WEBVTT - The Memory AI Trade 0:00:06.040 --> 0:00:06.640 Loking to train. 0:00:06.680 --> 0:00:11.160 So I'm Joel Webber and I'm Eric Belchernas. 0:00:11.800 --> 0:00:16.400 Eric AI trade has been a constant for about two 0:00:16.440 --> 0:00:19.360 or three years now, but today we're going to talk 0:00:19.400 --> 0:00:22.640 about a little corner of it that has become very interesting. 0:00:23.239 --> 0:00:27.000 Yeah, it's almost like downstream from AI, these AI data 0:00:27.080 --> 0:00:30.240 centers which they're building like crazy. This isn't news to anybody. 0:00:31.080 --> 0:00:35.199 They need memory and high bandwidth memory in particular. And 0:00:35.240 --> 0:00:38.760 there's like three companies that have almost the entire market 0:00:39.040 --> 0:00:42.920 s k Heiniz, Samsung, and Micron. Two of them are 0:00:42.960 --> 0:00:47.519 South Korea, one US. Now, there was a push though 0:00:47.560 --> 0:00:49.320 all three of those thocks are up a ton, so 0:00:49.360 --> 0:00:52.040 people were buying the South Korea ATF and we're seeing 0:00:52.320 --> 0:00:55.400 an inordinate amount of flows into that ETF earlier in 0:00:55.440 --> 0:00:58.120 the year and last year. But you don't get Micron 0:00:58.200 --> 0:00:59.960 with that, but you still get too at pretty good way. 0:01:00.680 --> 0:01:02.720 But you also get a lot of like other South korea'stocks, 0:01:02.760 --> 0:01:04.400 so like you're going to get K pop stuff in there. 0:01:04.600 --> 0:01:06.400 It's like a proxy, but not a one. 0:01:06.560 --> 0:01:09.280 Then Roundhill saw this, which again heads up play and 0:01:09.319 --> 0:01:13.200 they put out a memory ETF recently, very recently, and 0:01:13.240 --> 0:01:16.160 it just blew up immediately. We're talking two hundred million 0:01:16.200 --> 0:01:19.120 dollars of trading right off the bat. Now that's normal 0:01:19.160 --> 0:01:21.640 for like the spot bitcoin ETF for or the first 0:01:21.640 --> 0:01:24.440 ever like Ether ETF, that was pretty normal. But you 0:01:24.440 --> 0:01:28.119 don't see that ever, especially for them ETF. Normally you're 0:01:28.160 --> 0:01:31.360 lucky to trade two million, not two hundred million. I'm like, 0:01:31.360 --> 0:01:34.520 what is going on here? So I figured out the 0:01:34.560 --> 0:01:38.520 two South Korea companies aren't in ADR format yet, meaning 0:01:38.560 --> 0:01:40.880 you can't really buy them in the US, so your 0:01:40.880 --> 0:01:42.640 best way is to use the South Korea ETF, and 0:01:42.680 --> 0:01:44.800 that doesn't have micron. So this has let's give you 0:01:44.840 --> 0:01:48.800 all three in about a sixty five percent waiting and 0:01:48.920 --> 0:01:51.720 people were like going crazy for it. They must have 0:01:51.840 --> 0:01:54.800 let the market know. And it's at this point over 0:01:54.840 --> 0:01:57.840 a billion dollars now to keep that, to put that 0:01:57.840 --> 0:02:01.720 in perspective, the top four biggest the matic ETFs, it 0:02:01.760 --> 0:02:04.840 took them all over one thousand days. In some three 0:02:04.840 --> 0:02:08.120 thousand days, they hit a billion. So d Ram pulled 0:02:08.240 --> 0:02:11.400 off in ten days. So again we have a massive 0:02:12.040 --> 0:02:16.040 outlier smash hit in her hands. So I went and 0:02:16.160 --> 0:02:20.400 worked with the Bloomberg intelligence analyst who's a specialist to 0:02:20.080 --> 0:02:22.120 write a three Ways the Play note, which is one 0:02:22.120 --> 0:02:24.600 of our new recurring notes, and we're here to talk 0:02:24.639 --> 0:02:25.000 about it. 0:02:25.440 --> 0:02:29.040 That ETF by Roundhill is called d ram and here 0:02:29.240 --> 0:02:33.600 to speak with us about it is Jake Silverman, who's 0:02:33.600 --> 0:02:38.200 a semi conductor analyst at Bloomberg Intelligence. He's going to 0:02:38.200 --> 0:02:44.959 cover more than that as well, this time on trillions 0:02:45.440 --> 0:02:48.079 the memory trade. Welcome to trillions. 0:02:48.200 --> 0:02:49.160 Yeah, thanks for having me. 0:02:49.520 --> 0:02:52.040 Okay, So what Eric just laid out for us is 0:02:52.120 --> 0:02:57.720 a wildly successful ETF launch. It's been pretty interesting times 0:02:57.720 --> 0:03:01.240 in the semiconductor business. Why did it take so long 0:03:01.480 --> 0:03:03.480 for an ETF like this to come to market? Do 0:03:03.480 --> 0:03:03.720 you think? 0:03:04.280 --> 0:03:04.440 Yeah? 0:03:05.600 --> 0:03:07.800 Yeah, I think what Eric pointed out it was interesting 0:03:07.880 --> 0:03:11.200 is that the lack of ADRs, so the South Korean 0:03:11.240 --> 0:03:14.880 companies specifically in this case Samsung and s ke Heiniks 0:03:14.919 --> 0:03:19.040 really haven't had that ADR format before, and so I 0:03:19.080 --> 0:03:21.799 think there was just kind of folks are trying to 0:03:21.840 --> 0:03:25.160 understand the best way to play it, and so creating 0:03:25.160 --> 0:03:29.160 this ETF format was probably just sort of the key 0:03:29.200 --> 0:03:32.720 to unlock that. And so there have been a couple ETFs, 0:03:32.720 --> 0:03:35.640 as you pointed out, based on Korea, and you know, 0:03:35.720 --> 0:03:39.800 Korean stocks and et cetera. And so it is interesting 0:03:39.880 --> 0:03:42.800 now because there's such a memory up cycled, there's just 0:03:43.000 --> 0:03:44.640 such a renewed interest in memory. 0:03:44.760 --> 0:03:47.360 Okay, so when we think about chips most of the time, 0:03:47.880 --> 0:03:52.040 especially with AI, we're thinking, like Nvidia, maybe the Google stuff. 0:03:52.600 --> 0:03:55.720 Memory is sort of old school. I kind of forgot 0:03:55.760 --> 0:03:57.760 about it, to be honest with you, Like, why is 0:03:57.800 --> 0:03:58.560 this taking off? 0:03:58.840 --> 0:04:01.400 Yeah? I would say memory is usually considered the boring 0:04:01.520 --> 0:04:05.040 aspect of semiconductors, but it's gotten a lot more exciting, 0:04:05.080 --> 0:04:06.880 and a lot of it is to do with high 0:04:06.880 --> 0:04:11.080 bandwidth memory. So basically, you take these typical dram dyes 0:04:11.120 --> 0:04:13.840 that you would get off of just any wafer that 0:04:13.880 --> 0:04:16.160 would come out of these fabs, and you stack them 0:04:16.279 --> 0:04:18.760 up and you use these things called through silicon via 0:04:18.839 --> 0:04:21.320 to connect them microbumps, and you put them on a 0:04:21.360 --> 0:04:25.280 logic dye that then connects over to the GPU. It's 0:04:25.360 --> 0:04:26.400 super complicated. 0:04:26.440 --> 0:04:29.920 It's way more interesting than can we get that in English. 0:04:30.080 --> 0:04:33.599 It's basically just stacking DRAMs that you're maximizing the amount 0:04:33.600 --> 0:04:36.159 of space you get on a GPU board, and so 0:04:36.279 --> 0:04:39.520 this way you get super high bandwidth that you can 0:04:40.200 --> 0:04:42.800 by maximize. I don't want to get too into the 0:04:42.800 --> 0:04:45.679 weeds with it, but it's basically maximizing the bandwidth based 0:04:45.720 --> 0:04:48.240 on a limited amount of space you have on a 0:04:48.279 --> 0:04:49.359 package for a GPU. 0:04:50.400 --> 0:04:53.400 In general, the idea is if people are building AI 0:04:53.480 --> 0:04:55.960 data centers and they want to store all this data, 0:04:56.440 --> 0:04:59.480 they clearly want to they need memory to do that, 0:04:59.720 --> 0:05:01.320 and they I want it to be in a small, 0:05:01.520 --> 0:05:04.800 smallest physical form possible. So there's just just this a 0:05:05.000 --> 0:05:07.680 demand and then be pressure to make it small. And 0:05:07.720 --> 0:05:09.680 so that's what's going on, is that right? 0:05:09.839 --> 0:05:10.080 Yeah. 0:05:10.080 --> 0:05:11.839 So one of the things that probably to keep in 0:05:11.880 --> 0:05:14.200 mind is when you think about some of these models 0:05:14.240 --> 0:05:15.880 like from open AI that come out and they have 0:05:15.960 --> 0:05:19.359 like trillion plus parameters, you have to store that memory. 0:05:19.680 --> 0:05:22.800 And then when you're actually interacting with a model. So 0:05:22.920 --> 0:05:25.320 let's see if you're on chat, GBT or claude or 0:05:25.320 --> 0:05:28.000 whatever and you're actually typing into it. So what it's 0:05:28.040 --> 0:05:30.600 doing is there's this thing called kb cash where it's 0:05:30.600 --> 0:05:33.680 actually storing the amount of data and then it's appending 0:05:33.720 --> 0:05:36.000 it over and over again. But what you want to 0:05:36.040 --> 0:05:37.800 do is so that the model is don't have to 0:05:37.839 --> 0:05:41.360 run through your entire context window every time you're typing 0:05:41.560 --> 0:05:43.960 and keep doing it over and over again, it'll store 0:05:44.000 --> 0:05:47.239 that data. So basically what happens is you need memory 0:05:47.279 --> 0:05:49.880 to interact with the model, and then when you're adding 0:05:49.920 --> 0:05:52.400 to the context for the model, it needs to interact 0:05:52.440 --> 0:05:55.880 with that as well. So basically, the more the larger 0:05:55.920 --> 0:05:58.400 these parameters are, the larger these context windows, the more 0:05:58.440 --> 0:05:59.080 memory you need. 0:05:59.279 --> 0:06:01.880 You're in equity analyst first and foremost. And we've got 0:06:01.960 --> 0:06:06.200 three companies here esk Heinex, Micron and Samsung. Distinguish them 0:06:06.240 --> 0:06:09.120 from each other or are they literally just plug and 0:06:09.240 --> 0:06:10.479 play versions of each other? 0:06:10.560 --> 0:06:13.120 They're very different one I would say Samsung is the 0:06:13.120 --> 0:06:16.560 most differentiated because they have various business units. So when 0:06:16.560 --> 0:06:20.520 you're buying Samsung, for example, you're not buying just memory, 0:06:20.720 --> 0:06:23.760 You're buying their smartphone business, You're buying their foundry business 0:06:23.760 --> 0:06:26.200 and the various other businesses that they have. Within that, 0:06:26.760 --> 0:06:30.160 esk Heinex and Micron are considered, or at least more 0:06:30.200 --> 0:06:33.400 so in the past, we're considered interchangeable. But the difference 0:06:33.440 --> 0:06:36.560 between Micron and eske Heinix is twofold. One. You have 0:06:36.600 --> 0:06:40.120 the capacity that eske Heinix has, that's a lot more 0:06:40.240 --> 0:06:42.760 than say Micron, and then the other aspect is that 0:06:42.839 --> 0:06:45.320 they've been the leader on a high band with memory 0:06:45.360 --> 0:06:47.800 for a while now, so when you compare it to 0:06:47.960 --> 0:06:51.440 micron A, Micron was playing catch up and has been 0:06:51.480 --> 0:06:55.600 gaining share, but sk Heinix has still had about double 0:06:55.640 --> 0:06:58.200 that share over the last I don't know, say twelve 0:06:58.200 --> 0:07:02.200 months or so roughly, So it's a question of technology leadership. 0:07:02.279 --> 0:07:04.760 Micron is still definitely in the mix. And then there's 0:07:04.800 --> 0:07:07.159 a capacity element which we can get into in a bit. 0:07:08.080 --> 0:07:08.279 Eric. 0:07:08.320 --> 0:07:10.040 I'm not gonna ask Jake this because I think he knows. 0:07:10.080 --> 0:07:11.280 Do you know where Micron is based? 0:07:11.520 --> 0:07:17.640 I would guess like Seattle, No, you'll never guess, West Virginia, No, Mississippi, 0:07:17.680 --> 0:07:21.440 No Ohio, Missouri, no. 0:07:22.080 --> 0:07:27.560 Louisiana, all over the place. Idaho? Oh yeah, was that next? 0:07:27.800 --> 0:07:30.240 I mean the only thing I actually drove through Idaho 0:07:30.280 --> 0:07:32.840 once on a road trip, really cool, just long stretches 0:07:32.840 --> 0:07:36.360 of road built to spill is a famous indie rock 0:07:36.400 --> 0:07:37.960 band from Idaho. That's the only thing I think of 0:07:37.960 --> 0:07:40.440 in Potatoes, Potatoes and Micron. 0:07:41.840 --> 0:07:42.760 Like like that. 0:07:42.880 --> 0:07:47.720 Yeah, that is wild. Yes, I mean, look this industry, obviously, 0:07:47.800 --> 0:07:52.200 these companies have been around, but the AI demand you 0:07:52.280 --> 0:07:55.920 have here that it's seventy five billion dollars this memory industry, 0:07:56.560 --> 0:08:00.000 and you think it's going to double in only two 0:08:00.040 --> 0:08:03.480 two years to one hundred and thirty five billion. That's 0:08:03.480 --> 0:08:04.360 pretty wild. 0:08:04.640 --> 0:08:07.360 Is that bookcase or barecase? 0:08:07.680 --> 0:08:10.120 So I think in the past, I would have just 0:08:10.160 --> 0:08:13.160 talked about memory as an entire industry, and now it's 0:08:13.200 --> 0:08:15.800 like you have two product sets. You have high bandwidth memory, 0:08:16.080 --> 0:08:19.320 which is just completely outpacing I would say at least 0:08:19.440 --> 0:08:21.800 unit growth the thing that's gotten. So it's kind of 0:08:21.800 --> 0:08:24.840 important to understand. So because I was talking about the 0:08:24.880 --> 0:08:27.720 structure of high bandwidth memory, it actually requires three to 0:08:27.800 --> 0:08:30.880 four times the amount of space per wafer compared to 0:08:30.920 --> 0:08:33.720 the traditional products that we see that it's like you'd 0:08:33.720 --> 0:08:36.120 have in your smartphone or your laptop, anything like even 0:08:36.240 --> 0:08:38.880 traditional servers that go into like say the cloud or something. 0:08:39.400 --> 0:08:42.360 So what's happening is there's a limited amount of capacity. 0:08:42.240 --> 0:08:47.000 They're consistently continuously upgrading that amount of capacity. We don't 0:08:47.000 --> 0:08:49.320 have to get into the specifics of how they do that. 0:08:49.600 --> 0:08:53.720 But because you shift those wafers over to high bandwidth memory, 0:08:53.880 --> 0:08:58.080 you're taking away capacity from other products. And memory is 0:08:58.280 --> 0:09:01.720 very supply and demand driven way more commodity like than 0:09:01.800 --> 0:09:04.920 the rest of the semiconductor market. I mean, obviously GPUs 0:09:04.960 --> 0:09:08.480 are the least commodity like at this point, but memory 0:09:08.559 --> 0:09:10.960 is the most commodity like. So if you're taking off 0:09:11.200 --> 0:09:14.960 supply and demand keeps going up, pricing is going to 0:09:15.000 --> 0:09:17.760 increase at a fairly high rate, which we've already seen 0:09:18.000 --> 0:09:20.800 and this year alone is probably going to increase anywhere 0:09:20.800 --> 0:09:23.360 between one hundred and two hundred percent, if not more. 0:09:23.720 --> 0:09:26.000 That's just d ram Land is also similar. 0:09:26.320 --> 0:09:29.120 Well, when we hear all these big tech companies, some 0:09:29.160 --> 0:09:32.320 of the mag seven are spending all this cash, even 0:09:32.360 --> 0:09:35.360 taking out debt to spend cash on AI. This is 0:09:35.400 --> 0:09:37.559 where they're This is where they're spending it, right. It's 0:09:37.559 --> 0:09:39.880 not rocket science here. They got to they got to 0:09:40.120 --> 0:09:41.120 spend it on stuff. 0:09:41.760 --> 0:09:43.839 By the way, you said d RAM there, you were 0:09:44.040 --> 0:09:47.520 referring to the memory itself, not the ticker, but the ticker. Yes, 0:09:48.520 --> 0:09:50.200 we gotta talk about the ticker for a second. When 0:09:50.240 --> 0:09:52.480 you saw that ticker, d RAM, what did you think. 0:09:52.679 --> 0:09:56.080 I thought, this is the best name for an ETF 0:09:56.120 --> 0:09:57.679 for memory I could come up with. 0:09:58.200 --> 0:10:00.600 I honestly called it DRAM for a while, and then 0:10:00.640 --> 0:10:01.880 I got educated. 0:10:02.360 --> 0:10:06.240 Well, you're still still working on that, so Idaho and Drama. 0:10:08.320 --> 0:10:09.480 It's an ongoing education. 0:10:21.480 --> 0:10:24.040 So when you look at what this ETF holds, by 0:10:24.040 --> 0:10:26.280 the way, Eric, we've got those three companies. What else 0:10:26.360 --> 0:10:27.040 is in this sticker? 0:10:28.480 --> 0:10:30.280 Here are the rest of the holding. So those three 0:10:30.400 --> 0:10:34.880 stocks make up sixty seven, so two thirds of the portfolio. 0:10:34.880 --> 0:10:38.440 Then we got forgive me on my pronunciation. Okay again, 0:10:38.480 --> 0:10:39.280 I'm an ETF guy. 0:10:39.679 --> 0:10:40.440 Jake's gonna help you. 0:10:40.440 --> 0:10:44.079 Okay, Kyoksia, how's that pretty good? 0:10:44.120 --> 0:10:44.320 Good? 0:10:44.320 --> 0:10:49.880 Okay, Kia Holdings Japan, sand Disc, that's US another retro 0:10:49.960 --> 0:10:57.040 one Seagate, US Western Digital Retro, Nanya. Yeah, that's in Taiwan, 0:10:57.760 --> 0:11:01.319 Win Bond in Taiwan, and that's it. Those all the 0:11:01.400 --> 0:11:04.319 last ones I name probably collectively make up again to 0:11:04.520 --> 0:11:05.760 describe that portfolio. 0:11:05.880 --> 0:11:06.080 Jake. 0:11:06.320 --> 0:11:09.920 Yeah, So when you expand beyond eske Heinex, Samsung, Micron, 0:11:10.040 --> 0:11:12.520 you get into sort of more like nand specifics. So 0:11:12.600 --> 0:11:16.400 sand Disk and Kyoksa have a JV to do research 0:11:16.440 --> 0:11:20.000 and development and they share the same fabs together. But 0:11:20.080 --> 0:11:23.880 that's nand only. So the ticker for the ETF is 0:11:23.960 --> 0:11:26.720 d RAM. But within that you also have nand as well, 0:11:26.760 --> 0:11:31.120 because Micron, Samsung and Eskahannis make DIRAM and nand nand 0:11:31.200 --> 0:11:33.920 is more like storage. So if I if you bought it, 0:11:33.960 --> 0:11:37.840 buy an iPhone, you'll see like it'll have tiers of storage, 0:11:37.840 --> 0:11:40.880 for example, five hundred and twelve gigabytes, one terabyte that. 0:11:40.920 --> 0:11:43.280 Stuff that fills up with the gooey storage. Yeah yeah, 0:11:43.360 --> 0:11:45.080 I'm like, what's on my iPhone? 0:11:45.080 --> 0:11:46.800 It's like when I take too many pictures of my 0:11:46.880 --> 0:11:48.360 dog and now I need to back it up on 0:11:48.400 --> 0:11:51.280 the cloud. Yeah, and then I need more storage. So 0:11:51.280 --> 0:11:52.800 someone's got to buy that storage. 0:11:52.880 --> 0:11:56.400 Yeah yeah, Okay, So when you think about what we 0:11:56.520 --> 0:11:59.560 just rattled off, there, is there something that's missing or 0:11:59.640 --> 0:12:02.880 is this literally the perfect portfolio for this trade right now? 0:12:02.920 --> 0:12:05.240 It's pretty good. So non Ya Tech, for example, is 0:12:05.280 --> 0:12:09.360 pretty small. They are Taiwanese based, But there aren't that 0:12:09.480 --> 0:12:12.280 many pure play memory suppliers out there, and that's because 0:12:12.320 --> 0:12:16.160 memory has been historically tough industry, a lot of bankruptcy, 0:12:16.240 --> 0:12:19.520 a lot of consolidation. That's changed over the last two 0:12:19.559 --> 0:12:23.400 decades roughly as that's kind of stabilized, But there aren't 0:12:23.440 --> 0:12:26.560 that many players out there that exist today that simply 0:12:26.760 --> 0:12:30.040 just supply those products. Once you expand beyond that, you 0:12:30.080 --> 0:12:33.400 start to get into the suppliers of those businesses and 0:12:33.480 --> 0:12:36.400 those are a lot more diversified. So if you're looking 0:12:36.440 --> 0:12:39.200 for a specific way just to play memory, that's probably 0:12:39.240 --> 0:12:41.480 the best pure play opportunity that you're going to get 0:12:41.520 --> 0:12:44.080 that also has a little bit of diversification within there 0:12:44.080 --> 0:12:44.520 as well. 0:12:44.600 --> 0:12:48.440 You said that memory is like almost like a commodity, right, 0:12:48.480 --> 0:12:51.760 And if we think about how commodities trade, there are 0:12:51.800 --> 0:12:55.520 these fabulous booms and bus right, and like a commodity 0:12:55.559 --> 0:12:58.480 supercycle kind of means that everything's going up and then 0:12:58.520 --> 0:13:02.480 eventually the air comes comes out of the trade. Where 0:13:02.520 --> 0:13:06.679 are we in that supercycle commodity trade cycle? 0:13:06.679 --> 0:13:09.800 We're definitely a boom cycle. Where in that cycle is 0:13:10.000 --> 0:13:13.319 I think the million dollar question for every portfolio manager 0:13:13.360 --> 0:13:15.839 and retail investor out there, you know, the question is 0:13:15.880 --> 0:13:18.280 like are we in inning two? Are we in inning seven? 0:13:18.520 --> 0:13:20.959 Are we starting to approach the peak of the cycle? 0:13:21.400 --> 0:13:24.680 And that's where these stocks tend to You either tend 0:13:24.679 --> 0:13:25.920 to make a lot of money or you tend to 0:13:25.920 --> 0:13:28.320 get in a lot of trouble. There's no question right 0:13:28.320 --> 0:13:31.200 now that we're in an unprecedented upcycle like we've so 0:13:31.240 --> 0:13:33.079 we've done analysis where we kind of look at the 0:13:33.200 --> 0:13:36.280 length of these cycles and they typically last somewhere between 0:13:36.559 --> 0:13:39.560 I would say seven and ten quarters. We're well beyond 0:13:39.559 --> 0:13:42.559 that now based on the reporting and the guidance from 0:13:42.559 --> 0:13:45.679 these companies. So the question is, Okay, if we're beyond 0:13:45.760 --> 0:13:48.520 the sort of ten quarter roughly peak that we've seen 0:13:48.600 --> 0:13:51.520 in terms of that upcycle, can we go another ten quarters? 0:13:51.679 --> 0:13:55.920 The question is then becomes, okay, is AI demand enough 0:13:55.920 --> 0:13:58.360 to sustain that? And then if I'm if we are 0:13:58.400 --> 0:14:01.880 in a very commodity like industry, how to supply factor 0:14:01.920 --> 0:14:04.280 into this as well. Our base case has kind of 0:14:04.320 --> 0:14:08.280 been more like, okay, we think come maybe second half 0:14:08.320 --> 0:14:11.080 of twenty twenty seven, twenty twenty eight. There's a lot 0:14:11.120 --> 0:14:15.360 more of these green field fabs coming online. Historically, what 0:14:15.400 --> 0:14:18.959 you've seen is a lot of these fabs have unused space, 0:14:19.240 --> 0:14:20.800 and what you can do is you can install these 0:14:20.840 --> 0:14:23.680 tools and you can get additional capacity out of them. 0:14:23.720 --> 0:14:26.800 But now these fabs are essentially full, and we can 0:14:26.840 --> 0:14:28.560 get into the history if you want a little bit, 0:14:28.600 --> 0:14:31.160 but basically these fabs filled up. They need to build 0:14:31.200 --> 0:14:33.800 new fabs, but the lead time on fab construction is 0:14:34.000 --> 0:14:36.280 at least two years, and that's if you do it 0:14:36.400 --> 0:14:39.440 incredibly fast. So if you want to build it from 0:14:39.480 --> 0:14:41.440 the ground, if I have dirt that's just sitting there 0:14:41.480 --> 0:14:43.960 ready to go, if you've already leased that space the 0:14:44.000 --> 0:14:46.680 green field, then you want to actually construct the fab 0:14:46.760 --> 0:14:49.800 and you know, again they need to be absolutely perfect. 0:14:49.800 --> 0:14:51.880 You can't even have particles of dust in these things, 0:14:52.120 --> 0:14:57.240 so it requires a lot of really specific construction that's specialized, 0:14:57.280 --> 0:14:59.640 and then you need to get these tools like, for example, 0:14:59.680 --> 0:15:04.120 as mel Cell's EUV lithography machines. Those cost hundreds of 0:15:04.160 --> 0:15:06.960 millions of dollars. You have to get those in install them, 0:15:06.960 --> 0:15:09.520 and that's just one piece of equipment. So this is 0:15:09.560 --> 0:15:13.720 incredibly complex and takes probably at least three years really 0:15:13.760 --> 0:15:17.080 from just digging dirt actually getting these functioning, and then 0:15:17.120 --> 0:15:19.160 that's the first wafer that comes out. So it's a 0:15:19.240 --> 0:15:20.000 long lead. 0:15:19.800 --> 0:15:23.160 Time and we're talking lot of greenfield coming online. 0:15:22.720 --> 0:15:25.040 A lot of greenfield. So part of that is ske 0:15:25.160 --> 0:15:27.680 Heinez has a fab coming online pretty soon. They probably 0:15:27.720 --> 0:15:31.720 already just started getting wafers out. Micron is expanding in Boise, 0:15:31.960 --> 0:15:34.480 They're also going to expand in New York. Samsung has 0:15:34.480 --> 0:15:37.400 talked a little bit about it. But another underrated aspect 0:15:37.440 --> 0:15:39.920 is probably China. So there's a couple of fabs. There 0:15:39.920 --> 0:15:42.000 are a couple of companies that have fabs in China 0:15:42.360 --> 0:15:46.240 YMTC for nand and then CXMT for DRAM. They have 0:15:46.320 --> 0:15:50.280 been aggressively expanding their capacity and gaining market share over 0:15:50.280 --> 0:15:52.960 the last five or so. Longer than five years or so, 0:15:53.480 --> 0:15:57.400 and so basically does that start to impact the industry 0:15:57.440 --> 0:15:59.920 come and say twenty twenty seven, twenty twenty eight, because 0:16:00.040 --> 0:16:02.400 they don't have the market share yet that if they 0:16:02.480 --> 0:16:05.600 doubled their capacity over the next year or so, could 0:16:05.640 --> 0:16:09.240 that actually really drive negative pricing as opposed to not 0:16:09.360 --> 0:16:11.920 just slowing down the price increases we've seen. 0:16:12.520 --> 0:16:15.360 One interesting thing about this trade drool is that if 0:16:15.360 --> 0:16:17.600 you look at the other ETFs, you know, we look 0:16:17.600 --> 0:16:21.359 at these three companies. Obviously, DRAAM has sixty seven percenters 0:16:21.440 --> 0:16:25.760 of the three companies, and EWY has forty eight percent, 0:16:26.040 --> 0:16:28.200 but no Micron, but it's heavy, and there are other two, 0:16:28.800 --> 0:16:31.640 but then there's a huge drop off even the AI ETFs. 0:16:31.840 --> 0:16:34.640 There's one AIS that has all three but only eighteen 0:16:34.640 --> 0:16:37.240 percent total waiting. But here's what really shocked me, the 0:16:37.280 --> 0:16:40.960 semiconductor one like SMH which everybody uses, it only has 0:16:41.440 --> 0:16:43.760 micron and it's only at like a five percent waiting, 0:16:43.800 --> 0:16:46.240 so you miss a lot. And then here's why I 0:16:46.320 --> 0:16:48.800 really wanted to write about this theme and really give 0:16:48.840 --> 0:16:52.720 it attention, is that Spy has almost nothing like you 0:16:52.760 --> 0:16:55.080 own none of this trade and Spy and as big 0:16:55.160 --> 0:16:56.560 of a fan as I am of the Voo and 0:16:56.560 --> 0:16:59.520 should trade, the one thing that I would argue that 0:16:59.560 --> 0:17:02.000 you don't get with a voo and Chill is in 0:17:02.040 --> 0:17:05.280 these trends. Eventually, if this grows and grows and grows, 0:17:05.280 --> 0:17:07.240 Micron will be a bigger part of Spy, no doubt, 0:17:07.760 --> 0:17:10.880 and you'll get at part of it. But you if 0:17:10.880 --> 0:17:12.800 you want to front run that, yeah, you can get 0:17:12.840 --> 0:17:15.720 these these runs that are earlier than when they're fully 0:17:15.760 --> 0:17:19.840 grown adults. This is why theming ETFs exists, in my opinion. 0:17:20.359 --> 0:17:21.879 So I think that's part of what we did. The 0:17:21.880 --> 0:17:25.280 other thing we looked at DROEL was, you know, Jake 0:17:25.400 --> 0:17:28.720 just talked about these memory companies are like under the 0:17:28.720 --> 0:17:30.920 gun to build all this stuff. Well they need stuff too, 0:17:31.560 --> 0:17:36.159 So we looked at semiconductor equipment makers and that's a 0:17:36.160 --> 0:17:38.760 whole other thing. That's a whole nother trade. Now there 0:17:38.800 --> 0:17:41.800 is no semiconducting equipment ETF yet, so we try to 0:17:41.840 --> 0:17:43.480 look for one of those before we go over the 0:17:43.520 --> 0:17:45.920 ETF that we chose for this. Do you want to 0:17:45.960 --> 0:17:48.240 talk about that supply chain? 0:17:48.560 --> 0:17:48.760 Yeah? 0:17:48.840 --> 0:17:51.520 Yeah, So I mentioned ASML before, so they're one of 0:17:51.560 --> 0:17:54.800 the keys. Now, in the past, DRAM hadn't utilized as 0:17:54.880 --> 0:17:58.600 much EUV, so some of these older technology nodes that 0:17:58.640 --> 0:18:02.440 were used for these DRAM waves didn't require EUV. That's changing, 0:18:02.640 --> 0:18:05.560 so you need more EUV tools per fab. But then 0:18:05.560 --> 0:18:08.320 expanding beyond that, LAM Research has really been one of 0:18:08.320 --> 0:18:12.240 the leaders in different deposition and etch machines that specific 0:18:12.280 --> 0:18:14.600 tools that specifically get used for these tabs. And then 0:18:14.600 --> 0:18:16.600 you have applied materials and then you have a very 0:18:16.680 --> 0:18:20.320 long tail of equipment companies that are In addition to that, 0:18:20.400 --> 0:18:24.240 you have KLA. I couldn't name them all right here, 0:18:24.280 --> 0:18:26.800 I would bore everybody, But the point is you have 0:18:26.960 --> 0:18:29.480 all these players. And the thing about them though, to 0:18:29.560 --> 0:18:32.920 keep in mind, memory tends to be a smaller percentage 0:18:32.960 --> 0:18:36.879 of their overall revenue, so their biggest customers are almost 0:18:36.960 --> 0:18:40.360 always TSMC. Then on top of that, you'll have company 0:18:40.400 --> 0:18:43.840 like Intel, and you'll have Samsung. And again, if you're 0:18:43.840 --> 0:18:45.880 talking about Samsung, a lot of that is for their 0:18:45.920 --> 0:18:49.720 foundry for logics and not memory, and Intel they don't 0:18:49.760 --> 0:18:52.560 make memory anymore, so if you're selling to Intel, you're 0:18:52.560 --> 0:18:56.000 not getting memory exposure via that revenue. So again, and 0:18:56.160 --> 0:18:59.879 there's a long tail also fabs that have smaller markets 0:19:00.680 --> 0:19:04.240 like global foundaries for example, or tower semiconductor some of 0:19:04.280 --> 0:19:07.760 the others that you're just not going to be selling into. 0:19:08.200 --> 0:19:13.080 So I think the exposure from these semiconductor capital equipment 0:19:13.119 --> 0:19:17.280 companies has been probably growing for memory, but also declining 0:19:17.359 --> 0:19:20.120 from China. So there's a couple of different moving pieces 0:19:20.160 --> 0:19:20.920 to keep in mind. 0:19:21.000 --> 0:19:22.840 The ETF we found for this one's roll was the 0:19:22.880 --> 0:19:28.879 Pacer Blue Star Engineering the Future ETF good name, but 0:19:29.560 --> 0:19:31.679 doesn't really have a ton of assets, only five million 0:19:32.000 --> 0:19:34.199 charges forty nine basis points. But all the firms that 0:19:34.520 --> 0:19:36.880 Jake just mentioned are in the top like ten holdings 0:19:36.880 --> 0:19:39.800 of this one. And I've been doing this a long 0:19:39.840 --> 0:19:42.679 time and sometimes when you find an analyst who's like, 0:19:43.000 --> 0:19:46.680 here's like these interesting stocks that are probably the most relevant, 0:19:47.119 --> 0:19:48.440 you have to go digging, and a lot of times 0:19:48.480 --> 0:19:50.920 you do find these these like new ones or ones 0:19:50.920 --> 0:19:53.800 that were overlooked in favor of the big popular one. 0:19:53.920 --> 0:19:57.080 So this is part of why we try to get 0:19:57.080 --> 0:19:59.720 people to explore the toolbox more fully and something we 0:19:59.760 --> 0:20:01.840 sh do more often his analysts, which is unearthed some 0:20:01.880 --> 0:20:03.840 of these really you know, cool, what's interesting there. 0:20:03.760 --> 0:20:07.240 Is like this one ETF becomes like this new sensation 0:20:07.400 --> 0:20:09.320 and then it makes you realize, don't look at the 0:20:09.960 --> 0:20:12.919 sun directly, look off to the side slightly, and that 0:20:13.040 --> 0:20:14.960 helps you see the rest of the eCos totally. 0:20:15.040 --> 0:20:16.600