Micron thoughts 2.5 weeks out of earnings

Investment Analysis Clubs Saul’s Investing Discussions
1

I decided to do a deep dive into Micron because its showing considerable similarities to stocks that I have lost money on holding into earnings. The similarities is large amounts of non-centralized data and downward stock price. So I went and looked at the data and below is a collection of the data as well as my thoughts.

Micron sells two main categories of goods with DRAM Products being approximately 69.5% of total revenues and NAND products being approximately 30.5%. DRAM includes the hot new AI technology called High Bandwidth Memory (HBM). Micron has about 4-6% of the market share of HBM. Looking at HBM market predictions for 2025 range from 21B to 46.7B. Micron has targeted a 20-25% market share of HBM in 2025, which could be achieved by their 8-Hi HBM3E chip being utilized in the H200 Nvidia system. Micron also has recently announced 12-Hi HBM3E chip that is compatible with B100/B200 Nvidia system. If Micron can achieve 20% market share in 2025 that would be 4B to 9B in revenue for HBM alone in 2025. Compared to their several hundred million in HBM sales in 2024. The companies total revenue was 25B in 2024, so a 4-9B increase in one product would be a substantial gain for the company.

Micron’s own prediction of Q1 2025 has revenue at 8.7B which would be 12% gain QoQ. Micron also predicted it would have EPS of approximately 1.74 dollars which would be 217% gain QoQ. Micron has hit their revenue estimates 8/8 for the last 8 quarters. Micron has hit their EPS estimate 7/8 for the last 8 quarters. Also with Micron selling out of their production before the quarter starts lets me assume they some predictive abilities of what this quarter will hold.

With all this positive upside why then is Micron so cheap today? To gain market share of HBM, Micron is spending billions of dollars in capex. Spending 8B in 2024 and targeting to spend 13.5B in 2025. It will also have to cannibalize its DRAM production capacity to make HBM, which requires three times as many wafers as traditional DRAM products. DRAM and NAND prices have declined in recent months after Micron’s last conference call. Which is what the downgrading analysts have cited as their reason for their downgrade.

Prices of DRAM and NAND are public knowledge and have not gone in the direction that Micron predicted in their last conference call. With the biggest movers being DDR3 and cheaper DDR4. Those have been rapidly declining in price as the demand for them has stagnated as new consumables shift to the newer DDR5 standard. Micron predicted a slight increase in prices in the DRAM sector mostly because of limited wafers and manufactures shifting production to the more profitable HBM, which also require more wafers than DDR4 and DDR5. During the conference call Micron, said that they would be shifting away from low margin items to higher margin items like HBM and DDR5.

I think that the lowering prices of DRAM and NAND is just noise and takes away from the AI growth play of Micron. Micron has the most energy efficient HMB3E and in their conference call they said that their product sells at a premium compared to others. From my research the dominate provider of HMB3E is SK hynix, followed by Micron and Samsung taking up the rear. Samsung has had issues with getting certifications from Nvidia. Another recent down pressure on Micron is the recent statement from NVIDIA CEO that NVIDIA is “working as fast as it can” to certify Samsung’s New AI memory chip. This was said on the 23rd of November. Which helped MU stock price drop 6.3% last week.

Micron is aiming for 20-25% of the HBM market share which has been their traditional share of the NAND and DRAM when split with SK hynix and Samsung. Micron has already allocated majority of its HBM3E for the FY 2025. Demand for HBM3E is higher than production so even with Samsung finally getting orders from Nvidia won’t change the opportunity for Micron.

I believe there is a large upside for investing in MU. Especially if they can execute and deliver over 200% QoQ in EPS. But I’m interested in other opinions.

Drew

59 Likes
2

I am seeing the same potential upside for them as the current P/E is 141 but it includes this big trailing negative quarter. The last four quarters of EPS are,

-1.12 → 0.71 → 0.30 → 0.79, totalling 0.68

The last earnings press release projected EPS to be “$1.54 ± $0.08”, or 1.60 as the top of their guide, which would effectively be swapping the -1.12 quarter for 1.60, so the last four quarters would be,

0.71 → 0.30 → 0.79 → 1.60, totalling 3.4

Just landing at $3.4 over the last year will give them a P/E of 28, and analysts are projecting the forward P/E or next 12 months of EPS to land at a P/E of 11. The company is looking quite cheap to me as revenue is getting close to crossing over a 10B quarterly run-rate soon if the business goes according to plan.

28 Likes
3

wpr,

You need to be careful looking at MU in the lens of TTM p/e ratio. It is a boom/bust cyclical stock and in the past the only way to make money on it is buying when the P/E looks high and selling when it is low <10x. I’m not saying MU won’t do well from here with the explosion in AI memory (HBM demand).

The stock has been under pressure because past memory up-cycles have lasted ~8 Q’s and we are >4Q’s into this one. This one may be longer… But low end memory prices have been falling on the spot market with PC and mobile end markets continuing to be weak and inventories rising. DDR5 and HBM pricing is holding up well, but this needs a close eye and is not a buy and hold stock unless you want to ride out the cycles. Which is not fun (I know from experience). The good thing is Samsung is struggling a lot with HBM and same higher end DDR5 so I think they will be more careful with not oversupplying the market than they have in the past (the last down cycle they lost ~$10B)

Bnh

30 Likes
4

@bnh91 My thesis around Micron relies on the company transitioning to a non-cyclical stock. They follow the same pattern as semiconductors from having a difficult 2023, followed by an acceleration in growth.

There are a number of catalysts such as huge AI demand, but some lesser appreciated ones such as the next generation of PCs and smart phones spec’d out for using double the memory to support local AI on the device.

The High Bandwidth Memory they sell is higher margin than the other tech and they are selling more of that which is why the gross margin is ramping back up. Additionally, they’d said the TAM for HBM was 4B this year, and it is 25B for next year. Micron is already sold out on HBM through 2025, so the revenue is more predictable for this segment of the business.

With regards to valuation, I typically use EPS as a guideline for companies that have been profitable for awhile. However, if we want to look at a revenue basis, Micron is relatively cheap on P/S as well,

P/S
Micron 4.2
Broadcom 9.7
Nvidia 30.4

I could see Micron getting up closer to a ~10 P/S or so and be a reasonable valuation. Broadcom grew revenue at 47% last quarter yoy and Micron is at 93% growth yoy, and both companies are subject to similar cycles on computer hardware.

Here’s a number of things the CEO said last earnings which indicate to me this company is not following a cyclical pattern,

We are entering fiscal 2025 with the strongest competitive positioning in Micron’s history. We have leadership 1-beta DRAM and G8 and G9 NAND process technology and leadership products across our end markets. Robust data center demand is exceeding our leading-edge node supply and is driving overall healthy supply/demand dynamics. As we move through calendar 2025, we expect a broadening of demand drivers, complementing strong demand in the data center. We are making investments to support AI-driven demand, and our manufacturing network is well positioned to execute on these opportunities. We look forward to delivering a substantial revenue record with significantly improved profitability in fiscal 2025, beginning with our guidance for record quarterly revenue in fiscal Q1.

Multiple vectors will drive AI memory demand over the coming years, growing model sizes and input token requirements, multi-modality, multi-agent solutions, continuous training and the proliferation of inference workloads from cloud to the edge. Micron is focused on translating the opportunities from AI demand into value capture for all our stakeholders. Demand from data center customers continues to be strong and customer inventory levels are healthy.

We expect the HBM TAM to grow from approximately $4 billion in calendar 2023 to over $25 billion in calendar 2025.

As we have said before, our HBM is sold out for calendar 2024 and 2025 with pricing already determined for this time frame. In calendar 2025 and 2026, we will have a more diversified HBM revenue profile as we have won business across a broad range of HBM customers with our industry-leading HBM3E solution.

The PC market is in the early stages of our transformation, and we expect a significant shift towards AI-driven functionalities that promise to enhance user experiences and productivity. AI PC require a higher capacity of memory and storage. As an example, leading PC OEMs have recently announced AI-enabled PCs with a minimum of 16 gigabytes of DRAM for the value segment and between 32 to 64 gigabyte for the mid- and premium segments versus an average content across all PCs of around 12 gigabyte last year.

With the advent of AI, we are in the most exciting period that I have seen for memory and storage in my career.

And frankly, our HBM3E product is getting premium in the industry as well versus other products. So it just puts us on a good trajectory ahead as well.

27 Likes
5

Well, I proclaimed in about 1998 on the Semiconductor boards of the MF (on AOL as the then MF board moderator) that semiconductors were no longer cyclical! Boy was I wrong! 26 years later, they are always going to be cyclical.

There’s a couple factors working here. Semiconductors have always been viewed as “of national interest”. So we saw Japan subsidize wildly their manufacturing. They dominated Memory in the 1990s (Fujitsu, Hitachi, Toshiba, NEC). Micron found it’s foothold and others dropped away.

Later the South Koreans jumped in and became dominate as well (Samsung, SK Hynix) etc.

China was next but their struggles start and end with the Semiconductor Equipment. They have been stealing the ASML secrets for awhile, so soon enough, they will have their own Semi Equipment infrastructure created. Probably. not this decade though.

Micron is now building 10’s of millions of SSD drives for the Enterprise. These are high value products and include their own controller chip they design (ASIC) built by TSM. They use Micron made Flash chips.
As the world moves more and more to SSDs, less integrated folks are hard pressed to compete. Seagate is basically out of SSD market now for example.

So I would say this is a complex environment. Fabricating chips is an expensive game in terms of capital investment. Micron has competed well against the state subsidized folks. There’s less of that these days with China being handcuffed. There’s less investment given the ROIC equation being challenging. (but we are seeing renewed “country” investment - see the Chips act)… So this will likely result in more capacity a few years from now.

So I tend to agree, Micron may not be as cyclical for a spell as history would suggest. Until they become cyclical again!

23 Likes
6

There really should never be a comparison of MU to BRCM or NVDA.

They sell products to the same customers I guess. But the market dynamics and competitive landscape has always been quite different.

We’ve had some recent innovations (namely, HBM) that make MU look more like a NVDA. But in short order, their competitors will have the identical product. “short order” is measured in months here. In fact, this is built-in. Before anyone would adopt HBM, there is an agreed to industry standard for the interface, form factors etc. To assure there will be multiple suppliers of the exact same thing. And HBM will be a good market for a time. Until supply can meet demand and then you will have several companies all selling the same thing to the same customers. That could be a few years out. But most investors know that day is coming.

To expand: Customer is Dell. They will qualify as many HBM suppliers as are qualifiable. When that list is big enough and supply is ample enough, prices drop. Each month, there is a dutch auction to fill Dell’s supply. Dell will (for each new product design) decide up front, who will supply the GPU. Today, that is NVDA. Could it be AMD at some point - yes. But that decision would be made and the price agreed to well in advance of revenue. So it’s competive, but once agreed to, it’s locked in for a time. Dell will be locked into that choice for a generation and so while that scenario is competitive, it’s not ever like Memory where each order every month is competitively exercised. HBM will get there in time.

26 Likes
7

This is different than what the company is saying on their last earnings call where they said,

  • “And we provided you some of the expectations we have of our industry leadership of that 12-high product with the 20% lower power consumption versus other 8-high products”
  • “Micron can have 20% lower power than an 8-high product from the nearest competitor”
  • HBM3E product is getting premium in the industry (they have pricing power on a superior product)

There really should never be a comparison of MU to BRCM or NVDA.

These are the only three large caps in the same sector and industry that are growing. The data provider I’m using lists all three of them as,

Industry: Semiconductor and Semiconductor Equipment
Sector: Information Technology

What is your definition of cyclical? Do you consider Nvidia cyclical?

The classical definition I’ve seen of cyclical is a business which is very heavily tied to macro economics, with the most typical industry be housing. With housing the interest rate changes, and the general health of the economy drive where the housing market goes.

For example, I don’t see Nvidia or Micron having an outsized impact because of interest rates, or consumer spending habits. Previously, when home computer purchases and phone purchases were drivers of revenue for semis back in the 1990s they were more cyclical.

There was a thread awhile back about Supermicro before its 10x+ run-up where I was pointing out Supermicro was growing fast and had a P/E of 10. Most of the replies I got where that the company is not worth investing in because it’s cyclical, and that it’s just a commodity product.

If you are saying we can’t compare P/E and P/S to companies in the same sector and industry as Micron, how do you suggest we determine if the market is offering a good price?

11 Likes
8

I have not studied Micron. But any differentiation they currently enjoy with 12-high vs 8-high product will be removed in short order if history is our guide.

There is lots of differentiation within the Semiconductor industry. Chip designers are not the same as Chip Fabricators which are not the same as Semiconductor Equipment suppliers. Even within the Chip Designer group, there is massive differences between a CPU/GPU provider and an RF Chip provider such as MaCom. I’d not use these as comparisons.

A company who’s profitability goes up and down in cycles. These cycles for a memory company are somewhat tied to economic cycles (Supply) but also to how many competitors are supplying how much product vs that supply. The memory guys all operate their own fabs and thus have huge capital investment cycles and then need to run these at full capacity. Under-utilized fabs are cash drains (see Intel). History has repeated EVERY time with massive over production.

I consider all Semiconductor companies cyclical. The equipment guys especially. The fab operators as well, including TSM, Intel, ADI, TXN, Wolfspeed…

The fabless guys don’t have the capital outlay and are a different model. Their cycles are tied to economic demand (how many PC’s are being sold drove GPU demand for example) as well as how good their products were vs competition. NVDA was very cyclical depending on 3D Graphics and later bitcoin-miners. But AI is not currently cyclical, it’s pedal-to-the-metal. And that will eventually have a down cycle and become a mature industry (IMO of course). But their lead is huge, so no, I don’t expect a cycle for NVDA until the race stakes are lowered a bit by their customers. Forever is a long time.

Micron should be compared to other Memory and HBM suppliers.

You wouldn’t compare Caterpillar and Kenworth to Tesla or Volkswagon.
You can compare NVDA to AMD, and perhaps BRCM, MRVL and ALAB. You can compare ADI to TXN, NXPI, ON and others.
You can compare ASML to KLA, AMAT etc.

Your data service information is not nearly granular enough…

24 Likes
9

In the case of a long-term B/H thesis, I’d be ok with comparing $MU to any other memory manufacturer, IF I were willing to make the assumption that other suppliers’ HBM tech would eventually meet/exceed $MU’s.

But on this board we want to know which companies are growing, rapidly, right now.

So if we are assessing a thesis that $MU is about to inflect, right now on HBM, wouldn’t we want to narrow comparisons to just HBM suppliers?

If we do, that only gives us:

  1. An OTC company: SK Hynix
  2. Samsung, Intel and Fujitsu, all of whom manufacture a lot more than just memory

That’s not a lot to work with in terms of apples-to-apples comparables.

13 Likes
10

The CEO is on record multiple times the past few earnings, and conferences saying they have a competitive advantage on power consumption with the HBM product versus the other memory makers. That is leading to pricing power on their product versus competitors. I’ve provided multiple direct quotes of the CEO making this point. A 20% difference is power consumption is not some insignificant detail when the AI systems are reaching thermal heating limits already.

EDIT: Adding from the Q&A they had back on August 28, regarding this competition topic,

Analyst:

Well, to dig more into that, can you talk about that decision to maintain share and potentially at the risk of not gaining share, especially as we think about the competitive environment in HBM. Like what is your willingness to potentially gain share in that market specifically versus maintain?

Sanjay Mehrotra:

So let me be very clear. In HBM, of course, we are gaining share, right? I mean, our share in HBM so far has been pretty low and going from a very low share to our shares lining up with our DRAM shares sometime in calendar year 2025 is a tremendous opportunity for us. We are excited with our strong product of HBM3E, which is industry’s best product when it comes to power that is 30% lower.

Here the CEO is saying the power consumption is even lower than the previous 20% cited. The CEO is literally saying let me be very clear, we are gaining share, and have the industry’s best product.

NVDA was very cyclical depending on 3D Graphics and later bitcoin-miners. But AI is not currently cyclical, it’s pedal-to-the-metal. And that will eventually have a down cycle and become a mature industry (IMO of course). But their lead is huge, so no, I don’t expect a cycle for NVDA until the race stakes are lowered a bit by their customers. Forever is a long time.

Here you are saying AI is not cyclical currently, which implies the companies getting the boost from AI are not cyclical. The widely accepted definition of a cyclical stock is a company that depends on heavily on macroeconomics and when you look this up the most commonly cited examples are: automotive, airlines, housing, construction, retail, and travel. If we are just saying a business is cyclical because it has ups and downs in profitability, then every company is cyclical.

Micron should be compared to other Memory and HBM suppliers.

You wouldn’t compare Caterpillar and Kenworth to Tesla or Volkswagon.

Catepillar/Kenworth and Tesla/Volkswagon are in different industries and sectors. The point you are making is more like saying Tesla shouldn’t be allowed to be compared to Volkswagon because Tesla makes electric cars and Volkswagon makes gas cars.

Historically this board has compared companies in the same industry even if they were not working on the exact same product. What you are writing is the equivalent of saying a company like Crowdstrike is only allowed to be compared to SentinelOne, and Palo Alto.

What I am saying is it can make sense to compare a CrowdStrike to other SaaS companies. We wouldn’t say you are not allowed to compare CrowdStrike to Snowflake, because one does information security and the other does data warehousing. They are in the same industry, so it’s at least some basis to come up with a rough valuation.

14 Likes
11

I’ve been clear:

Micron had a competitive advantage today. History has proven that in memory, those gaps close. It won’t last years.

Semiconductor business is cyclical. This cycle could elongate due to AI. Most semiconductor markets are in a down cycle and AI and Mil/Aero is up. But we are in a down cycle as an industry. If you are not selling into AI, you are down. This cycle will likely last awhile longer than most. See AI.

You mention a thesis that stated the segment is no longer cyclical. Then you later came back with that it’s not a long term buy/hold play.
I misunderstood because if you are not long term, why care about cycles?

10 Likes
12

Well, earning were last night, and while the quarter was mostly in line with expectations (which for me was below my expectations), they guided under the expectations for EPS, and in line for revenue.

While the stock is cheap (even so after the haircut), I am considering selling out here because I don’t see a clear growth trajectory anymore.

I will take my time however, because despite a great year so far (still over 70% YTD even after yesterday) I have made several mistakes in the past weeks (holding on to TMDX and ASPN, which are now so beaten down they really shouldn’t be sold).

Curious to hear other opinions.

14 Likes
13

I also sold my position this morning after the report. They guided almost exactly for where they landed,

revenue 8.7B → 8.71
adj GM 39.5% → 39.5%
adj EPS 1.74 → 1.79

They forecast for next quarter is significantly down though,

revenue 7.9B
adj GM 38.5%
adj EPS 1.43

This guide breaks their streak of them growing revenue, gross margin, and EPS sequentially for many quarters, as all three metrics are guided down for next quarter.

Long story short the consumer side of the business is reducing their inventory they stockpiled rather than purchasing new. It sounds like some of the companies overbought on supply previously and are churning through that inventory.

On the other side of the business things are going well, data center revenue grew 400% yoy and 40% sequentially, accounting for over 50% of total revenue. If I was more inclined to be a buy and hold investor, I’d be okay holding here. However, the sequential guides down was enough for me to sell.

23 Likes
14

Micron’s earnings beat my personal estimates. With the surge in data centers overcoming the headwinds that the rest of the memory market faced. I’m disappointed in the guidance with it down 10% QoQ, but its still guiding up 36% YoY.

Breakdown of business section performance and % of revenue.

  • Micron’s Data Center sales up 48% sequentially.
  • Storage Business is up 3%.
  • Embedded Business down 10%
  • Mobile Business are down 19%
  • Micron guides total Revenue 10% for Q2
    Picture1
    Picture11652×992 134 KB

The cyclic nature of the DRAM and NAND market has hit the downwards portion. If everything goes down at 20% except the data center sales the company will have a 10% drop in QoQ revenue. Portions of their business dropped almost 20% this quarter and are expected to do worse next quarter. So realistically most of Micron’s business will decline but the data center. With Micron guiding at 10% revenue decrease it shows they don’t have that much confidence in data centers growing faster than 10-20% QoQ.

With the current drop in price the P/E of Micron has dropped to 24.66 with 3.5 EPS for the TTM. If they hit their guidance for next quarter they would have a P/E of 20.55 and growth YoY of 36% during a downturn of their NAND and cheaper DRAM products. The following quarters they have the continued on ramp of their HBM3E and HBM4 as well as the recover of the NAND and DRAM markets.

Take aways from the Conference Call on HBM

  • Still selling HBM at a premium
  • Higher yields than expected
  • Faster production ramp of HBM than expected
  • Raising TAM of HBM for 2025 to 30B

Overall, I’m pleased with the HBM growth and the Data Center sales. I’m disappointed in their guidance. I decided to trim my position in MU because my confidence has been lowered to a low level position. I did not fully sell out because it still has the potential to 2x to 3x its Market Cap in the next two to three years if it can continue to execute in the HBM and data center market.

Drew

22 Likes
15

So, you sold because Micron is cyclical? Because the HBM stuff is right on track (see above dance).

5 Likes
16

I think it makes sense to compare MU to NVDA simply because people would invest in MU due to the AI investing theme ( same as NVDA). AI growth is driven NVDA’s top line growth and also MU’s top line growth.

But when we look at gross margin, operating margin, and FCF as a percentage of revenue MU lags so far behind NVDA that I’d have to ask the question: why would anyone invest in MU when they can invest in NVDA? I asked the same question about SMCI when folks were piling into that one.

Gross margin:
NVDA 75%
MU 37%

Operating margin:
NVDA 66%
MU 27.5%

FCF margin:
NVDA 47.9%
MU 1.3%

NVDA keeps sooooo many more of the dollars from the stuff it sells than MU does. That tells us that NVDA has pricing power, a much more differentiated product offering, and a just a better business model. MU’s products have more completion which is reflected in gross margins and market share metrics…look at these metrics for NVDA and MU.

GauchoRico

76 Likes
17

I don’t own MU but some folks might want to own it because it looks undervalued. 2 year Fwd EBITDA CAGR is 62%, while the NTM EV/EBITDA is 5.2.

This suggests a lot of potential upside over the next couple of years, especially if NAND and DRAM start to recover. I would definitely buy MU over wildly overvalued Axon or Tesla, which could easily drop 50% at any time, as they are story stocks, whose valuations are not based on the fundamentals.

Regards,

Jeff

9 Likes
18

Be careful using EBITDA for Semiconductors given the massive amount of capital expenditures to build and equip a fab facility. Massive.

Tax incentives also may play a short term catalyst role so you want to account for it.

17 Likes
19

Micron investors might be interested to hear Gavin Baker’s take on the best investment class for 2025:

(start at 1:14:05).

It’s High Bandwidth Memory (HBM), and only Micron and Hynix make it (Samsung is trying).

13 Likes