The AI Memory Supercycle: Micron Reports June 24 With 81% Margins and HBM Sold Out Through 2026

On June 24, Micron reports earnings. And I am paying attention, because this is not a normal chip report. The company has already told the world what it expects: revenue of around $33.5 billion for the quarter, give or take $750 million, with diluted earnings per share near $18.90. The number that actually made me sit up though is the gross margin. Micron is guiding for roughly 81% gross margin. A year ago that figure was less than half of that.

Let me say that again so it lands. A company that makes physical memory chips, a business that has historically been one of the most brutal, boom-and-bust, low-margin corners of the entire stock market, is about to print software-level margins.

That is the story. That is the whole story. And almost nobody outside of hardcore semiconductor nerds is talking about it the way they talk about Nvidia.

When people think about the AI boom, they think about one company. Nvidia. The GPUs. Jensen in the leather jacket. And fair enough, that has been the trade of the decade so far.

But here is the thing nobody explains to you when you are starting out. A GPU on its own is a very fast brain with no memory. It can calculate at insane speed, but it needs somewhere to hold the data it is calculating on, and it needs to feed that data in and out fast enough to keep up. That is what memory does. Specifically, the kind of memory that matters for AI is called high bandwidth memory, or HBM.

HBM is not your normal RAM. It is specialized, stacked, ultra-fast memory that sits right next to the AI accelerator. Without enough of it, the most expensive GPU on earth sits there starving, waiting for data. So when companies build AI data centers, they are not just buying GPUs. They are buying GPUs plus a mountain of HBM to go with them.

There are only a handful of companies on the planet that can make HBM at scale. Micron is one of them. And Micron has said something that changes how I read this entire sector: its HBM capacity is sold out through 2026, with reported demand outstripping available supply by something like 50 to 67 percent.

Sold out. Not "selling well." Sold out, with a line of customers wanting far more than exists.

People in the memory business love the word "supercycle." Every few years someone declares one and then prices crash and everyone looks silly. So I want to be honest with you about why this time looks different, and also where the risk hides, because there is always risk.

Historically, memory was a commodity. DRAM and NAND chips were basically interchangeable. When demand was hot, every manufacturer cranked up production, flooded the market, and prices collapsed. Then they all cut production, supply got tight, prices spiked, and the cycle started over. It was a meat grinder. Micron has lost money in entire years before. That is not ancient history, that is the normal life of this industry.

What is different now is that HBM is not a simple commodity. It is hard to make. The manufacturing yields are difficult, the packaging is complex, and the customers, the big AI players, sign long agreements and care more about getting supply than about squeezing the last dollar out of the price. When your customer is desperate for the product and there are only three serious suppliers in the world, the pricing power flips. That is how you get a memory company guiding to 81% gross margins.

Demand is reportedly outpacing supply across HBM, DRAM, and NAND all at once, and industry supply constraints are expected to last well beyond this calendar year. When all three product lines are tight at the same time, the whole company prints money, not just the fancy HBM part.

I look at all that and I do not think "this removes the risk." I think "this is a real structural shift, and I want to understand both the upside and what could break it."

I want to walk you through my process, not give you a verdict, because the process is the part you can actually use on any stock for the rest of your life.

First, I separate the expectation from the reaction. Micron has already told us it expects a monster quarter. That means a monster quarter is, to some degree, already baked into the price. The dangerous mistake beginners make is seeing great numbers, assuming the stock must go up, buying right before or right after the report, and then watching it drop anyway because the great numbers were already priced in. A stock can report record results and fall 10% the same day. That is not the market being irrational. That is the market having already known.

Second, I look for the guide, not the print. The number for last quarter matters less than what management says about next quarter and the year ahead. If Micron beats on the current quarter but sounds cautious about the back half, the stock can sell off. If it meets expectations but raises its outlook and says supply is even tighter than thought, it can rip. I always read the forward language first.

Third, I ask what would prove me wrong. For the memory bull case, the things that would scare me are simple to name. A sudden glut if competitors bring capacity online faster than expected. A slowdown in AI data center spending if the big buyers decide they have built enough for now. Or a broad recession that drags down the PC and phone markets, which still make up a big chunk of memory demand. If I cannot name what would break my thesis, I do not actually have a thesis. I have a hope.

This connects to something I have written about before in position sizing in a high valuation world. When a stock has run hard and expectations are sky high, the size of your position matters more than being right about the direction. You can be completely correct about the company and still get wiped out if you bet too big and the timing goes against you for a few months.

I keep coming back to one idea. The AI buildout is not one stock. It is a supply chain. And the further down the supply chain you look, the more interesting it sometimes gets, because that is where the crowd is not staring.

The crowd stares at Nvidia. I get it, and I have written plenty about why Nvidia in 2026 is not as simple a story as the bears think. But the same demand that lifts Nvidia has to lift the things Nvidia needs to function. Memory is one of those things. So are the networking and custom silicon plays I dug into when I wrote about Marvell being one of the quietest AI monsters out there.

Here is the mental model I use. Imagine a gold rush. Everyone wants to buy shares in the company that found the biggest gold mine. Fine. But somebody has to sell the shovels, the pickaxes, the water, the railroad ties. In an AI gold rush, HBM is one of the shovels, and right now the shovel maker is sold out with a waiting list. That does not guarantee the stock goes up. Plenty of shovel makers have been terrible investments over the years. But it tells you the demand is real and physical, not just a story.

I cannot write about any stock right now without talking about the macro backdrop, because it shifted hard this month.

This was the first Fed meeting under new chair Kevin Warsh, and it did not go the way the bulls wanted. The Fed held rates steady, which was expected, but the projections told a different story. Nine of the eighteen policymakers now see at least one rate hike by the end of 2026. A hike. Not a cut. The market spent most of this cycle praying for cuts, and instead got a coin flip on a hike.

Warsh kept hammering the words "price stability," which is Fed-speak for "we care more about killing inflation than about keeping your stock portfolio happy." Stocks tumbled into the close on the day of the decision. It was reportedly the worst "Fed day" under a brand new Fed chair since 1994.

And yet. Look at how the week actually finished. Despite that ugly Fed reaction, the S&P 500 still gained about 0.9% for the week, its eleventh winning week out of twelve. The Nasdaq jumped 2.4%. Tech strength and optimism around the US-Iran deal more than offset the hawkish Fed.

I wrote about the mechanics of this Fed shift in more detail when I covered how nine officials flipping toward a 2026 hike is repricing everything. The short version for a stock like Micron is this. Higher-for-longer rates are generally a headwind for high-growth, high-valuation names, because future profits are worth less when you discount them at a higher rate. But a company that is literally sold out and printing 81% margins has something most stocks do not have in a hawkish environment: actual current earnings, right now, in cash. That is a very different risk profile from a story stock that needs cheap money and a friendly Fed to survive.

Micron does not report into a vacuum. The same week brings FedEx earnings, and a little further out we get first quarter GDP and the May reading on PCE prices, which is the Fed's favorite inflation gauge.

I always pay attention to FedEx for a reason that has nothing to do with shipping stocks. FedEx moves the physical economy. When FedEx talks about volumes, it is telling you whether real businesses are shipping more or fewer actual boxes. It is one of the cleanest reads on whether the economy is speeding up or slowing down. So when Micron tells me about the AI future and FedEx tells me about the present-day economy in the same week, I get a stereo picture instead of a mono one.

And that PCE number at the end of the month is the one that could decide whether that Fed hike talk gets louder or quieter. If inflation comes in hot, the hawks get bolder and rate-sensitive stocks feel it. If it comes in soft, the whole "hike in 2026" conversation could fade fast. I am not predicting which. I am telling you what I will be watching, so you can watch it too.

Here is where I get protective of you, because I have seen this movie too many times.

When a stock has a famous "sold out, record margins, supercycle" story attached to it, that story becomes a magnet for exactly the wrong behavior. People who have never bought a single share suddenly want to go all in the day before earnings because they heard it is going to be huge. That is gambling with extra steps.

The story being true does not make the trade safe. Micron could report a genuinely incredible quarter and still fall, because the bar was set impossibly high. Or it could fall for a week and then grind higher for a year. Or the supercycle could roll over in eight months and the whole thing could deflate. I do not know. Nobody knows. Anyone who tells you they know is selling you something.

What I do know is the difference between investing and reacting. Investing is deciding ahead of time what a company is worth to you, what could break the thesis, how much you are willing to risk, and then sizing your position so a bad outcome does not ruin you. Reacting is seeing green numbers and FOMO-buying with money you cannot afford to lose. One of those is a skill you build over decades. The other is how young investors blow up their accounts and quit forever.

I am treating the Micron report as a window into the health of the entire AI hardware supply chain, not as a one-day bet on a stock pop. If the company confirms that HBM and DRAM are still wildly supply constrained and raises its outlook, that tells me the AI infrastructure spending is still very much on, which matters for a whole basket of names, not just this one. If it sounds cautious about supply loosening or demand cooling, that is a warning flag for the entire sector that I would take seriously.

Either way, I am not letting a single earnings report bully me into a rushed decision. The report is information. What I do with information is decide slowly, size carefully, and never bet the rent.

When I evaluate any company, one of the first questions I ask is "how many serious competitors are there, and how easy is it for a new one to show up?" In a lot of industries the answer is "tons, and very easy," which means margins get competed away fast. In HBM, the answer is "almost none, and extremely hard."

There are essentially three players in the high bandwidth memory game worldwide. That is it. Building a leading-edge memory fab costs tens of billions of dollars and takes years. You cannot just decide to enter this market next quarter. The knowledge, the equipment, the yield expertise, the customer relationships, all of it takes a decade to build. That kind of barrier is exactly what lets a company hold onto fat margins instead of watching them erode.

Now, here is the honest other side. A three-player market can flip from "everyone is sold out and minting money" to "everyone overbuilt and prices are crashing" faster than you would think, because when only three companies decide how much to produce, their collective decisions swing supply hard. If all three race to add capacity at the same time chasing these juicy margins, two years from now there could be a glut. That is the eternal memory risk, and no story about a supercycle erases it. I just keep it on my radar instead of pretending it does not exist.

If half of this post sounded like a foreign language at points, that is normal, and it is not your fault. The finance world hides simple ideas behind fancy words on purpose. So let me knock down a few walls.

Gross margin is just the percentage of each sales dollar left after paying the direct cost of making the product. An 81% gross margin means it costs about 19 cents to make something Micron sells for a dollar, before other expenses. Higher is better, and 81% for a hardware company is wild.

HBM, DRAM, NAND are three kinds of memory. DRAM is fast, short-term memory. NAND is slower storage that holds data when the power is off, like in your phone or SSD. HBM is the premium, stacked, ultra-fast memory built for AI chips. All three being in short supply at once is the unusual part.

Priced in means the market already expects something, so it is already reflected in the stock price. Good news that is fully priced in does not push a stock up, because nobody is surprised.

The guide is management's forecast for the next quarter or year. Markets often care about the guide far more than the result just reported, because investing is about the future, not the past.

Learn these four ideas and you can follow 90% of earnings coverage without feeling lost. The jargon was never the hard part. The discipline is.

I want to leave you with a habit, not a stock tip. Whenever I find a trend everyone is obsessed with, I force myself to ask "what does this trend physically require, and who supplies that?" Electric cars require batteries and lithium and charging networks. AI requires power, cooling, networking, and memory. The headline winner gets all the attention and usually all the valuation. The suppliers one layer down sometimes get overlooked, and overlooked is where opportunity occasionally lives.

This is not a magic trick that always works. Sometimes the supplier is a terrible business with no pricing power and you are better off ignoring it. The point is the question, not a guaranteed answer. Asking "who supplies the supplier?" turns you from someone who chases headlines into someone who maps an entire industry. That map is worth more than any single trade, because you can use it for years.

## Your move this week

Here is my challenge to you, and it does not cost a dollar. Before Micron reports on June 24, write down three things on paper or in your notes app. One: what you think the stock will do. Two: why you think that. Three: what would have to happen for you to admit you were wrong.

Then after the report, go back and read what you wrote. Do not trade on it. Just compare your prediction to reality. Do this ten times across ten different earnings reports and you will learn more about how markets actually work than any course could ever teach you. You will start to feel, in your gut, the gap between a good company and a good investment, and how often the market has already priced in the obvious.

Build the muscle of predicting, recording, and reviewing. That muscle, not any single stock, is the thing that compounds.

*Disclaimer: MentorSurge is not a financial advisor and this is not financial advice. This post is for educational and entertainment purposes only. Nothing here is a recommendation to buy or sell any security. Investing involves substantial risk of loss. Numbers cited were accurate when written and change constantly. Always do your own research and consult a licensed professional before making decisions with real money.*