Memory Stocks Just Hit Records & Here's Why That's Dangerous

Added: 2026-05-12

[00:00:00]Everyone is bullish on memory right now.

[00:00:02]Micron is up. SanDisk is printing record margins. And the thesis is absolutely everywhere. AI is driving a structural shift. The old boom bus cycle is over.

[00:00:11]And this time it's different. Analysts are calling it a super cycle. A new era.

[00:00:15]A permanently higher baseline for the entire memory industry. But here's the thing. They might actually be right. But the last time this exact argument was made this confidently with this much consensus at this kind of valuation was 2018. And what happened? Micron fell 55% over the next 12 months. Today, I want to take you inside the memory super cycle thesis. Not to dismiss it, the structural case is actually genuinely compelling, but to show you exactly where it holds, where it breaks, and what every investor needs to be watching right now if you've got exposure to this sector.

[00:00:47]So, the setup is obvious. Memory is running hot, and that's just not even up for debate. And the market, to be honest, is responding exactly how it should. Stocks are moving higher because earnings and expectations are moving higher. But this is also where people are starting to get very lazy. Memory has always looked like the strongest when the supply is tight and pricing is peaking. That's when margins expand the fastest, sentiment flips bullish, and the narrative becomes this time is actually different. How many times have we heard this? Now, that doesn't automatically make it wrong. But it does mean expectations are rising very quickly. So, the key question is not whether things are good. They obviously are. The question is whether current conditions represent a sustainable baseline or is this the best part of the cycle where everything just happens to be working all at once. If you're honest right now, you have tight supply, aggressive demand, and expanding margins all hitting simultaneously. And you know what? That's usually very late cycle behavior, even if the structure underneath is actually improving. So, I wanted to talk about the real framework and kind of why structural demand versus market expectations is so important to run through. And to do that, you need to separate what's actually happening in the business from what the market is currently pricing in.

[00:01:48]>> The bottlenecks within the AI trade because he said that the markets love bottlenecks. So, it used to be that the bottlenecks was GPUs. Remember when we were talking about Nvidia's GPUs and they were hard to get etc. Well, now it's CPUs >> because like I said structurally demand is improving. AI workloads are requiring more memory per system. Data center buildouts are only getting larger and customers are spending more aggressively and that is raising the baseline like I said earlier for the entire industry.

[00:02:14]But right now the market doesn't just price improvement. It prices the future of that improvement. So right now expectations are being pushed towards a scenario where demand is staying strong through 2028 29 and even 2030 pricing remains elevated and margins are actually holding and that's exactly where the risk builds because even if the long-term story is correct the path will definitely not be linear. Demand does not grow in a straight line and supply does not stay constrained forever. This time might be different or it might not. The mistake most people are making right now is they're assuming that the strong current fundamentals mean strong forward [music] returns. And that's not how this works in a structurally cyclical industry. The business can be improving at the same time, but the trade can actually become less attractive because expectations have already moved so far ahead of reality. And if expectations are pricing in peak conditions holding, then even a normalization from great to good can still disappoint the market. And that's when you see a pullback before they actually show up in the numbers. And with this said, we actually have to understand what AI is actually changing in the memory space. Because AI is not just increasing demand, it's changing the actual structure of demand. In the past, memory is scaled with units. So what this means is more devices meant more demand. Now it scales with intensity. Each AI system actually requires significantly more memory than a traditional system. Training models require high bandwidth memory and running those models requires large amounts of DRAM. And storing and retrieving that actual data requires nan. And so that's why this demand is hitting all parts of the stack at once.

[00:03:36]Companies like Nvidia are effectively driving this because their systems require massive amounts of memory bandwidth to actually even function. And even consumer companies like Apple now are are increasing memory in devices to support new AI features. So the demand is obviously real and it is structurally higher than ever before. But that doesn't mean it grows at the same pace forever. And that's exactly what us investors have to figure out and price accordingly. It just means that this starting point now is higher and the slope of growth has been steeper in the earlier phase of adoption. The key is understanding that this is front-loaded growth. Strong early acceleration followed by eventual normalization as infrastructure gets built is exactly what might happen. And one of the key risks that no one is really talking about and we really have to concentrate on is the risk of demand actually being pulled forward. The biggest thing people right now are missing is timing. So demand is strong like I've said, but it also is being pulled forward. And what that means is the buyers have changed.

[00:04:27]Instead of fragmented customers reacting to price, the market is driven by a small amount of hyperscalers building out their infrastructure. These companies are already buying ahead of demand. They're locking in supply, securing capacity early. And that creates a very strong near-term setup, which is exactly why we're seeing names like Micron and SanDisk run like crazy.

[00:04:45]So demand is spiking. Supply looks tight. Obviously, prices are moving higher. But it also means that some of that demand would have happened later.

[00:04:51]At some point, those customers stop buying at the same pace because they already have the capacity installed and their cash flow simply will not be able to keep up with demand. That's when growth normalizes. Not because the end demand will necessarily disappear, but because it was front-loaded in the first place. And this is how cycles reset exactly in this environment. It's not through a collapse, but it's through a slowdown in incremental demand. And that shift usually happens before it's obvious in the numbers. And by the time it shows up in those reported results, the stocks have already started reacting. they've fallen 10 20 30% and it's already too late for investors to get out and trust me you do not want to be in that position. But what's interesting about this now is there's the super cycle debate. All right, if we're actually going to define the super cycle properly because this is where people again are getting lazy and just repeat what they've been hearing without doing any work. When people say we're in a memory super cycle, what they're really saying is the business has structurally improved and to a degree that definitely is true. And Micron is putting up these numbers that don't look like a normal recovery. So they're saying okay this is actually going to stay around for a long time perhaps even forever and the businesses have structurally changed where in history that's not what we've seen. Perfect examples are Micron and SanDisk. And for some of the numbers for Micron you got revenue nearly tripling year-over-year.

[00:05:59]Every segment has hit record levels and forward guidance that implies a single quarter could exceed any full year they did prior to the AI ramp. So that tells you that obviously demand is real and is very broad. It's in HBM, it's in DRAM, it's in NAND, it's at data center, it's all of it. And on top of that, Micron is talking about multi-year agreements and even a 5-year deal, which historically you didn't see in memory. So that supports this idea that demand is more strategic and less transactional moving forward. But this is again where we have to take a step back. Structural improvement does not mean the cycle disappears. The earnings model for memory is still driven by pricing, supply, and demand. And right now all three are aligned in the best possible way, which is why these stocks have exploded. So pricing is elevated because supply is tight. Supply is tight because capacity was cut during the original downturn and hasn't fully actually even come back. And demand is so strong because hyperscalers are aggressively building this AI infrastructure. And that combination is creating extreme margins and extreme earnings for these select companies who effectively have a monopoly over the industry. So for example, right now Micron Technology is being framed as cheap. You'll hear people say it's trading at roughly four to five times forward earnings, which is true, and that the expectations are around $35 to $45 in earnings per share over the next year. On the surface, of course, that looks like an obvious and compelling buy. Because a company growing this fast, tied to AI trading at that kind of multiple should be a no-brainer. But that only works if you assume that those earnings are actually sustainable. And that's the key assumption people are making without actually thinking through what's driving those numbers. Because those earnings are not coming from a steadystate environment like we've talked about.

[00:07:28]They're coming from an exact period where pricing is elevated, supply is tight, utilization is high, and a product mix is improving all at the same time. That is definitely not normal.

[00:07:36]It's something we've never seen before and it's actually peak alignment. And to understand why this matters, we also have to understand how the memory companies actually generate their earnings. So the business itself is extremely sensitive to pricing. Revenue is essentially price multiplied by volume. But costs just don't move nearly as fast as pricing. So now when pricing is increasing, revenue jumps quickly while costs stay relatively stable, which is exactly why we're seeing these margins expand aggressively. And right now when we look at Micron with the gross margins pushing into the 70 to 80% range, it seems rather absurd. But the reverse can also be true. If pricing comes down even modestly, revenue declines while costs remain relatively fixed in the short term. And this would compress margins much faster than people expect and ultimately compress the share price as well. And this is exactly why memory earnings are so volatile and why they have been for prior cycles. It's not because demand disappears right away. It's because small changes in pricing have an outsized impact on profitability. That's a very important point. So, let's actually put some numbers around this because this is where again some more misunderstanding becomes very obvious. If Micron is earning around $40 per share at peak margins, that assumes those elevated pricing conditions will in fact continue. But if gross margins move, say from 75% down to 55%, which is historically strong earnings, they don't just dip slightly, they compress meaningfully. that $40 earnings per share can quickly become $25 earnings per share or even closer to $20 depending on how pricing moves across DRAM and NAM products. And now the valuation looks extremely different.

[00:09:03]That four times forward earnings that you thought was so cheap is no longer viable. It's now suddenly 8 to 10 times earnings on a more normalized earnings basis. [music] So the stock didn't actually even get more expensive. The earnings just got less sustainable. And that's the part people and young investors miss. The multiple looks low because the E is actually inflated. So the right way to think about this is not asking whether Micron is cheap today because structurally it does look that way. The real question is what level of earnings are you actually going to underwrite into the future. If you believe these current conditions, tight supply, strong pricing and peak margins are sustainable, then yes, it looks cheap. But if you believe those conditions normalize even slightly, then the earnings power comes down and that valuation becomes much less compelling.

[00:09:46]And so this is where discipline comes in as investors because like I've said in memory, you do not need demand to collapse, for numbers to come down. You just need simply for supply to catch up a bit or pricing to soften modestly. So when you zoom out, the super cycle debate becomes much clearer. Yes, the floor is structurally higher. AI demand is real and it's not necessarily going away. The cycle could last longer because supply is harder to scale and more capital intensive. But the idea that these peak conditions become the new baseline is actually where this math might break. And that's what this market is going to start to assume at some point. And you do not want to be on the wrong side of it. So what we also have to keep in mind is is when expectations have shifted towards peak pricing, peak margins, and continued acceleration all holding at once, the setup actually becomes more fragile, not less. And that's not because the story is even wrong. It's because the expectations are too clean, too linear for a business that has never behaved that way in history. Now to understand this even further, we can talk about SanDisk. This was where the story will actually simplify. SanDisk is essentially a NAND and storage business. So when you look at it, you're looking at a much more traditional way of looking at the memory cycle. And the numbers make that very obvious. They just reported roughly 5.95 billion in revenue, which is up 250% year-over-year and nearly 100% sequentially with gross margins around 78% and about 3.6 billion in net income in a single quarter. And for future growth, they're also guiding the next quarter to around 7.75 billion to 8.25 billion in revenue with 79 to 81% gross margins, which is essentially unheard of in this industry. That is definitely very extreme. These are not steady state numbers and those are what peak pricing and full utilization looks like in memory. And nan pricing has clearly moved. Inventory has normalized and capacity is definitely being absorbed.

[00:11:33]But again, when all three happens at once, earnings explode just like this.

[00:11:38]But when we compare this to Micron, Micron is putting up the similar strong growth which we've seen, but the product mix is actually very different. It's not just NAND, it's DRRAM and now HBM and data center exposure layered on top of that. So that's why Micron's revenue is scaling even faster and that's why it's a bigger company and why the narrative around it is even stronger. It's benefiting from both the cycle and the structural AI demand. SanDisk, on the other hand, is mostly the cycle with a little bit of AI demand as well. And that's the key difference. So when you see SanDisk's margins go from roughly low 20% range last year to a high 70% today and [music] earnings swing from losses to billions in profit. That is not a structural transformation. That's pricing doing the majority of the leg work and it tells you where we are in the cycle. Supply was tight like I've said pricing moved and now we're seeing peak profitability flow through. So the way to think about this is very simple as investors. SanDisk would be your more clean read on the actual cycle itself.

[00:12:32]And Micron is that same cycle but with a structural AI tailwind layered on top.

[00:12:36]If you're bullish, the argument is that AI demand will keep pushing the cycle higher and longer than normal, meaning pricing will have to stay elevated.

[00:12:44]Margins will have to stay very strong and companies like Micron will continue to outperform while even SanDisk holds better than prior cycles. So in that scenario, the earnings power is higher than what people would actually even assume and what many Wall Street analysts are actually saying and multiples would actually expand even higher from here and you the prices would run even further. But if you're more cautious, the argument is that SanDisk is already showing us what peak conditions look like. So it really comes down to what you actually believe. If you think this is a longer stronger cycle driven by AI, you're going to lean into it, especially on names like Micron because they're the premium producers.

[00:13:17]If you think this is peak alignment of cycle plus AI, then you recognize the numbers are great, but the riskreward would no longer be early. So when we step back and we strip out the noise, the picture is actually very clean. The memory industry is no doubt fundamentally better than it was in prior cycles. AI is not a short-term demand spike. It's a real shift in how compute works, and some are saying it's obviously a technical revolution, and memory just so happens to sit directly in the middle of that. Systems definitely need more of it. They need faster versions of it and the buyers are definitely stronger and that's why companies like Micron and SanDisk are seeing demand across every segment like I said between HBM, Drram, NAND, data centers all at the same time. That definitely didn't happen before and it's exactly why it's confusing some very smart investors. So both things are actually true at the same time which actually makes this way trickier. The businesses are structurally improving.

[00:14:10]The current earnings are still cyclical but they're also elevated. And the mistake the market is trying to make is blending those into one idea. And they're treating peak earnings as if they are the new baseline. And that's exactly where the risk will lie. So the way to think about this positioning is very simple. If you believe on one hand that AI demand keeps accelerating, HBM remains constrained and supply takes longer than expected to catch up, then you stay bullish, especially on higher quality names like Micron because the cycle stretches and earnings will stay elevated for longer and they can reinvest that cash flow and give it back to shareholders. But if you believe that hyperscaler demand starts to even normalize a little bit, supply ramps into 2026, maybe 2027 or 2028, so we're looking even further ahead, and pricing softens even modestly, then that's your sign to get out, and you're likely close to peak earnings rather than early in earnings, and the easy money has already been made. And this is exactly where the riskreward that we so often talk about at TLI becomes less attractive at current levels. So, this isn't a bull versus bare call. It's a timing and expectations call. The story is definitely right. The numbers are structurally real, but the question is, are you going to be buying the long-term structural shift or are you actually buying the best part of the cycle? So, as always, if you found this helpful, subscribe for more breakdowns like this and feel free to drop a comment with any other companies or sectors that you want me to cover. And don't forget to click the Patreon link below. We'll see you again next week.