May 12, 2022
Sanjay Mehrotra (Chairman and CEO)
• Mehrotra talked about how much data the world is creating and how high the rate of this growth is. He cited specific examples in AI/ML, mobile phones, and automobiles. He often talks about these kinds of examples. They are nice anecdotes, but not useful as an investor. What is useful is the fact that there are a diverse set of end markets for memory. This increases stability of supply and demand.
• The memory market is estimated to be $161B in total revenue in 2021
• They are forecasting DRAM bit demand CAGR to be mid-to-high teens between 2021 and 2025. Data center will grow faster than the overall industry. NAND bit demand CAGR will be in the high-20% range over the same time. For overall memory, mobile and PC will grow slower than the overall industry, and data center will be faster. SSD penetration in PCs is now approaching 90% and total mobile units have stabilized worldwide.
• By 2030, DRAM + NAND will be ~$330B in annual revenue, more than 30% of total semiconductors. This is a slight increase in share from today.
• In the last eight years, average industry gross margin has increased from >20% in 2006 to 2013 to >40%, a 2000-bps improvement
• The bending of Moore’s Law is causing a natural reduction in bit supply growth. This is happening because each node adds fewer bits and investment is higher, so discipline to add supply is higher. Industry WFE spend is down to 30%, from a level of 57% at the peak of the 2011 to 2012 supply growth cycle.
• Rising industry complexity can be understood with five dimensions that make it difficult for competitors to enter. This is their why-China-can’t-succeed-in-memory slide. The five factors are capital intensity/scale required, IP and cumulative knowledge, breadth and complexity of product portfolio, importance of manufacturing and quality expertise, and customer relations. Regular readers of my posts know I am short China entering the memory industry. These reasons sum up nicely why I believe this.
• Micron believes their replacement value is $100B (manufacturing and R&D facilities plus net cash). There is >$20B of R&D investment (cumulative since FY10) and 50,000 lifetime patents.
• Micron has a total of 45,000 employees across 17 countries
• The company leads in process technology, both DRAM and NAND. It is important to note this is a statement on their leading node being the most advanced. Micron’s overall average process technology is not leading, they trail Samsung, which leads to Micron’s cost structure being behind Samsung.
• A new greenfield DRAM fab will need to be added by the end of the 2020s
• Included a review slide of the commitments from their last investor day, in 2018. They are now leading in technology, have more than 80% of NAND bits sold into high-value solutions, improved $9B in their operations cost structure, and returned 50% of free cash flow to investors. The high-value solutions is a squishy measure to check as an investor. The operations improvement is impossible to check, but the company’s gross margin improvement indicates they have made a lot of progress here but can’t confirm if it is $9B worth.
Sumit Sadana (CBO)
• Micron seeks to keep their overall bit share flat, to keep the DRAM and NAND markets in balance. They seek to gain share preferentially in the markets that are more stable, higher growth, and lower seasonality. They have #1 share in automotive and industrial, which are examples of these kinds of markets.
• The shift away from homogeneous data center compute-memory architectures (waning Intel dominance) will increase DRAM demand, as the processors need more memory because of changing workloads
• From now to 2030, the company believes total DRAM-NAND spending in data centers will grow at a 14% CAGR. This was presented on a slide that predicts bits shipped per server will 3x in NAND and 2x in DRAM from 2021 to 2025. For DRAM, this just means bits per server will grow at the same rate as the overall DRAM market, which implies the outsized DRAM bit demand growth from servers comes all from more units. This is inconsistent with his message that heterogeneous architectures will increase memory density unless he is talking just about NAND growth. The tripling of NAND bits per box in the next four years implies this growth will be higher than the overall industry.
• DDR5 DRAM will be over 50% of total server content by the end of calendar 2023
• They see a massive inflection in NAND applications in data centers, driven by the shrinking cost difference between SSDs and HDDs. From 2021 to 2025, NAND bit growth will triple. From 2025 to 2030, when this inflection kicks in, they foresee 5x growth in NAND bits consumed. Note there is one more year in the latter period, so the difference isn’t as large as 3x to 5x.
• Sadana had a slide on each memory segment. The message is that Micron is a leader in memory-based products, not just making commodity bits as was the case five years ago or more. This section also reinforces the message of how the markets that consume memory are becoming more diverse.
• Automotive memory revenue will grow at a 28% CAGR from 2021 to 2025. This is from a smaller base, however, relative to other segments.
• There were testimonies from several Micron customers sprinkled throughout the presentations. The Lenovo leader mispronounced Sumit’s name, which makes a statement of how important the relationship with Micron is to Lenovo.
• Micron has increased the share of their revenue that comes from long-term agreements from 10% in 2017 to more than 75% today. This is most valuable because of the insight it gives Micron into customer demand plans. They aren’t take-or-pay agreements, nor do they lock in pricing.
• Micron announced that they have entered their first forward pricing agreement with a DRAM customer. This agreement locks in pricing, on a declining slope for the customer. The price is set in such a way to pass through Micron’s cost reductions, such that Micron’s gross margin will be stable through the cycle. This first agreement with a “top ten” customer and is worth >$500M per year in revenue. They are targeting a similar level of pricing/gross margin to the through-cycle average. The total value to both parties is to be similar to what would be gained without the arrangement, but it will be predictable. This is an experimental level of revenue, $500M is less than 2% of Micron’s annual sales.
Scott DeBoer (EVP of Technology and Products)
• The company will ramp EUV on their 1-gamma node in CY24. They are last among the three DRAM makers because (they say) the tools and materials will be advanced sufficiently by this node. The 1-gamma node will use EUV on “multiple” layers.
• The 1-alpha node was 9-12 months ahead of competition, at introduction
• The 1-beta node is the first that will introduce high-performance CMOS on all products across the node
Manish Bhatia (EVP of Global Operations)
• They are reaching mature yield 30% faster on DRAM and 20% faster on NAND on the current process nodes as compared to the previous node. This means lower cost and faster ramps to volume production.
• Micron is the #1 company in semiconductors, and in the top 5% across all companies globally, as measured by Gartner’s culture of quality. This index measures employee commitment to quality and the systems they have available to ensure high quality products. Surveys are okay, their ranking as #1 or #2 in quality rankings from customers is more meaningful.
• Micron believes they are #2 today in DRAM front end cost, and #1 in NAND front end cost. They are targeting high single digit DRAM bit cost declines over the long term, and low to mid-teens reduction in NAND bit cost.
Mark Murphy (CFO)
• Micron’s gross margin is now 2% better than industry average, blended for both DRAM and NAND, up from being 13% below the industry average five years ago
• Revenue has steadily climbed for the last seven years, but the step up in the last 12 months is among the lowest over that period. Operating cash flow declined for the first time in the last year and free cash flow has dropped since peaking in FY20. Though FCF has declined recently, to $3.6B, it is still above the $2.9B in FY18, when the last DRAM pricing cycle peaked. The new management team has wrung a lot of cash flow improvements out of the company.
• Since FY18 (the last strong upturn year), Micron has made $70B in cash from operations before R&D investment. Of this $70B, $51B went into either R&D or capex. They increased their cash balance by $10B, and returned $8B to shareholders. If the next four years look like the last four, and in the future all the cash after investment were returned to shareholders, that would be $18B. Divided by 3.5 years, that is $5.14B per year in cash to owners. At the current share price in the mid-$50s, this is a 7.9% owner’s cash flow yield. Not bad, but not great, considering some of the other valuations available out there right now.
• Micron has a strong balance sheet and an investment grade rating from all three agencies
• Cross-cycle financial model: revenue CAGR in the high single digits, operating margin ~30% (>70% from DRAM), CapEx in the mid-30% as a portion of revenue, with >10% of revenue returned cash to shareholders. 10% of revenue is well under the model I roughed out above.
• They are increasing the dividend for the first time, a 15% raise, to $0.115 per share. This is still a meager amount, meant to qualify Micron stock for investors who only buy dividend-paying companies.
• At the point of this presentation, they had bought back $700M worth of stock, and said that will continue at the share price at the time, which was $10 to $15 higher than where it trades today. The recent share price decline happened after the 3rd quarter ended, but hopefully the management team will give an update on buybacks during the conference call on June 30th.
• Share buybacks will be a mix of programmatic and opportunistic
Analyst Q&A
• Sadana declined to speculate on how large a portion of total revenue their forward pricing agreements could reach in the future
• Little rapport between the Micron executives and analysts
• They expect the cycles to moderate further in the future
Summary
By far the most useful information of the day is the first ever long-term financial model provided by the company. It is not compelling for investors, promising to return >10% of revenue to shareholders through dividends and share repurchases. At the revenue growth rate given, this is in the range of $3.6B to $5.0B in owner’s cash flow per year. Since the last five years were $5.15B in cash available to owners, the company must believe the 2018 upturn was an anomaly. They are forecasting lower cash available to owners from a larger stream of revenue in what they characterize as a healthier industry with more moderate cycles. Most of the 2.5 hours of presentation time was spent describing details on products, technology, and manufacturing. While this is all interesting to hear about, the financial model provided is a reminder that all of that talk is about activities Micron has to execute just to keep up. This management team has done an excellent job of improving the company. They should be recognized for that. But I don’t see anything in the plans they laid out that will improve Micron much beyond where it is as an investment. As it always does, Micron’s stock price performance will come down to the vagaries of memory pricing. Investors can make (and lose) money by buying and selling the pricing cycle. At least for me, that has gotten much harder since COVID hit and disrupted several of the key variables in the machine. This investor day left me less optimistic for Micron’s share price getting to $100 or better in a future of moderating cycles. Dampening the cycles means less opportunity to make money for those wanting to invest in them. I am continuing to hold my shares, but I have a lower probability than I did a few months ago on Micron’s gross margins getting to the mid-50% level or better in the next year.
-S. Hughes (long MU)