Q4 and FY-22 Earnings Release and Analyst Call

9/29/22

Press Release and Investor Presentation Commentary

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

  • Capital expenditures on wafer fab equipment will be reduced by 50% in FY-23 compared to FY-22. Micron’s fiscal year runs from September to September. As WFE reductions take a quarter or two to flow through to supply, this change will lower Micron’s bit output early in calendar 2023. They certainly already started delaying equipment delays in the fourth fiscal quarter (Summer) of 2022.
  • Capital expenditures were $3.58B in the FQ4-22 and $11.98B for all of FY-22
  • In FQ4, the company spent $784M on share buybacks at an average price of $59.40 per share. For the full fiscal year, they used $3.21B for share repurchases at an average price of $68.64. Management was too early with this $3.21B in buybacks, another example of Mehrotra’s poor capital allocation skills. He and his CBO came from the NAND industry and still don’t seem to understand the nature of the DRAM cycles. A positive is at least they have switched from mechanical buybacks to a mix of mechanical and opportunistic. The company will continue to be held back in this area by the CBO’s pessimistic disposition, likely keeping management from taking smart risk with larger buybacks during this part of the cycle when the share price is most depressed.
  • Guidance for FQ1-23 is revenue of $4.25B +/-$250M, GM of 25.0% +/- 2.0%, operating expenses of $1.09B +/-$25M, and diluted EPS of ($0.09) +/- $0.10
  • Data center revenue fell Q/Q and Y/Y driven primarily by ASPs, though the industry continues to face some supply constraints that are limiting server builds
  • PC client revenue fell Q/Q and Y/Y driven by PC unit sales declines and customers reducing inventory
  • Mobile revenue fell Q/Q and Y/Y though the last two fiscal years, FY-21 and FY-22, were records for Micron’s mobile business. That doesn’t say much other than the company kept their market share, because the overall mobile memory market grew over those two years.
  • The automotive industry grew 30% Y/Y
  • Industrial is now softening, falling Q/Q and Y/Y on soft macroeconomic conditions
  • The pricing environment for both DRAM and NAND is “extremely aggressive.” They must be getting killed on ASPs to use such strong language.
  • DRAM bit growth: CY-22 industry demand is expected to be low-to-mid-single digit percent. In CY-23, they expect supply growth to be “well below” demand growth. They are modeling mid-single digit percentage growth in DRAM industry supply in 2023. Management revised down their view of long-term bit demand growth to mid-teens percent.
  • NAND bit growth: CY-22 industry demand expected to grow slightly higher than 10%. Supply growth in CY-23 is expected to be below demand growth. Management didn’t use nearly as strong language for NAND as for DRAM. I believe this is a reflection of what other NAND producers have said, about being more likely to risk oversupplying the market by outgrowing industry bit growth rates. Long-term bit demand growth will be around 28%.
  • Supply growth is expected to be “significantly above” demand growth in CY-22 for both DRAM and NAND
  • They expect CY-23 demand growth to be closer to long-term demand growth for both DRAM and NAND. But this far out the executives at Micron or any other company don’t know anything. This is also an almost meaningless statement. It only says demand growth will be better in CY-23 than in the dismal CY-22.
  • The company is selectively reducing factory utilization in both DRAM and NAND, to reduce bit supply
  • DRAM was 72% of total revenue in the quarter and 73% of revenue in FY-22. Bit shipments were down 10% Q/Q and ASPs declined low-teens percent Q/Q.
  • NAND was 25% of total revenue in both the quarter and the year. Bit shipments were down low-20s percent Q/Q. ASPs declined mid-to-high single digits percent Q/Q.

Financial Statements

Statements of Operations

Revenue in the quarter was $6.64B, a sequential decline of 23%. Gross margin was 39.5%, a 720 bps drop from last quarter. Of the financial figures available, I judge the memory cycle by Micron’s gross margin. In the upturn of 2016 to 2018, GM rose for eight quarters before falling for the next six. In the next cycle, GM rose for nine quarters and reached a peak of 61%, a level unheard of in more than two decades for Micron. Some consider this to have been a super cycle, caused by a large shift up in memory demand to build data centers. This cycle peaked in Summer of 2018 and gross margin fell for five quarters before starting to rise at the end of 2019. The next cycle was scrambled by the Pandemic. From the cyclical bottom in the December 2019 quarter, gross margin percent rose for three quarters, fell for one, then rose for three more before staying between 47% and 48% until the early Summer of 2022. Long time readers of my posts know I base my timing of buying and selling Micron stock on the duration of the ups and downs. I assume upturns will last eight quarters and downturns will last six. This worked well until the Pandemic hit in early 2020. With that exogenous shock to demand, I declared force majeure on my model. I didn’t know when to sell but thought that the GM% would get above 50% before the end of the cycle. I was wrong. I would have been better off sticking with my timing and forgetting where I thought GM would peak. Had I done that and assumed the upcycle would last eight quarters, I would have sold in the December 2021 to February 2022 range, as that was a full eight quarters after the bottom. I didn’t, so here I am wondering what to believe. Do I ignore the Pandemic and stick with my eight-up-six-down strategy? If I do, then what was the peak? The stock went down-up-down at the top. If I count eight up from the bottom, then the peak was FQ1-22. Six quarters from there is FQ3-23, which starts in March of 2023. Mark Murphy, the CFO, said that is what they are modeling internally, for the market to begin recovering in the second half of their fiscal year 2023, which would be March through August of next year. R&D spending in Q4 was $839M, a quarterly record. SG&A was also a quarterly record in Q4. This surprises me as the current management team has controlled expenses better than ever. The downturn was clearly underway for all of the last quarter, so it is surprising they didn’t ratchet operating expenses down more. It may be just having to turn a big ship, or it could be they are feeling pressure to spend on labor to be competitive in the market. They guided OpEx to be down in Q1, so it is probably more the former than the latter. Operating income was $1.52B, nearly a 50% drop from the third quarter. That is an operating margin of 23.4%, down 1140 bps sequentially. Higher interest rates almost tripled interest income. Since their debt is all fixed-rate, interest expense was almost flat. Micron, because of their strong balance sheet and fixed-rate debt, can start making money from the interest rate spread. Before tax income was $1.55B, out of which a mere 3.6% was provisioned for income taxes. This is much lower than their normal low-teens rate, so earnings will be artificially large this quarter. Net income was $1.49B, or $1.35 per diluted share on 1.106B shares.

Cash Flows

For the full year, Micron had $15.18B in operating income. This included $7.12B of depreciation and amortization. Compare that with the $12.07B in capital expenditures the company made over the same period. That is a stunning amount of investment. This certainly contributed to the overcapacity the market is seeing today. It is also driven, in part, by how much more expensive it is getting each node to add bit growth. With the market softening in the second half of the fiscal year, $2.18B of cash was invested in inventories. There was also $514M of stock-based compensation in the year. Starting with operating cash flow before changes to working capital, the company brought in $16.4B in cash flow during FY-22. From this, they invested $12.1B in capital expenditures, nearly three-quarters of the cash flow from a year that contained the peak of the upturn. That fact right there – that 75% of the cash flow from Micron’s best year needs to be reinvested in capital expenditures – is reason enough that MU is only good as a cyclical investment and not a buy-and-hold. They shifted a net $155M into available-for-sale securities. They also brought in $888M from selling their Lehi, Utah fab to Texas Instruments. The company also added $115B in cash from government incentives. Of the $5.15B remaining, the company used $2.55B to buy back stock, $2.03B to retire debt, $461M in dividend payments, and $352M in other items, including pre-payments for EUV scanners. The company also issued $2.0B in new debt, so the debt retirement was really a rollover to loans with more favorable terms. All this means, in the peak of the upturn, Micron added $510M in cash to their balance sheet. Almost five times this amount went to share buybacks. I’ll review below how effective that has been for shareholders. My owner’s cash flow will be a range, including and excluding stock-based compensation. In both cases, I will take out the effects of changing working capital, because so much went into inventory this year as the memory cycle turned. The high end of the range is $4.36B, or $4.33 per share. The low end is $3.48 per share. At a share price of $54.62 per share, these are owner’s cash flow yields of 6.4% to 7.9%.

The Balance Sheet

The Ben Graham (what?!) net-net value of Micron is $5.4B, subtracting all liabilities from current assets. That is $4.89 per share. Until Micron is down below $10, we can’t even start talking about it as a classic value stock. The company’s full book value is $49.9B, or $45.12 per share. 77.5% of this book value is PP&E. Leading edge semiconductors are such a capital-intensive business. The company has $9.33B of cash and short-term investments versus $6.9B in debt. I would like to see them be bolder with the share buy-backs given the strength of their balance sheet, though the executives already missed their big chance by using up so much cash repurchasing shares too early in the down-cycle. Coming back to inventories, this is an item to watch closely over the coming two quarters. The $6.66B of inventory the company is holding today is by far the most they have ever carried, by more than a billion dollars. The first three quarters of COVID-19 were the previous record, when inventory was around $5.5B for three quarters. It declined to around $4.5B as the cycle peaked, reaching a condition the previous CFO characterized as “lean”. There are two things I’m watching for here. First, will Micron have to take an inventory write-down? That would happen if the market price they are likely to get for the inventory falls below their cost to make it. Second, will they make good on their public statements to hold inventory rather than sell into a weak market and make it even weaker. Their end of quarter inventory, using their guided Q1 revenue, is 140 days. The CFO said in the conference call he expects inventory to grow to above 150 days in the first quarter.

Share Repurchases

In this earnings release, management called out that from FY-19 to FY-22, the company bought back 121 million shares at a cost of $6.5B. This is an average per-share cost of $53.72. I checked the share count over this time and it did indeed decline by 121 million shares, thus the buyback figure they gave is net of dilution. Since the Elpida purchase in 2013, a transaction that included significant new shares, the low point of Micron’s share count was 1,036 million, in 2016. In the upturn of 2017 – 2018, share count peaked at 1,238 million. This increase of about 200 million shares happened as holders of convertible debt exercised those converts into new stock. That upturn also allowed Micron to clean up and strengthen their balance sheet, including the retirement of all remaining convertible debt. With no convertible debt and a strong cash position, Micron is able to weather this and future downturns without issuing new shares. That will increase how well share buybacks will bring down share count. The meaningful stock-based compensation the company issues to executive and other employees causes dilution every quarter. If management times their buybacks well, they can significantly reduce the share count over time. The measure of the executive team’s performance with buybacks changes with the stock price, so I will continue to check in on this regularly. The count of 1,129 million shares at the end of FQ3-20 is a good starting point to use to measure the new management team as that is when the converts were fully cleaned up from the balance sheet.

Conference Call

Numbers given during the conference call are non-GAAP.

Sanjay Mehrotra (President and CEO)

  • They are reducing FY-23 capital expenditures as well as lowering utilization in their factories
  • Industry-leading 232-layer NAND was introduced in the fourth quarter. The company is on track to begin ramping their 1-best DRAM node in manufacturing by the end of calendar 2022.
  • Their investment in the Boise site manufacturing fab will be $15B from 2023 through 2030. They will soon announce a second high-volume U.S. DRAM manufacturing site. It doesn’t make sense to do two U.S. sites instead of one from the perspective of cost, since it reduces the advantages of scale. That means it is likely they are doing this to get more grant money from the U.S. government than they could get with a single site. That, or maximizing a combination of state and federal grants.
  • While end-demand for data centers is strong, cloud and enterprise customers have lower memory demand because of supply chain constraints and inventory reductions in response to macroeconomic uncertainty
  • They lowered their unit growth forecast for PCs down again, to a mid-teens decline. The mobile market also declined both sequentially and year-over-year. Automotive demand continues to be strong while industrial is softening similar to PC and mobile markets. Autos and data centers are good, rest is bad. I think autos will roll over in the next two to three quarters as semiconductors finally catch up, and some demand declines.
  • The pricing environment is “extremely aggressive” right now. They are getting hammered by their customers on pricing.
  • CY 2022 industry bit demand growth for DRAM will be low-to-mid single digits percent and for NAND it will be slightly higher than 10%. They expect supply growth to be significantly above demand growth in CY-22, contributing to very high inventories in DRAM and NAND
  • For CY-23, the expect demand growth to be closer to long term growth trends. This just says it won’t keep being completely horrible.
  • In CY-23, the expect industry DRAM supply to grow well below demand growth. They are modeling mid-single digit percentage growth in DRAM supply in 2023. NAND supply will also be below demand in 2023.
  • Long-term DRAM bit growth will be mid-teens percent, a slight reduction in their previous view. NAND will grow 28% annually over the long term.
  • They are reducing CapEx to $8B for FY-23, down over 30% year-over-year. Construction CapEx will double year-over-year while WFE CapEx will decline nearly 50% as they slow ramp rates on new nodes.
  • They are selectively reducing output at select factories for both DRAM and NAND

Mark Murphy (CFO)

  • DRAM revenue was up 12% year over year while NAND revenue was up 11%
  • Year-over-year change for all BUs was down. CNBU was off 25%, Mobile down 23%, Storage down 34%, and embedded down 9%. The memory market is terrible in all segments. The pain only varies in degree.
  • Gross margin declines in the quarter were caused by declining ASPs
  • It is comical that a company with the level of capital expenditures that Micron has includes EBITDA revenue and margins in their presentation. Yet they do.
  • The company has $13.6B of liquidity as of the end of fiscal year 2022. This is down from $14.5B ninety days ago.
  • Average share price was $59.39 for buybacks in Q4, expending $784M to buy back 13.2M shares
  • Average days of inventory in the quarter was 139 days, reflecting how bad the market is right now
  • Macroeconomic uncertainty is high. They expect bit shipments and pricing to decline in both DRAM and NAND in Q1. The second fiscal quarter is expected to have similar financial performance as is forecasted in Q1 but bit shipments will weakly start to recover.
  • They are forecasting both bit shipments and revenue to recover in the second half of the year. Days of inventory are also expected to recover in the second half of FY-23.
  • CapEx is expected to be front-loaded in the year. Free cash flow is forecasted to be negative $1.5B in FQ1-23
  • Their tax rate will be elevated in fiscal year 2023, because of coming changes to how R&D expenditures are handled. Their taxes will be at least $300M higher because of this. This will lead to a tax rate in the low-to-mid teens percent range.
  • FQ1-22 non-GAAP guidance: $4.25B +/-$250M, GM of 26.0% +/-2%, operating expenses of $1B ± $25M, and EPS of $0.04 +/- $0.10. They forecast revenue to be similar in Q1 and Q2 of FY-23. They believe revenue and volumes will recover in the second half of FY-23, which is the period from March to August, inclusive.

The lousy state of the memory market was nice for the executive comments, because Sanjay didn’t go on and on about worthless product updates, because he was busy explaining why their financial results and forecast is so terrible. They went on for 35 minutes in prepared remarks, but the comments were more useful they typically are.

Analyst Q&A

  • Utilization cuts are in the mid-single digits range on products that they have the highest inventory on. Gross margin impact from these cuts will be later in FY-23, after Q1 and will be a two-point hit to gross margin.
  • They will opportunistically repurchase shares even with negative free cash flow. Free cash flow is expected (hoped) to recover in the second half of FY-23.
  • They expect DRAM demand growth to be mid-teens percent in CY-23
  • Second half of the year bit growth should be up year over year. He didn’t say if this is calendar or fiscal, but I think he meant fiscal.
  • Management still expects CapEx intensity, cross-cycle, to be in the mid-30s percent of revenue
  • They expect inventory to increase to over 150 days in the first quarter of fiscal 2023. A healthy level is 100 to 110 days for them.
  • In 2023, Micron expects PC unit demand to be flat year over year. They are modeling a rebound in mobile phone unit sales next calendar year, as China opens back up.
  • Sanjay said they plan to shift their bit supply to market segments that are most profitable. They also don’t intend to give up market share with their planned supply reduction. This is a signal to the other companies that they also need to cut supply to bring balance back to the market.

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

  • The company expects to return to free cash flow generation in the second half. Murphy declined to comment on Q2 FCF.
  • They are expecting a “robust recovery” in the second half of FY-23. Right now, they are shipping well below end demand because customers are filling their needs with internal inventory. When this adjustment is completed, it will lead to higher demand from memory makers.
  • There are portions of cloud demand that are weaker on macro issues and others are strong. It nets out to “healthy” cloud demand. These customers do have high inventory.
  • OpEx will be lower in FQ2 than the level guided in FQ1 and will continue to reduce as the year goes on
  • Micron’s tax structure is optimized for higher profitability. The downturn in the market will raise their tax rate, that is what is inferred from this comment. The new law to amortize R&D will be “several points” of higher taxes.
  • Because PC and mobile demand weakened first in the down cycle, they expect these segments to return to health first as well. One assumption in this is that the Chinese government will improve in the coming quarters and that will increase consumer demand.
  • They expect the customer inventory correction to be mostly completed by the end of their second fiscal quarter, in February of 2023
  • Inventory will be “well over” 150 days in the second quarter, which is the level they will reach in the first quarter
  • With their current forecasts, the company is not predicting they will need to take a write down on their inventory this year
  • The WFE reductions are on both DRAM and NAND. Their construction spending will be higher in FY-23 than it was in FY-22. This is lumpy spending and FY-22 was a lower year.
  • The crossover point for DDR5 DRAM, which has a larger die size than a similar density DDR4 and thus will lower bit growth (all else equal), has pushed out as the processor vendor timelines have move further into the future
  • Their long-term projections for cost declines are high-single-digit percent annually for DRAM and low-teens-percent for NAND. In FY-22, the company exceeded both of these figures for actual cost reductions. The costs referenced are per bit. In FY-23, they expect both cost reductions to be below the long-term model, because of lower CapEx, reduced utilization, and a couple of local issues (power and ForEx).
  • Sadana and Bhatia were both excellent when answering their questions
  • End demand in automotive remains strong. Excepting that, all segments have been impacted by slowing demand. For customers, the mindset has shifted from keeping semiconductor inventory to protecting free cash flow.
  • Within the server segment, enterprise (on prem) has been weaker than cloud
  • YMTC supply has contributed to the current imbalance in the NAND market

Summary

This downturn has been swift and severe. Gross margins peaked at 47.9% in the fourth fiscal quarter a year ago, were sideways for two quarters, then started down in the Spring of 2022. Since the second fiscal quarter of FY-22, ended in early March, gross margin went from 47.8% to 40.3% and is guided down to 26.0% in FQ1-23, which will end in early December. The executives said financial performance will be similar in Q2 as the Q1 guidance before recovering in Q3. Management doesn’t know more than a quarter out what is going to happen, though they do benefit from industry experience and insights into customer demand that aren’t available to the wider market. That said, cyclical investors in Micron can’t bet the cycle fully based on management or analyst commentary. I don’t know when the cycle will turn up but historically when supply is pulled down as quickly as Micron is reducing their bit growth, it works to bring the market back into balance. I will be watching Samsung, Hynix, and Nanya closely in the next couple weeks for their DRAM capacity comments. My best prediction for the DRAM market is it will turn over in the Summer of 2023. Capacity growth will come down sharply in the next two quarters. The wild card, and there is always a wild card, is if the macroeconomic environment gets worse and keeps bringing demand down, preventing supply moderations from getting the market back into balance.

S. Hughes (long MU)

1 Like