Hey everyone, welcome back to Chip Stock Investor. This is another episode in our series in how to invest in the semiconductor industry, and we're gonna be focusing on IDMs integrated device manufacturers, but specifically we're gonna be focusing on the memory chip market in 2025. Watch out for these three key points. What are memory chip IDMs and how do they make money? And memory has changed a bit in the AI era. It's almost seen as a bottleneck. Surely that's a reason to invest in memory IdMs, is it? And finally, we'll focus on the strange things that are happening in memory. Is memory more than just a commodity? Here is a look at our semiconductor industry flow. The IDMs, integrated device manufacturers. This is the old semiconductor industry business model. This used to be the supply chain. IDMs, think like Intel, or Micron. They made some of the tools, they did the engineering, they actually manufactured the chips and then they sold the chips. But as the industry has gotten more complicated, gotten more expensive, companies have specialized and disaggregated, out of the IDM model. So this is why they're called IDMs. These companies still engineer and a manufacturer and sell the chips. Memory IdM sales will make up around $185 billion in 2025. That's a mid-teens percentage growth, over 2024. Previous to that, in between 2023 to 2024, they had over an 80% growth rate at that time because we were coming from the depths of the bear market. Let's get that first point out of the way. How do IDMs make money? IDMs design and manufacture the chips and then they sell to their product partners. Some IDMs focus on off the shelf or merchant silicon. Others provide custom engineering to meet specific customer design needs. And barriers to entry or, sometimes often referred to as a moat for these businesses, we're going to rank as medium to high. We're gonna say medium because as you'll find, a lot of IDMs, including the memory chip makers, all make basically the same thing. And the reason for that is because anybody can go out there and, acquire the IP from other companies. And also purchase the equipment from the fab equipment makers to fire up a manufacturing facility, a fab, and start cranking out memory chips. We've seen this, especially in China, a lot of new memory chip makers have popped up in China. However, we can also rank this as a high barrier to entry because, purchasing that ip, engineering it into a chip, acquiring the equipment to, to start a fab, billions and billions of dollars to get all of that up and running. So it's not like anybody can just enter this market. So you could view this as a high barrier to entry, but is this actually a moat? Is this in of itself a reason to invest for the long term? Before we continue, I wanna remind you to hit that subscribe button and consider a membership to Semiconductor Insider. Link In the description below, there's going to be a lot of slides and a lot more explanation of the memory market in general, in a published manual that goes along with this video only available to our Semiconductor Insiders. Check out the link below in the description. There's three main parts of a computing system. There's the logic, that's the actual compute power, the GPUs or CPUs that do the thinking part of the computing system. Networking, that's where all the information gets moved between different parts of the network, like between the logic chips from data storage to logic or into and out of the data center. And then there's storage, that's what we're focusing on today. Where information is stored either in queue waiting for compute processing. Or in long-term storage that's gonna be used later on command. With storage being a main part of the computing system, why is it such a commodity? Yeah. Each of these on its own, are a commodity, right? I think oftentimes, a lot of investors hear the word commodity and they think cheap, non-differentiated piece or ingredient, and that's not entirely true. Sometimes you can just simply call a commodity something that is an input, used to make a more complex thing. And so all of these, even A GPU, all by itself is a commodity. NVIDIA designs commoditized parts, so just making a storage device as complex a process as it might be, it is still simply just an ingredient that a company needs to complete a whole product or a whole service. So that's why we say these memory IDMs produce commodities and this is a key reason why they may not have as competitive a moat or a as high a barrier to entry as you may think. And it's a reason why they haven't been the greatest of ultra long-term investments. And I think as we discuss the different parts and pieces of the memory market by product type or very broad product type, I think this will become apparent if you just look at these companies and look at the historical stock price over the course of 20, 30 years, if the company has that type of history. It's a mixed bag to say the least. And the DRAM market is the largest portion of the memory market. Players like Sk Hynix, Micron, and Samsung, those are the three biggest. And then we have other IDMs, Infineon and CXMT in China. Yeah, and we call out Infineon in particular because a lot of the IDMs that you may think of when you think of auto, industrial, or power chips also have segments that developed some sort of dram or maybe even flash memory. Sometimes they sell that as a discreet memory, or sometimes it's embedded directly into the systems that they help design. A lot of IDMs here participate in this market, but as you said, the three big ones that you need to know about in dram, SK hynix and Samsung and South Korea and Micron in the US. There's a lot of excitement surrounding the memory market because of DRAM chips, specifically because of high bandwidth memory. High bandwidth memory, our HBM is used specifically with those accelerated computing GPUs like Nvidia has. And what is HBM exactly? We have a few illustrations for you from SK hynix as well as from Micron to illustrate what exactly HBM is. It sounds pretty high tech and it is, but what it is at its base level is DRAM core chips stacked on top of each other. And so this is why the DRAM market has come into the spotlight as a high growth portion of the memory market. It's really just a subset of DRAM. HBM rapidly ramping up to help supply those accelerated computing systems, and there's a lot of technology that goes into these things. They get co packaged close to the logic chip, and so this is why some people say, Hey, HBM has changed the game. Memory is no longer a base commodity. However, remember the fact remains and Kasey, you talked about this in your SK hynix video and deep dive. The complexity of the ingredient doesn't change the fact that it is still an ingredient, a commodity, and as a result, there are still going to eventually be some cyclical swings up and down in HBM revenue. But let's jump to the next big portion of the memory market, NAND. This is a different type of memory. It's used for longer term storage, typically anyways, used for longer term storage. You see some overlap. The dominant DRAM players are also the dominant players in NAND. SK hynix, and Samsung and South Korea and Micron in the United States. But we have a couple of other businesses in NAND chips as well. Yeah, we've talked about recently Kioxia. They have a partnership with one of our portfolio holdings, Pure Storage, and then SanDisk, which has been spun off from Western Digital. And then on the bottom left you see the logo for Yangtze Memory Technologies, that company is based in China. Nick, you mentioned that NAND is primarily used for longer term storage. How is it manufactured differently than dram? You can check out some of our previous videos on this. When we especially discussed Lam Research and the equipment technology they've developed. For things like NAND flash. So DRAM dyes get stacked on top of each other to make HBM. NAND is also built in layers, but it's actually manufactured that way. These are not individual chips or die that are then stacked on top of each other. Different process here that makes these layers of different silicon that ultimately create these cells that store the data. A lot of work has gone on in the last decade to increase the density, the amount of data that can be crammed into a NAND chip. A lot of progress being made here. Typically, longer term storage does have that bottleneck where not enough memory can be moved out of the long-term storage directly to the logic chips. And that's where DRAM kind of acts as almost like a buffer between the two. It's a short term memory device where data is held in queue before getting computed. But some interesting things happening here that could change the game. But before we get to that, let's talk about the cyclicality inherent with all of these businesses, because there is a pitfall here for investors to be aware of before you jump on board thinking that you know, the game has changed, these are no longer commodities. These are gonna be great long-term investments. SK hynix has been a bit of a standout in the memory market. We have revenue from the company from 2013 to 2024, and you can see pretty much SK hynix has been able to go up and the right in their revenue growth. However, one thing we do need to point out is that investors did seem to pile in in 20 23, 20 24, thinking that high bandwidth memory is gonna be the new thing, it's gonna continue with a high growth rate. Between 2023 and 24, there was a high growth rate in the memory market, however, that was at the bottom of the bear market. So coming out of that, of course there's going to be a very high growth rate. Yeah, HBM is not going to make these businesses hyper-growth companies for the foreseeable future. They're going to remain cyclical. And as you mentioned at the outset of this video, Kasey, the memory market is expected to grow at a much more normal rate of low to mid teens percentage year over year in 2025. So HBM not a cure all for this cyclical looking chart. Let's look at Samsung very briefly. Their semiconductor business, which is just one part of Samsung, is a bit opaque. The financials are difficult to read, but, some of these bars are goofy looking because in blue prior to 2021, Samsung Semiconductor also included the Samsung Display manufacturing business. They separated that out starting in 2021, and so those gray bars starting in 2021 and onward are just the semiconductor business. However, even that is still kind of opaque because there's a lot going on in Samsung semiconductor revenue. There's Samsung Foundry, which competes with of course TSMC, Taiwan Semi Manufacturing and Intel's Foundry business. And then of course there's the segments that manufacture DRAM and NAND chips. And Samsung doesn't specifically call out how much revenue they're making from each. They just kind of give a, a market share breakdown. So keep that in mind. There's a lot going on in Samsung. We wouldn't recommend you invest in Samsung just because you think they're going to catch up to SK hynix or even Micron in the high bandwidth memory race. We've done a lot of videos about Micron in the past, talking about their cyclicality. We still have a small position of Micron in our portfolio, but you can see over the years extreme cyclicality in the memory market. Yeah, and the problem with that cyclicality is in a good year, Micron is very profitable, but in a bad year, like their fiscal 2023, even fiscal 2024, the company can flip, to losses very deep in the red losses. And so you have to be aware of that. The best time, to buy stocks like this are probably very early on or before a new growth cycle takes off. As a lot of investors were, very rudely reminded of, especially at the tail end of 2024, any change in revenue trajectory or even a perceived change coming in, in revenue, trajectory for a company like Micron can send the stock price crashing from peaks. Now let's look at the revenues of a couple of NAND storage companies, Kioxia and SanDisk. Kioxia recently had an IPO. The history is a little bit complicated for actually both Kioxia and SanDisk. Yeah, complicated to say the least. Kioxia which focuses on NAND or flash was originally part of Toshiba, the big Japanese conglomerate. Toshiba spun off its hard disk drive business and sold its NAND flash business to a consortium of investors led by Bain Capital. And now it's been IPO'ed and late 2024, but its ownership structure is still complicated. Bain Capital is still a controlling shareholder, as is Toshiba. Toshiba still has a stake in Kioxia and SK hynix also owns a bit of Kioxia as well, especially via some convertible debt that is on Kioxia's as balance sheet. SanDisk actually was also an independent company up until about 2016 when it was acquired by Western Digital, the hard disk drive company. Actually at that time, SanDisk's, CEO, was Sanjay Mehrotra. You probably know him now as the CEO of Micron, but Western Digital has decided to spin SanDisk back off as an independent company. As of the first quarter of 2025, and in the midst of all of this NAND chip maker IPO talk, there was a little bit of a wrinkle here when at the time Western Digital, which still owned SanDisk, offered to acquire Kioxia from that consortium of investors at Bain Capital. Ultimately, they decided to just forge ahead with their own IPO and then Western Digital said, okay, we'll IPO San Disk as well. But there is a strategic partnership between these two NAND pure plays. It could make for possibly at some point in the future, some sort of interesting, merger between the two, or maybe that doesn't happen at all. Maybe sk hynix acquires Kioxia, who knows what could happen here. But this is the current state of the NAND market. It's an exciting place of technological development, but it has gotten a little bit crowded, I guess you could say as of late. Let's round out this discussion with our third point. Are strange things really afoot in memory land. We've discussed the fact that memory is a commodity. we've settled that, at least in our minds, but what are some things that are changing in the memory field that could make it more interesting? Yeah, the re IPOs are definitely kind of a curve ball here as of late, but some new technological, developments. And I, this really also goes back to the fact that memory is ultimately a commodity. It's just an ingredient. And companies that make the total computing system have lots of options to choose from. So SanDisk, around the time of its. IPO had an investor day event where they talked about this reimagining of. NAND which again, remember is typically used for long-term memory and the reason why you would use it for long-term memory versus something like DRAM co packaged really close to the logic chips is because there is that bandwidth bottleneck. However, SanDisk thinks it has come up with a solution there on the right of this chart that could rival HBM, which is those DRAM chips stacked on top of each other, co packaged with things like GPUs. HbF, high bandwidth flash, move over HBM. . We don't have a ton of details on this, but it looks kind of similar to HBM, HBF or high bandwidth Flash Stack one on top of the other. But we'll have to see how this technology continues. Yeah, check out Semi Insider, we'll be talking about this as we get more details on what exactly this is and if SanDisk is about to get like a big co packaged, memory deal with Nvidia or Broadcom or AMD or whoever, exciting things. Now there is this other part of the memory market, HDDs, hard disk drives. Seagate and Western Digital dominate this space. Toshiba still has a bit of a hand in this, but by and large, we think the average investor, can probably just stay away from HDDs for now. The reason being is this revenue chart. Seagate one of the leaders in hard desk drives, also Western Digital. Obviously the flash business is now SanDisk at stand alone but the blue bar, their HDD revenue, kind of a similar picture. Tough to manage an investment in a business like this if this isn't your day job, if you're trying to pick long-term winners, so probably just stay away from HDDs, it would be the, the blanket catchall here. As we mentioned, there are a lot of other ways to invest in the memory market without this extreme cyclicality. These other areas of the semiconductor industry, chip fabs, wafer fab packaging and assembly equipment, fabless chip designers, and of course EDA and IP companies. If you haven't caught that video that we did on how to invest in ip, that's one of the most underrated areas of the semiconductor industry and one that you absolutely need to know about. So we'll link that video right here. Until then, we look forward to seeing you over on Semi Insider where we can talk a lot more about memory chip companies, investing in them and what's coming down the pike for more memory technology and how that is going to change the landscape for computing. See you over there. Thanks everybody. Make sure you sub at the channel, hit the like button, share with your friends, and we'll see you all again soon at Chip Stock Investor.