the last six months while WIP has risen slightly. Expect to see finished goods inventory continue to come down.
Conference Call
Numbers given during the conference call are non-GAAP.
Sanjay Mehrotra (President and CEO)
- · Mehrotra mostly just read the slides to the audience, taking up twenty minutes of time that could have been saved with a shareholder letter.
- · The company’s capital expenditure plan for FY-24 remains unchanged at $7.5B to $8.0B, with WFE spending down year-on-year in FY-24.
- · The company’s current profitability levels are still well below what is needed to support continued investment in R&D and capital expenditures.
- · In DRAM, they expect tight supply conditions across all end markets.
Mark Murphy (CFO)
- · Cloud revenue more than doubled sequentially.
- · $382M of benefit was realized in Q2 from selling previously written-down inventory.
- · Variable compensation expense was higher than expected in the quarter, leading to OpEx above the midpoint of the guided range. The company set aside more money for bonuses because they think the fiscal year will be better than they did three months ago.
- · The large tax benefit in the quarter is from applying a different global tax rate than was used in prior guidance.
- · Inventory finished the quarter at 160 days. Leading-edge supply in both DRAM and NAND is tight. They expect to reach their target level of 120 days of inventory by the end of FY-24.
- · Previously written down inventory has completely cleared as of the end of FQ2-24. R&D expense is increasing OpEx in the third quarter. The company expects total OpEx for the fiscal year of $4.0B.
- · In FY-25, they expect their effective tax rate to be in the mid-teens.
- · Positive free cash flow is expected in both FQ3 and FQ4 of the current fiscal year.
Mehrotra and Murphy spent 35 minutes in this call on prepared remarks. This is still way too much (write an investor letter) and continues a trend for this executive team. There was a disruption before the analyst Q&A that lasted for several minutes. I’ve never experienced this. They had technical difficulties.
Analyst Q&A
- · Most of their HBM3e supply for FY-25 is allocated. FY-24 supply is sold out. The CEO reiterated they will match their overall DRAM share to their HBM share sometime in FY-25.
- · By the end of FY-24, their wafer capacity will be low-double-digits percent lower than the peak reached in FY-22.
- · CapEx in FY-25 for the company will be higher than in FY-24, both for WFE and in total.
- · High demand for HBM has put non-HBM DRAM into tight supply.
- · The supply situation in DRAM is tight now, a combination of low CapEx from the downturn, shift in capacity to HBM and to DDR5, and lower fab utilization.
- · Their leading-edge capacity is running at full utilization.
- · Margin is expected to increase from the third to the fourth quarter of FY-24 at a similar step size to the increase from FQ2 to FQ3. Murphy didn’t say, but the best guess is he is referring to gross margin. That would be a further 650 basis point increase if he is indeed referring to non-GAAP gross margin.
Post Earnings Q&A call with Mark Murphy (CFO), Sumit Sadana (CBO) and Manish Bhatia (VP of Global Operations)
- · Sadana started his first answer by saying the company targets steady market share in both DRAM and NAND, a signal to the other manufacturers that Micron will not seek to gain share in the current strong market. He also said “disciplined capex” twice.
- · The company’s HBM3E is built on their 1-beta node, which reached maturity a year ago. It is a better product since it is on a mature node.
- · The CFO said they expect margins to increase through FY-24 and through FY-25. They can’t possibly know what will happen 5-6 quarters out, but a statement like this indicates how strong they believe the market recovery will be.
- · From FQ3 to FQ4, they expect DRAM bits to be up “slightly” and NAND bits to be up “somewhat.” For the remainder of calendar 2024 after that (September and on), they expect flatter bit shipments with a somewhat higher mix of NAND.
- · In the fourth quarter (this must mean fiscal quarter), the company is forecasting gross margin above 30%.
- · Outside of HBM, PC has been the first segment to recover in DRAM. Data center was expected to be the last to recover, and in the last few weeks they are indeed seeing upside from their data center customers.
- · They have “far more” DDR5 demand than supply. Strength in data center SSDs is also notable.
- · Sadana said again that, when they do add capacity, it takes 14 months for the wafers to come online. That is 9 months of lead time on the tool delivery and 5 months through the front end and back end of manufacturing. I think this is somewhat longer than what would actually happen, specifically in tool delivery, but the point stands. It will take upwards of a year for capacity addition decisions to flow through to become wafer output.
- · When asked about capital returns, the CFO said the top priority is their balance sheet. He said he is really happy with the state of the balance sheet, but that can’t be true with the amount of debt they are carrying. I think we’ll see them reduce debt significantly before they meaningfully increase the dividend.
- · Demand for traditional (non-AI) servers were weak for the first couple of months of calendar 2025, but that has shifted in the last few weeks, with higher demand seen. The company expects 2025 (he didn’t specify fiscal or calendar) to be the strongest year for traditional data center in several years.
- · Today, the HBM die is margin accretive to DRAM, so the higher cost is more than covered by better pricing.
Summary
After analyzing last quarter’s results, I felt worse about the market. This time, I feel better, much better. The stock price reacted the following day with the strongest rise I can remember. The enthusiasm around Micron’s stock is a combination of the memory cycle turning, high demand for AI compute capacity, and equity markets at all-time highs. A year ago, I could see the stock reaching $110 this cycle, but not this early. This is only the second quarter of rising ASPs in DRAM and NAND. It feels like the cycle is further along than that. What is most encouraging for memory investors is that the company is forecasting bit growth in both DRAM and NAND to be almost flat for the rest of 2024. ASPs are already rising sharply. They were up almost 20% for DRAM this quarter and over 30% for NAND. They see bit demand in calendar 2024 to be up mid-teens for both DRAM and NAND. It is March. That means there are three more quarters coming where demand is up around 15% and bit supply is flat. Pricing will rise sharply in that environment, assuming other memory suppliers don’t increase their bit supply. With Micron’s capital expenditures at downturn levels through at least August, plus the lead time from the decision to add capacity to new wafers being at least a year, Micron won’t significantly increase their production of bits until at least the spring of 2025. If Samsung, Hynix, and Kioxia/WD have similar plans to what Micron laid out, the memory industry is entering a period of severe undersupply that will last through at least the end of calendar 2024. That is good because the industry is still far from healthy. Micron’s operating margin is around break-even this quarter, following losses totaling more than $8B over the last two years. The memory makers need to hold the line on adding capacity. Otherwise, the through-cycle finances don’t support investments in R&D and capacity. Customers need to be screaming for more bits for at least three quarters before Micron and its peers add more wafers. Pricing needs to rise 50% from here. If that happens, it will support more investment. It will also lead to record results for Micron. In a scenario where memory prices increase by 50% over the next six quarters and bit shipments rise 10%, Micron will have quarterly revenue approaching $11B. If gross margin peaks in the mid 50% range, between the peak of the last two cycles, Micron would have quarterly earnings around $5 per share. That is over $5B of net income. It seems an absurd possibility, but things are aligned for that to happen. The severity of a downturn typically determines the strength of the upturn that follows. The 2022 – 23 downturn was historically bad. Capacity investment has been low, AI is a new source of demand, and many economies around the world are healthy. When looking forward at this cycle in early 2023, I thought the high case was $125 per share for Micron. Now, $150 per share doesn’t seem farfetched.
- Smooth Hughes (cyclical long MU)