Micron sells semiconductors for memory and storage. Its “Not And” (NAND) chips offer non-volatile data storage, which maintains stored information when a system’s power is turned off. It’s ideally suited for longer-term data storage use cases where processing latency isn’t the primary concern. For many of the longer-term storage tasks as well as things like saving model progress during training, this is the cheaper option. These chips provide the foundation for its solid-state drives (SSDs), which are used in computer data storage and things like USB flash drives. Micron sells standalone NAND chips as well as SSDs with NAND chips in them. SSDs replace hard disk drives (HDDs), as they’re more power efficient, durable and resilient. It provides basic memory cards for things like gaming devices and cameras as well.

Separately, its Dynamic Random Access Memory (DRAM) chips offer volatile memory storage for personal computers, data centers, and mobile devices. “Volatile” means that storage isn’t maintained when a system’s power is turned off. DRAM is a big piece of model training and inference as the primary, low-latency workhorse. If you tried using NAND to create intricate chatbot outputs, we’d be waiting all day. It’s more expensive but it’s needed in these cases. Perhaps most interestingly, Micron offers a type of DRAM called high-bandwidth memory (HBM) to meet the massive data needs of GenAI. It sharply improves data processing capacity by stacking more memory into one single unit. Nvidia is a big customer, using Micron’s HBM in its Blackwell and future Vera Rubin systems. While Micron entered the HBM space years after SK Hynix and Samsung, they quickly racked up nearly 20% market share thanks to power efficiency strengths and overall memory demand levels.

AI is a sizable tailwind for both major buckets, as storage and memory needs have notched a profound step-change in overall demand stemming from the AI boom.

As a reminder, the CHIPS and Science Act banned Micron from buybacks for 2 years. That expiration date is December 9th. There’s no minimum set cash balance Micron has in mind, but they do plan to significantly lean in soon. They just reported $33B in FCF for Q4 and think they’ll be “around the cash target level” after next quarter. After that, the buyback pace should sharply accelerate. In terms of CapEx, while they did raise guidance here for next year (more later), they do think capital intensity levels for this cycle will remain below other cycles.

For fiscal year 2027, they expect Q/Q revenue growth every quarter. At the same time, the fantastic performance will mean higher employee compensation levels, which weighed a bit on the slightly disappointing Q1 GPM guidance. Emphasis on slightly, as an 86.25% guide is still eye-poppingly great. They see gross margin improving beyond Q1 as this headwind eases. In terms of price increases, they talked about a more moderate pace for 2027. That might lead to a small amount of disappointment from some investors as they question whether or not the pace of price increases has now permanently peaked. And again, I’m being very picky here. Finally, thanks to stronger demand trends and rising visibility thanks to multi-year contracts, they plan to spend more than their previous $40B FY 2027 CapEx guide. Micron now expects $25B in Q1-Q2 CapEx and even more for the second half of the year. The slower pace of price hikes, the GPM guidance and the higher CapEx forecasts were the only modest negatives amid a much bigger sea of more important positives.

Other Guidance Notes:

Micron trades for 7x forward EPS. EPS is expected to grow by 791% this year, by 119% next year and by 13% the year after. Estimates will rise following this report.

Cycle Visibility:

Micron somehow enjoyed an incrementally strong demand environment compared to last quarter. Interest levels went from historically amazing to even more amazing. And while that’s encouraging, it’s not what is going to positively surprise investors at this juncture. Micron has trained the world to expect masterful outperformance. Everyone knows they’ve flawlessly taken advantage of AI tailwinds and that those tailwinds are as strong as they’ve ever been. This is both a rising tide raising all boats, and Micron building a better boat than some of its competitors. That’s how it has racked up nearly 20% global HBM market share despite starting years after Samsung and SK Hynix.

So what will deliver that positive surprise and share price returns beyond the grand slam that bulls have already hit? Incremental confidence in cycle longevity and rising visibility into demand beyond calendar 2027. Stocks are forward-looking and this one is notoriously famous for fantastic momentum unexpectedly turning on a dime. That’s why this thing trades for 7x forward earnings. That’s investors telling us they’re still very worried about that eventual cyclical peak and what happens to Micron’s P&L during the next inevitable downturn. There’s a small minority that does think this company is no longer cyclical, but I just do not agree. I think demand factors for Micron are stronger than they’ve ever been and this cycle is more impressive than any others before. With all of that said, the best thing Micron can do right now is add conviction to the idea that this down period is nowhere in sight. And? I think that happened. The robust 2027-2028 commentary briefly discussed in the outlook section is a byproduct of insatiable demand and tight supply conditions (for DRAM and NAND). Both of those things are expected to intensify during 2027 and 2028 compared to fiscal 2026, which just wrapped up as their best financial year ever (by a long shot). Simply put, business is booming and the boom is getting bigger and louder, if anything.

“We do not have line of sight to when supply and demand will return to balance.” – CEO Sanjay Mehrotra

For additional compelling evidence pointing to this cycle having plenty of runway, its multi-year strategic customer agreements (SCAs) are becoming a much larger part of its backlog. That’s so important. If so many are worried about aggressive, sharp declines in financial performance whenever pricing power and supply tightness issues peak, long-term contracts can help alleviate much of that concern.

Micron’s business isn’t solely and woefully reliant on one-time transactions, short-term contracts and healthy spot markets anymore. While I again think this company is still absolutely cyclical, this change will likely make the severity of the dips somewhat less extreme in the future. Since last quarter, Micron signed an additional 10 deals to reach 26 total. Backlog from these deals jumped from $100B to $150B, while SCAs in total are expected to be 35% of revenue through 2030 compared to a mid-20% proportion as of last quarter. Better yet, a little more than 26% of its overall revenue (three quarters of the 35%) has “defined pricing” with pre-set floors that are above previous cycle peaks. Again, while Micron would be far from immune if demand unexpectedly soured, this would put them in a better spot compared to previous cycle troughs. Impressively, as customers seek supply certainty over the next several years, Micron has signed SCAs through 2031. Micron is finding it very easy to trade a bit of potential gross margin excess in exchange for clearer demand over the next few years. I think that’s a smart trade to make.

Other Cues & Positioning:

It’s nice to hear from Micron that things are remarkable, but it’s even better when that’s backed up by clear evidence from others. In this case, there’s plenty. Muse is becoming the most successful consumer product launch of the AI boom (and maybe ever). The free access to hefty amounts of compute and historic traction make it clear why this is so good for Micron. All of those workloads and side chats require memory. Micron is a big partner for Meta. Elsewhere, 2027 mega-cap CapEx commentary has been uniformly upbeat, models simply keep getting bigger, and AMD is leaning heavily into offering more HBM capacity.

On the other hand, Nvidia has shown more interest in shifting some DRAM needs to SRAM with their licensed Groq technology. DRAM-based HBM needs enormous data processing capacity at world-class latency to help AI models and agents stand out from the pack. Although SRAM can be used in some places to trim costs, DRAM remains the workhorse. Even in the version of their Rubin platform that uses the most SRAM, there are still 168GB of HBM4 per chip (vs. 288GB for other iterations) and the two infrastructure titans also continue to work together in other ways. For instance, Micron and Nvidia are collaborating on the first custom HBM4E implementation (called NV-HBM). Really great to hear as rumors of Nvidia despecifying (or reducing HBM usage) run rampant. Micron is not concerned about this and expects their work with Nvidia to increase going forward.

Generally speaking, running models, conducting agent-based work, storing everything and moving data where it needs to go for all of this cannot happen without memory vendors like Micron. That, paired with the vast supply demand mismatch, is why their results look so unbelievable right now. They have all the pricing power and they’re taking advantage just like their customers have done to them during periods of inventory gluts. Knowing this, the world is trying very hard to lower reliance on HBM players as inflation runs rampant and customers look to control exploding costs. Every single mega-cap is enduring some level of margin pressure because of this requirement and is trying whatever it can to alleviate things. Despite all of that work, the industry has made very little progress and HBM remains vital in most places.

In terms of how Micron stacks up vs. other HBM vendors enjoying this same momentum, they think quite well. Every quarter, Micron leadership calls itself the clear technology leader in the field. They believe performance and efficiency are why they’ve so seamlessly overcome a later mover disadvantage and see their memory performance as a durable differentiator that lets customers run better AI products. Aside from that, rising U.S. manufacturing capacity is a compelling geopolitical risk eliminator. Finally, Micron does not care if open-source models continue gaining popularity. They all use a lot of memory too.

Roadmaps – Manufacturing & Technology:

Everything is on track. Construction across its New York, Idaho and Virginia facilities is advancing on schedule, as they notched an important concrete pouring milestone in New York this quarter. In Japan and Singapore, all DRAM and cleanroom footprint expansion is also going as planned. Wonderfully boring execution here. Micron is working extremely hard to add more supply and diminish the crippling shortages its industry is facing. 

And similarly, on the technology roadmap, their newest 1-gamma DRAM nodes and G9 NAND nodes are ramping as meaningfully as hoped for, with the next generation of manufacturing technology (you guessed it) moving forward without any hiccups.

More on Data Center Demand:

Overall industry server unit growth is expected to stay just under 20% for calendar 2026 and 2027 while price hikes provide plenty of growth on top of that. In DRAM and HBM, almost all unit supply through the end of 2027 is spoken for with favorable pricing that will narrow the GPM gap between HBM-based DRAM and its overall DRAM business. Specifically, prices rose by around 18% since just last quarter while DRAM unit volume growth was around 5% Q/Q. Most of this stellar growth is due to higher prices, as Micron is already selling everything they can make.

The NAND story is very similar. While this segment may not be as sexy as memory, storage is still a key cog in the AI machine. Models and agents need to access massive sums of data on a moment’s notice with storage being maintained whether a system is on or off. AI also leans on KV cache to store previous model outputs for reuse. That all requires NAND and is why pricing rose 30% Q/Q while unit volume growth was a healthy 10% Q/Q.

Other Notes:

Great quarter. Demand levels remain unfathomably strong while we’re still several quarters away from any kind of supply-based relief. They’re taking HBM market share, delivering excellent NAND-based momentum and improving the quality of their business model through more SCAs. I do think some will fixate on the “slower pace of price hikes” commentary and will wonder whether that marks peak pricing growth rates. They’ll also highlight the 86.25% Q1 GPM guidance, which was below 87% consensus, and the raised 2027 CapEx guide. To me, focusing on those three details is missing the forest for the trees. And in this case, the forest is mighty, healthy and still voraciously growing. Everything they said in this call points to this cycle still having plenty of room left to run and leaves me with far more reason to be optimistic than skeptical. Well done, again.