Micron MU fiscal Q3 2026 earnings review

Investment Analysis Clubs Saul’s Investing Discussions
1

Here is my take on the Micron earnings from this evening. The headline for me is that “supply is structurally constrained”. The management detailed how both DRAM and NAND supply will remain tight beyond calendar 2027. Later they detailed how they see tightness continuing beyond 2027. They said there is “no line of sight”, as to “when memory supply will be able to catch up with increasing demand”.

As Micron is partnering with customers more on their roadmaps, they have now signed 16 SCAs or long term agreements, with most of these deals having 5-year terms. The company is measuring RPO in regards to these SCAs, and that RPO metric is “approximately 100B”.

Micron’s expansion efforts are timed well with some new fab capacity coming online in mid-2027 and through 2028. The company is expanding massively with 10B of CapEx guided for next quarter and they projected 40B+ CapEx for the upcoming year, saying they will spend 10B+ per quarter. From my perspective, I like this aggressive expansion plan Micron has laid out as it does not seem like the supply shortage is going away any time soon.

Possibly even more impressive are the quarterly numbers and guide. I mentioned on Reddit I was looking for at least 36B of revenue this quarter, and that number landed at 41.5B. That revenue number of 41.5B is +74% quarter over quarter, and +346% year over year.

For the next quarter guidance I said I would like to see over 45B guided for, and the company guided for 50B. Stunningly GAAP gross margin came in at 84.6% when they had guided for 81%, and they now project 86% margins next quarter.

It seems the company is firing on all cylinders with practically every financial measure hitting a record. The transcript reads incredibly well, and I found this to be an exceptionally strong report. Here were some more details from the transcript,

  • “Exceptional fiscal Q3”
  • Data center revenue exceeded 25B, run-rate over 100B
  • SSD revenue 5B+, more than doubling sequentially
  • Now signed 16 strategic customer agreements SCAs, fundamentally transform our business model
  • Smartphones, high end PCs, new consumer devices also driving demand with AI enabled features
  • “We currently do not have line of sight as to when memory supply will be able to catch up with increasing demand”
  • Memory process technology is getting more complex with every new node
  • “Supply is structurally constrained”, “Despite our comprehensive efforts to increase supply”
  • SCAs accelerate partnerships in technology and innovation, providing customers with supply assurance
  • Working with automotive on shorter term 3-year contracts, “commitment to the important sector”
  • Take or pay agreements, binding commitments (later detailed how customers give them unrestricted cash)
  • SCA has “price bands”, “floor price enables a very robust gross margin for Micron, well above our peak quarterly margins in any past cycle”
  • 1-gamma DRAM node, G9 NAND node both ramping well and on track to become highest volume nodes in Micron’s history, volume production 2nd half 2027
  • Ramp of significant greenfield capacity in the years ahead
  • On device AI for PCs and smartphones has “pent-up unit replacement demand”
  • Automotive, ADAS remains a powerful driver of content growth
  • Humanoid robots carry 10x the amount of memory as average L2+ vehicle
  • On robotics, “We expect a sustained substantial multi-decade memory demand cycle to begin in the latter part of the decade”
  • Acquired Tongluo site and production ramping in Taiwan, volume in mid calendar 2027
  • Japan and Singapore expanding, Singapore building another center of excellence for advanced packaging
  • Total revenue 41.5B, +74% qoq, +346% yoy
  • DRAM revenue 31.1B +343% yoy, 76% of total revenue
  • NAND revenue 9.9B, +361% yoy, 24% total revenue
  • Cloud Memory Business Unit CMBU 13.8B revenue, 33% total, +78% qoq
  • Core Data Center Business Unit CDBU 11.5B, 28% total, +103% qoq
  • Automotive and Embedded sector 4.6B revenue, +71% qoq, gross margin 79%, up 11 percentage points sequentially
  • Adj EPS +106% qoq
  • Record cash balance 30.2B, reduced debt by 4.4B, now 5.7B debt, net cash 24.4B
  • Guiding next quarter 50B rev, gross margin 86%, EPS $31
  • “We hold the cash” in regards to SCAs, indication of a transformed business model at Micron
  • Fabs that begin volume in mid 2027+ will bring on operating leverage
  • 2027 overall tight supply, “tightness continuing beyond 2027”
  • “So supply even in 2028, when supply begins to improve gradually, we see that demand will continue to be on a robust trajectory as well because these AI trends are very long-term trends. AI is still very, very early innings.”

The management added “we feel great about the trajectory of the business”. I agree with that take, and thought this was an incredible report from Micron. I had been adding to my position a few times before earnings, and still thinking to add again after this report to have them be a higher confidence position. It is incredible that even at Micron’s already massive scale, they have “no line of sight” when the increasing demand will end.

68 Likes
2

To think that this had a market cap of only a few hundred Billion $s until recently and has only just surpassed $1 Trillion.

If you annualise the Q4 guide across a full year it will deliver the highest FY earnings in corporate history with the exception of Aramco in 2022. That makes $1 Trillion market cap look very cheap.

Throw in pricing environment continues to hold out and and extend its horizon, supply build out coming on stream in the medium term to add volume expansion to pricing strength together with a change in business model with longer term recurring revenue like customer agreements, technology leadership and incredible operating margins across the board…

then this is an easy hold if not a buy!

Ant

21 Likes
3

My main concern is that the company’s near-term future growth is limited by their production capacity. That’s what’s driven their margins, enabling them to essentially over-charge for their products (as does Samsung and SK Hynix). And, yes, management is responding by building new production facilities (euphemistically labeled “greenfield”), but those facilities won’t produce until 2028.

That tells me that sometime pretty soon, perhaps early 2027, and lasting throughout 2027, probably into early 2028, the company will hit a revenue/profit growth wall. This isn’t like Nvidia, which was able to secure additional production from TSMC in 2023.

Now, if you’re in Micron for the long haul (5 years plus), then holding through this period will probably work out fine, even well. But, Wall St doesn’t like to wait and in my view will consider waiting for production growth to be “dead money” and many investors will sell to get into something else that is growing faster, and revisit when to come back in, say 2028 as the new facilities come online.

I suspect this isn’t a popular view, and my track record on these things is, to be honest, not great. So, I’d like to hear counters to my expectations.

20 Likes
4

Hi Smorgasbord1 - much of this is “greenfield” and 2028 scheduled but they have along the way acquired and repurposed an existing fab in Taiwan for more immediate capacity and expanded their operations in Singapore off existing footprint for a faster set up - these should bring interim benefit ahead of the larger longer term scheduled production facility projects.

In fact in the earnings call they shared that timelines were accelerated for one of their projects…
“Manufacturing/footprint timing language tightened in one area: the Tongluo site moved from “beginning in fiscal 2028” (prior call) to “mid-calendar 2027, about a quarter earlier than our prior expectations” (current call), per CEO, President & Chairman Mehrotra.”

Also as they move from one technology format to the next within the existing facilities, that secures node yield increases from existing capacity so to a degree enlarges “volume” capacity. Further node volume expansion is taking place as production moves to each successive technology generation in NAND & DRAM within 2026 & 2027 time frame. With supply constraints all around then pricing could still keep increasing.

From the Earnings Call…
“In DRAM, we expect industry DRAM bit shipments in calendar 2026 to grow in the low to mid-20s percentage range, slightly above our prior outlook. In NAND, we expect industry NAND bit shipments in calendar 2026 to grow approximately 20%, unchanged from prior expectations. We expect Micron DRAM supply to grow approximately in line with the industry supply growth, while Micron NAND supply grows somewhat less than the industry supply growth in calendar 2026.”

Agree that Wall St doesn’t like to wait though and there could be some “mind the gap” risk here.

In the meantime - just in case anyone feels that this is purely an AI Data Center HBM driven story…

From the Earnings Call…
"In automotive, ADAS remains a powerful driver of content growth. L2+ and above vehicles, which feature progressively increasing levels of autonomy have over five times the memory and storage content of an average vehicle. The mix of L2+ and above vehicles is more than doubling this year to over 20% and is expected to exceed 40% by 2030. Average auto memory and storage content is expected to further increase as mix shifts towards higher levels of autonomy with progressively higher levels of content.

**In robotics, continued advances in simulation, foundation models and integrated hardware and software stacks are accelerating physical AI. This creates a growing content rich opportunity for high bandwidth, low-power memory and storage that powers real-time perception, inference and control. Humanoid robots carry 10 times the amount of memory as an average L2+ vehicle, and we expect a sustained, substantial multi-decade memory demand cycle to begin in the latter part of this decade."

Ant

28 Likes
5

Yes, fully agree. This was an absolutely phenomenal quarter. I also added a few days ago because I thought they would produce numbers that would knock it out of the park… and they went beyond what I’d expected.

It’s already up over 16% pre market right now. This will most likely bring Micron to my number 2 holding, only topped by Nebius in my portfolio.

Looks like it’s going to be another great day.

Jonathan

15 Likes
6

@Smorgasbord1 I think that is a fair concern that the new “greenfield” fabs are a year or two away from volume production. Still it is worth considering many of these projects “broke ground” in 2023 and are getting to the half way point. The company is spending 10B+ per quarter now on CapEx, so they scale is large.

It is a possibility they have a couple quarters that stall out around ~75B revenue. Additionally, their margin is getting practically maxed out. However, I’m not too concerned about this as the valuation on the stock isn’t pricing in enough growth from my viewpoint.

The company can still get efficiency gains from current production facilities. They mention 20% bit growth which means essentially 20% more production but more efficient memory chips. I was asking AI about this, if they are effectively supply blocked themselves from scaling up right now. The responses I got mentioned a few ways they are gaining additional production, or efficiency,

  • Process node transitions - migrating to smaller nodes with tech such as 1-gamma DRAM, more chips per wafer on the same equipment
  • Yield improvements - as new process nodes mature, defect rates drop, meaning more usable dies per wafer
  • HBM stack optimization - producing more memory chips through architectural improvements

I do think it is a fair callout about the production capacity limitations though. Many of their current gains are from price increases, and their longer term 5-year deals may lower pricing power. However, they are getting unrestricted cash up front from customers, which can further fuel expansion plans.

16 Likes
7

I defer, especially to the cutting edge tech savvy in this group, concerning this company and so have to limit my exposure to MU.

The expansion of memory demand as an industry or signficant industry subsection is quite new to most of us. It does appear however, that AI has changed the memory industry (for the present) from a cyclical industry just providing a resource to a structural bottleneck creating upward price pressure and enormous investment.

See Chris Zeoli substack article. The idea being that MU is part of, and is now benefiting from a durable DRAM triopoly with SK Hynix and Samsung, fully booked HBM (High Bandwidth Memory) capacity, and rapidly rising HBM mix. These companies appear to be straining to keep up with demand.

One risk, and perhaps remote, is that we return to the cyclical industry/commodity type characteristics, say if there is a reduction or slowdown in AI capex or slower hyperscaler spending or HBM oversupply. But when everyone is buying tulips, that could be a good time to think about the downside to owning flowers.

Cheers,

Bill

3 Likes
8

@anthonyms and @wpr101, thanks much for your replies. Good to know that there is room for production volume improvement before the new sites come online, even if somewhat limited.

I did a little research on Samsung and SK Hynix, and while both companies are also out of advanced cleanroom square footage, they have larger base-wafer capacities. Apparently, the danger for Micron is that if they achieve a breakthrough in production yields they could ramp up production to a higher level faster than Micron, which won’t have the cleanroom space until its 2028 greenfield expansions.

SK Hynix’s M15X fab in Cheongju is scheduled for utilization in mid-2027, and the Yongin cluster is expected in late 2027. Samsung’s Pyeongtaek P5 fab has been pushed to sometime in 2028.

Also, note that there are different kinds of memory that are used for “AI purposes”: HBM, NAND, and DRAM. Here’s a table on them:

Memory Type Core AI Function Production Complexity Operating Profit Margins
HBM Direct GPU execution (e.g., LLM training & inference) Extremely High (Requires advanced 3D stacking & TSV packaging) 50% – 65% (Commanding massive enterprise price premiums)
DRAM System memory for AI host CPU servers & caching Moderate (Relies on traditional wafer scaling to 1b/1c nodes) 50% – 90% (Currently peaking due to severe artificial supply deficits)
NAND Massive data ingestion pools & model storage (SSDs) Low to Moderate (Highly optimized vertical 3D-NAND stacking) 35% – 50% (Experiencing a strong recovery led by enterprise QLC)

The businesses have a number of decisions to make on how they allocate their production. For instance, Micron stopped making consumer product memory to convert to data center focused memory since the margins are higher. And, it takes 3X the capacity to make a byte of HBM versus a byte of DRAM, and so the shift to HBM for GPUs created a shortage of DRAM, sending those prices and margins up. But, Micron and Samsung can’t just switch since they already contracted to produce HBM. FWIW, SK Hynix has about 58% of the global HBM market, with Micron and Samsung splitting the rest about equally.

And switching production between memory types is costly and takes time. These companies have to predict where the market will be. If they decide that DRAM’s current shortage means they should produce more of that, but so does the other 2 companies, than that shortage may go away and they would have been better off making HBM instead. Or, if GPU build-outs stagnate or decline, then producing more HBM instead of DRAM would have been the mistake.

Now, Micron has gone from one customer contract to a reported 16 in last 3 months (including 4 hyperscalers and 3 enterprise customers), with large upfront payments ($22B) and large non-cancellable commitments ($100B), so the risk for what they’re producing being the wrong mix seems almost non-existant. Micron says these 16 agreements lock up about 1/5 of their total DRAM volume and 1/3 of its entire NAND flash volume through 2030. Micron said their HBM production capacity up through the end of 2027 is already spoken for, as well. And, that that is only about half of what the current demand is.

While Mr. Market seems to like demand outpacing supply, I think that’s somewhat short-sighted. If there isn’t enough memory, what does that mean for GPU or ASIC production and sales? And what does that mean for data center build-outs? And then what does that mean for AI adoption - and pricing?

If everyone from ChatGPT users to Palantir to enterprise customer wants AI services but as a planet we can’t expand AI server production deployment to satisfy that demand, what happens? Are Nvidia’s sales limited? Do OpenAI and Anthropic have to raise prices, if only to reduce demand? Where is all the capex from Google, Amazon, Oracle, etc. going if there’s not enough memory to satisfy the capex expectations?

13 Likes
9

@Smorgasbord1 Not sure if you saw but the SK Hynix ADR is set to launch on US markets in a few weeks. They have English transcripts, are cheaper than Micron and possibly have more innovation or production capacity. I’m planning to start a position in them once it starts trading.

There some technologies being created which will likely alleviate some of this pressure, even though memory output is not keeping pace.

The one I know most about is Credo’s Weaver product. Here are the supposed performance improvements,

Credo's Weaver product
Credo's Weaver product2242×900 252 KB

Basically the ASIC buildouts will have to start looking at doing more with less, by using products such as Weaver or equivalents. Maybe this comes back full circle to the other thread where you mentioned data center capacity and power won’t be able to keep up. It seems like the industry will be forced to innovate even further in terms of power, compute, and memory.

9 Likes
10

Yes, I’ve been watching that. I also currently have small positions in EWY and FLKR, which are South Korean ETFs. EWY is almost 30% in SK Hynix and 20% in Samsung, and FLKR is similarly allocated.

My understanding is that this is only good for inference. On the training side, HBM is still required for its speed. It’s still early days for OmniConnect Weaver, with design wins only from some AI startups, not any hyperscalers, although reports are that Google is evaluating it. I think real revenue from it isn’t expected until 2028 anyway.

Yeah, probably. And I’m sure there are research projects at the hyperscalers and universities working on optimizing software to require less. Exciting times.

6 Likes
11

An interesting article about Micron in Barrons today. I’ll link to the story at the end of this post. I’ll also summarize key points in case you do not have a subscription.

  • Micron is working to smooth out the supply and demand cycles that have plagued them and the industry for decades.
  • MU CEO says that the industry supply will improve gradually in 2028. There is no visibility as to when supply will catch up to growing demand.
  • The 3-5 year take-or-pay contracts that MU has been announcing will already make up 40% of MU’s revenue over that period and they expect to sign more of these favorable contracts. This locks in demand and pricing.
  • Maybe most eye opening to me was MU’s forward forecast on their gross margins. The high water mark for GM in past cycles was 61% in Q4 2018. GM is now 85%. The CEO said that 61% GM will be the low point going forward…the expected range will be 61-85% which means massive cash flow and profitability for the next five years+.

https://www.barrons.com/articles/micron-stock-price-still-cheap-12408e32?mod=hp_LEDE_C_1

13 Likes
12

No subscription here. Obviously don’t quote a huge block, because that would be against policy, but could you maybe provide a sentence or two from the article to support this? I’m interested because I haven’t heard that anywhere else and I don’t see anything about it in the transcript from earnings.

Bear

4 Likes
13

The price floor will be “well above our peak quarterly margins in any past cycle,” Mehrotra said, which was 61% in the fourth quarter of 2018. Since 2010, half of Micron’s quarters had gross margins below 32%, so anything in the 60% range will prove immensely profitable—and it’s locked in through 2030 for a big portion of its customers.

Here is a link to a free Yahoo Finance article that basically says the same but in less detail.

5 Likes
14

GM will be the low point going forward…the expected range will be 61-85% which means massive cash flow and profitability for the next five years+.

I’m guessing this part about 5 years, and the margins, comes from the nature of the agreements that Micron and other memory makers have signed recently.

  • Micron has 16 mostly 5-year deals that is 100B in RPO
  • Sandisk has three 5-year deals worth 42B in RPO

Micron related points from last earnings,

  • “Typically, these agreements have a 5-year term from calendar 2026 through the end of calendar 2030.”
  • “These SCAs accelerate the transformation of our business model, enhance partnership in technology and innovation and provide customers with contracted supply assurance.”
  • “These SCAs are structured as take-or-pay agreements, we are binding commitments to purchase specific volumes over this multiyear term”
  • “For our SCAs with price bands, the floor price enables a very robust gross margin for Micron, well above our peak quarterly margins in any past cycle”

Sandisk related points from their last earnings,

  • “We believe our margins are sustainable”
  • “We provide supply, they provide demand and we have visibility for many years all the way to five years”
  • “… get a fair return for our products and get the cyclicality out of the business”
  • “what’s happening is there are now customer sets that very substantial customers that don’t want to play the quarter-by-quarter price game”
  • “Fundamentally change the way this business has worked over the last several decades”
  • Looking to get setup where different deals have different end dates, so there is no cliff

It seems like both companies are working to reduce the cyclicality in the business through these agreements. While it may impact the max price they can charge quarterly, it should remove some of the boom/bust element to their business. The deals are very favorable to both Micron and Sandisk in terms of cash up front.

27 Likes