AI Made Everything Expensive (ft. GamersNexus)

0:00This is an iHeart podcast. Guaranteed human. Run a business and not thinking about podcasting? Think again. More Americans listen to podcasts, then add supported streaming music from Spotify and Pandora. And as the number one podcaster, iHeart's twice as large as the next two combined. Learn how podcasting can help your business. Call 844-844-iHeart. This is Chelsea Handler from Dear Chelsea. Every week, the news gets worse, the world gets crazier, and Yamanika is here to tell whoever's responsible, you're the problem. Do you know I just found out who Sidney Sweeney was? If he got a bunch of women, then I should have a bunch of men.

0:35Do better or do less so I don't have to do so much. I'm Yamanika and I'm out. Listen to You're the Problem with Yamanika on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

0:52Call Zone Media. Hello and welcome to Better Offline. I'm your host, Ed Zitron.

1:11Subscribe to the newsletter, join the subreddit, buy a t-shirt, all the links are in the notes. And here's a joke to kick us off, what do you call the CEO of a French RAM company? That's right, Emmanuel Micron. And that's what we're here to talk to Steve Burke of Gamers Nexus about today. Steve, welcome back to the Octagon. Emmanuel Micron. Yeah. You didn't tell me that one before we started. Yeah, because you would have killed me before I read it. But look, we're here to talk about RAM. So we'll start with a simple question. Why is RAM so expensive right now? What is doing this? I mean, the reason it's so expensive is not grounded in reality, I think.

1:50yeah it's the the surface of it is that all the data center build out is consuming all the memory and therefore you know price higher because everyone wants it but i do think there's a little bit more to it than that right so from what so when you build chips we're going to really simplify this one for the audience because it's pain in the ass to explain otherwise but when you build chips you have something called a wafer you you have the silicon out and then you etch it using an orthography machine, but you can make less high... So there's the high bandwidth memory that they use inside GPUs, and then there's the regular DRAM that they use in everything else, right?

2:31Is that... It's a simplification, but is that fairly accurate? I think the most base level stuff to know is that memory, some form of RAM or memory, is in almost everything. A common misconception, I'm not sure... I think your audience is fairly technical, but a common layperson misconception is that storage is the same as memory they'll use it interchangeably storage is like your drive that you you know you put data on like photos or whatever and then memory is um is a temporary form of storage a volatile storage that gets wiped when it's not actively being used when you restart and it's and when they put stuff in memory that's where that's like ram and such where they like that's where you put the stuff that you're going to need quickly right yeah so yeah Yeah, memory is much faster than storage, permanent storage.

3:19And so it acts as this sort of transactional buffer to help get things to you quickly. So memories in phones, refrigerators these days, cars, there's all kinds of different grades of memory. And it's used in pretty much everything to do some kind of quick, like in a car, maybe you need to retrieve something related to like if there's a built-in map on the screen or something like that. or in a refrigerator, who the fuck knows, honestly. But I think that's... So that you can have a giant, horrible screen on your Samsung fridge. That displays you ads occasionally. Yes, exactly. Yeah, and so, yeah, to go to your wafer count, yeah, there's different types of sort of production for memory.

4:03And you'll find DRAM is used in all of it in some capacity for the most part. So you'll find DRAM dyes in HBM, like inside of an HBM package. But data center high-end GPUs will use high-bandwidth memory, and basically every consumer device uses something called GDDR, which is just a lower bandwidth, kind of more normal solution that's not as overboard on the performance for a consumer application. But they all kind of come from the same place, which is a wafer. They're all cut from place. so right now they're taking up more they are using more of the how much of their capacity are they actually using to build high bandwidth memory so the stuff that's going in the GPUs it appears to be an awful lot I mean so I know that the recent class action lawsuit or complaint I should say that was filed notes that the profitability the plaintiff's attorneys claim that the profitability of traditional, like conventional DRAM is greater than the profitability of HBM.

5:15That was one of their claims. And yet they sort of argue that, and it seems like their evidence is good, but they argue that why would you shift more and more capacity to producing this lower margin thing, HBM, if the easier thing to make is potentially higher margin. And so it does seem like more of it's shifting over there. you know open ai caused that sort of big kerfuffle previously when they claimed 40 percent of just all dram supply globally but that didn't happen though yeah it didn't happen it was all like basically one of the most insane just an insane story by the way just last year for those of you don't remember open a story came out that open ai was buying 35 to 40 percent of all the world's dram wafers it never happened yeah but we just talked about it like like we did like it did but So it did affect the price.

6:07It did. But here's the question. So why is everything so expensive? Is it because are they just jacking the price up? I think so. I mean, so, yeah, there's a couple things here. I'll lay out some of the highlights, and you can figure out how you want to go through them. But one of them is the memory industry has convictions, actually, in its past of price fixing collusion and, as the DOJ dubbed it, cartel-like behavior. So this goes the early 2000s. There was a bust. 2016, 2018, there was an attempt at a similar lawsuit. In the early 90s, the memory manufacturers were talking about how they thought their competitors were price fixing and colluding.

6:53And so this is a long and storied history where they, and now there's an active lawsuit, punitive class action for price fixing. So there's that antitrust aspect, which that obviously, you know, prices would be higher if you're just doing illegal things to make the prices higher. there's also the sort of physical manufacturing side which is that to make memory it's not quick to get another fab online and the companies seem to sort of signal to each other through earnings reports whether they're going to expand capacity or maintain capacity of their fabs which are factories for memory so there's that aspect which is if there's a real or not a sudden increase in demand And there's going to be a long latency to react to that with more supply if they even want to.

7:44So the current projections are 2028 or 2030 if we're getting more fabs online. The other reason it goes up, though, is really simple, which is that in a server that goes in a data center, some of NVIDIA's newer solutions will use one and a half terabytes of system memory. So that's DRAM. They'll use... Go ahead. Don't they also use the same RAM as we find in smartphones as well? Yeah, LPDDR typically, yeah. So Grace will use, I think it's 5X, I think is what they're on. Yeah, I pulled the numbers and it's something like 20 terabytes for an NVL72, which is a 72 GPU rack. Yeah, so they do, the single blade for the big boy is one and a half terabytes of system memory alone.

8:36And for perspective, a consumer desktop computer might have 16 gigabytes. With Windows these days, you'd probably want like 32. And they're taking one and a half terabytes for one server solution. But that's just the system memory. They also have other processors on there, like the Bluepoint or Bluefield, I think it's called DPU. It takes 128 gigabytes, if I remember correctly. the gpus take hundreds of gigabytes to like close to half a terabyte per gpu of hbm and the hbm uh has a there's kind of a yield factor at play where you know you can when you're talking about manufacturing that means you're also talking yields how many good products do you get for every bad and uh hbm is more difficult to make if you have a failure in manufacturing then you're also potentially losing more silicon dyes that could have gone to more and you and hbm is made by stacking the memory right so it requires just physically more memory forgive me if that's kind of an apes description no no it's fine yeah so hbm high bandwidth memory is literally just really high bandwidth uh it's kind of like if you were to if you were a city and you wanted to solve a public transit problem and you just built a shitload of i don't know let's take a massive tunnel you build like a bunch of trains you build a ton of trains for passengers but you don't have enough tracks to put those trains on they can't actually get on the track without blocking each other right you haven't really solved your problem and so high bandwidth is trying to solve that by saying okay we've got all this capacity for passengers which is the memory itself now we need something like a bigger road system to actually feed that capacity through to access it and so hpm tries to solve that especially because you look at gpus that might have like 12 plus times the amount of capacity that a consumer gpu would have and so you don't want to starve the the gpu core of the data that sits in the memory by not being able to get it fast enough and so hpm tries to solve that problem.

10:50So yeah, to go back to your point of stacking, high bandwidth memory is physically vertically stacked, as in they take silicon dyes for memory and they stack them atop each other and then they connect them all together as opposed to consumer memory where it's closer to, there's still layers, but it's closer to a flat plane. Right, but it just means that by virtue of that, it takes up more space in the place where they built the chips. Yeah, yeah. If you're on a consumer device, then the consumer memory, you'll need more physical area to put the same amount of capacity, whereas with the high-end server stuff, you just stack that capacity vertically.

11:34And then it's also, HBM's really significant advantage is that it's very low power comparatively. Right. Yeah, and so... But just to be clear, it takes up more space on the fabricator itself. oh yeah yeah if you're talking about the fab space then yes hbm requires a lot more space for a few reasons uh in terms of manufacturing area in terms of like footprint on a product it's less but in a fab it's more and um why is that in the fab uh one of the reasons so there's a few but one is that hbm requires additional silicon pieces to be able to function so with With GDDR, consumer GPU memory, you pretty much have the memory module.

12:22It looks like a black rectangle. And then if you were to use acid or something to etch away the black plastic part, you'd be left with pretty much a singular silicon die underneath. That's all it is for consumer. For HBM, you also need a logic die. So there's another piece of silicon that interfaces with the memory and the stack of memory. because you need something to kind of command and control and understand the vertical tower you've got of memory. And then you also need an interposer die, which sits underneath everything. So this is a solution that will sit under your GPU core, which is a very large piece of silicon and is shared with the HBM stacks.

13:08So all of your core and all your HBM is stacked on top of the same. It's called an interposer. It looks like a PCB, and they all communicate through that. So that's another piece of silicon. And so for these reasons alone, there's also others, but these alone, now you're involving more fabs. There's more logistics, shipping things around, and there's more fab or fabrication plant area allocated, even if you're just talking about making, someone's got to make the logic dies. TSMC makes the interposers. Right. There's just more moving parts, and it kind of takes up more space. exactly yep so why so to to my early question so this stuff is obviously lower margin but they seem to be able to sell an absolute shit ton of it so is that potentially why they're doing this that just they know there is insane demand on a level that for a high well i guess they can charge what they want for it as well i'm just having trouble trying to work out what it is they're doing other than just filling their boots have you tried asking uh chat gpt uh no okay i could ask my cat and get an answer i'd respect more but it's genuine but it's genuinely because it seems that unless i'm misunderstanding here that these micron sk hynex and uh samsung they're just kind of cranking up the price of regular ram yeah because they can i mean to some extent yes i I think this is where the price fixing allegations come into.

14:38I don't know if you try to take their side, which I really don't want to do. I'm trying. I'm trying. Yeah, if you try to. I think their argument, whether it's for truthful or not, but would be, we've got this crazy high demand for HBM. These are guaranteed customers. I can sell terabytes of it to Jensen, and it's one point of contact. it all moves quickly and easily and this there's this data center demand right now and we don't know if it's going to be here forever but consumers you know they're they're always going to have to buy electronics or whatever so i'm not exactly sure if that's to me it seems like maybe you make some sort of you you assign some value to the b2b relationship over b2c or something maybe but um also i guess if they yeah i'm not sure if you make why would that mean that they charge more for regular ram oh as far as why more for regular that part's easy that's just because uh the hbm itself takes the same a lot of the same machines as consumer memory and so if you're making more hbm you're naturally making less consumer uh which i think is where you start seeing the supply squeeze and the price go up.

15:59Likewise, a single piece of HBM can use up to something. I think it's like, it might be like 32 dies max or something of DRAM. You'll commonly see like four referenced, but a single piece of HBM will take more DRAM dies to build. And so one of the problems is if, let's just say four, if you have a piece of HBM that's got four DRAM dies in it and you have one failure somewhere in that stack or in packaging and assembly, now you've destroyed four individual pieces rather than if you have a yield failure with a consumer device, you destroy one piece. Sorry to ask a stupid question. What is a die exactly?

16:43Is that just what we would think of as a stick of RAM? No, a die is... So a wafer... We'll start at the wafer level. A wafer is just a disc, right? It looks like a CD, typically 8 or 12 inches. And they process it and they pattern it, or they literally project a pattern, or they lithographically etch a pattern onto the surface of this disc. Once that's done, these patterns, they're called patterns because it's a repeating sort of set of shapes that ultimately form the logic that makes the silicon. And the word dye comes from the process dicing. So with a wafer, when they're done producing it and patterning it, and it's basically ready to become a product, they need to cut that disc into a bunch of small squares.

17:42Right, so it's like a big slab that they cut up. Yeah, I mean, it's like making a pizza with a pizza cutter. except they're doing a pattern that only psychopaths would do on a pizza, which is making a bunch of squares. And so they cut it into the squares. That's called dicing. And then each of those pieces, the squares that yield from that, those are the dyes. And so if you were to pull up an image of any GPU, typically you'll see one large, it's called a monolithic die, which means one singular sort of piece of shiny reflective silicon mirror like finish that's your die and then for memory same exact concept just a lot smaller right so so right okay so there's less space to make regular rams so they just crank up the price that's really that really is it that's their entire thing right but i saw micron's earnings and they had 84.9 margin so it just kind of feels like they're setting the price to whatever they want it to be you don't think they're you don't think that is i is that not enough i think they need more don't they should get 95 well it's just kind of 95 why are you setting the setting it so low they should get 110 margins yeah no it's just it's very frustrating as well because i've read now arguments in favor and against micron where it's like oh micron got screwed by apple and consumer electronics companies during the last supply chain crisis because they kind of made a bunch more RAM than they needed to, from what I understand.

19:21Like Crimea River, you know? No, it really is. And it seems to be there's this boom and bust cycle in memory where they get a bunch of demand and then what? So do they build more capacity? Is it that they just order too much RAM and then people cancel their orders? like what happened what happened last time it is cyclical um yeah so it's cyclical i will say right now is nothing like i've ever seen and we spoke to a bunch of companies purchasers of memory like gskill and corsair who sell the sticks of ram that you put in a computer and those guys are telling us on record that um likewise they've never this is not like part of the cycle this is special and uh as for how is it different how the difference is how much the difference is the percentage increase it is this time so you know you're looking at like 700 increase in price in some situations in a very short period of time normally cyclically you'll see some ebbs and flows but it might be like back in the day a 50 kit of ddr4 became 85 or 90 maybe even 120 if it's a really bad year you know but you're still at like a little worse than double whereas now a 200 kit of memory is suddenly a thousand to twelve hundred dollars uh and so it's just the the amount of the price increase is what's different and then also um this time the sort of contest for the supply is uh it's it's coming from within the own industry where the very companies that would be the cause for an increase in sales of consumer supply are now the cause of these insane, unprecedented volume of sales for AI.

21:18Components.

21:23Get that Amex gold card ready. I'm way too tired to cook tonight. You read my mind. With the gold card, we can get up to$120 a year in statement credits. Are you feeling the Cheesecake Factory, Five Guys? Either of those sound good? Yes. Which one? Both. Pay with the gold card to receive up to$10 a month in statement credits at participating partners. Uncover more ways to reward yourself at AmericanExpress.com slash Explore-Gold. Enrollment required. Terms apply. Run a business and not thinking about podcasting? Think again. More Americans listen to podcasts than ad-supported streaming music from Spotify and Pandora.

22:01And as the number one podcaster, iHeart's twice as large as the next two combined. So whatever your customers listen to, they'll hear your message. Plus, only iHeart can extend your message to audiences across broadcast radio. Think podcasting can help your business? Think iHeart. Streaming, radio, and podcasting. Call 844-844-IHEART to get started. That's 844-844-IHEART. This is Chelsea Handler from Dear Chelsea. Every week, the news gets worse, the world gets crazier, and Yamanika is here to tell whoever's responsible you're the problem. If you come over here to play games, I'ma check you. Okay?

22:37If you do some in the news that don't sound good, I'm gonna play you. Join Yamanika Saunders as she breaks down the week's most problematic stories on her new podcast. You're the problem with Yamanika. Do you know I just found out who Sidney Sweeney was? New episodes weekly every Wednesday as part of my new network, the Dear Chelsea Network. If he got a bunch of women, then I should have a bunch of men. Do better or do less so I don't have to do so much. So join Yamanika each episode as she answers one question. Who's the problem? I'm Yamanika and I'm out. Listen to You're the Problem with Yamanika on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

23:30but it isn't based like so the 2021 supply chain crisis chip crisis was because there was a flood of consumer spending after covid yeah and that led to a bunch of consumer demand for products and overall spending across the board but like consumer demand drove it this one seems different because it's not demand outside of whatever psychosis the various hyperscale says. They're not spending a trillion dollars in CapEx really for like, it's not like consumers can't access whatever microsoft copilot. Well, I think what's interesting is the big three memory makers decided not, I think it was 2022 or three, to bring on more capacity in some of these situations because they were talking about how they got burned by oversupply and sort of following COVID, right?

24:21But what's interesting is if you go and look at some of the small suppliers, not many of them, but there's a couple of small memory makers that make up like less than 5 % of the market. Win Bond, Nanya are two of them. Nanya's headquartered at least in Taiwan. I don't know where their fab is. But these two companies, as far as I'm aware, I was just looking at this this morning, they increased their production capabilities at the time when the big three were saying that they weren't going to do it because they got burned. It's interesting to me because if you're, this is the argument the lawsuit makes too, if you're a small company, I would think you're at the biggest disadvantage, the biggest risk to increase your production capabilities with this massive amount of spend.

25:12If your biggest competitors are saying like this it looks like bad times ahead you know we don't want to spend into this but if you're a small guy like you're at the greatest risk of going under if you fuck up your your cat back right why didn't that this maybe this is a dumb question why didn't them increasing supply allow them to kind of undercut the microns of the world yeah that that one is just because the there's a micron does have a technology advantage so they can make memory that these other guys can't hbm is one of them maybe this is i didn't think about this till now but um maybe there is a very a small part i think but um possibly part of an answer to your question of why hbm that is something that's that other companies are struggling to make successfully in current generations so it's yeah i know only like sk hynix samsung and micron are the ones who can make HBM.

26:08Yeah, and the CXMT has also struggled with HBM, and they're a new Chinese memory maker that's actually done really well with DDR5. And so there may be a little bit of a pulling the ladder up after them factor. Well, here's a question as well when it comes to high bandwidth memory. Surely this is going to lead to an oversupply situation regardless, because HBM is used in what video cards god that's such an anachronistic term in the ai bubble um but like it's used in video cards and ai gpus basically right yeah it's very limited there's almost there's effectively zero consumer use case so when the ai bubble bursts where does the high bandwidth memory go uh oh yeah i don't oh okay yeah because i don't talk about the bubble burst in it well that's the thing i saw samsung is talking about increasing their high bandwidth they're spending like 50 something billion 500 billion dollars across an indeterminate period but like also building an entire new high bandwidth memory like factory yeah and that but i was looking i think i sent you the graph for coming on there was i saw this chart where it's like Like they were making what?

27:26A couple hundred million gigabytes of HBM in 2022. And they're now selling like seven or eight trillion a year. Yeah. So what's crazy is the HBM bit demand. I think I first saw this on semi-analysis, if I remember correctly, when I was at the sort of site but the hbm bit demand from nvidia was uh recently projected to exceed all hbm bit demand of all other companies combined from 2025 good and that's just nvidia and uh and again you have more wafer area per bit than with dram consumed that counts yield losses right and so increasingly as they make more and more of this, presumably for Nvidia's devices, you will continue to see consumers really get shafted if there's not new supply online soon.

28:29And likewise, if they're focusing all their efforts and resources on HBM for whatever reason, who's to say that the new fabs will even make consumer memory anyway, right? They might just make more HBM. But the thing is, if they're all built to make... Maybe I'm missing something. They're building these fabs specifically for high bandwidth memory, but high bandwidth memory is only used really for the thing in the AI bubble. Right. What are they going to do with all this capacity when it's gone? They would switch it over. They'd switch the line. I mean, they would, but it's like... I guess they'll just have...

29:05Aren't they just creating a future nightmare? uh i mean it's always good to have more capacity for them i guess i yeah i mean i guess it's good to have more capacity there's a lot of government money floating around too right like micron got a five billion dollar tax break in new york um as part of the uh i guess big beautiful bill but five billion like micron said they're gonna spend 250 billion right samsung and sk hynix have made 2.1 trillion dollars worth of commitments this just feels like it's it's almost like they saw what happened in 2021 and 2023 which ended with all of them doing at least a billion dollars of write-offs and when they were like fuck that we can go bigger yeah well i mean i i do think personally i i think that they are colluding again i i think they are price fixing um it might be through like a parallelism rather than a literal behind closed doors, meaning like we're just going to announce our changes in earnings reports and then our friendly competitors will kind of mirror us.

30:19And so, yeah, like it's kind of a prisoner's dilemma. Like if they don't raise prices, the other guy will, if they, and if they don't build capacity, the other guy will, they won't get the long-term deals. Right. Yeah. And also, the wafers themselves, I have some interesting numbers I wrote down, but I was looking up, when you were asking about HBM, I went and pulled up some of the original coverage I did of HBM, which started to come online, or at least was being actively developed somewhere around, I don't know, 2013 plus or minus a bit for AMD GPUs. So yeah, when AMD was working on Fiji and its Fury GPUs, they debuted HBM actually for consumer devices.

31:08They had HBM on Vega and on a couple other architectures in like the 2015 to maybe 2018 area, plus or minus a year or two. And massively expensive, just like it is today. the power efficiency was great which is really why they used it although they were bandwidth starved too but i i pulled up i had done some reporting at the time where i spoke to a bunch of the people in the know in the industry and at the time 2017 or so 2016 area the hbm2 memory for one of the devices um should have cost around 150 cost for them to to buy and the interposer in the packaging should have been another $25,$175.

31:55And this is old technology now. I'm sure a lot different. But the point of bringing it up is that if you compare sort of time for time and at the same capacity, at the time it would cost AMD $175 or so to put HBM on its device. Whereas with GDDR memory, the non-HBM stuff, they would have been at somewhere around like 68 dollars uh so they would be just so i guess the thing is they absolutely would make more money selling d ram or sorry ddr like regular people ram it's just that right now they have the business to kind of it's not just shitty it's not these bullshit video cards for loser gamers it's selling um ai gpus to big strong boys yeah right and yes like sam altman and yes The big, strong, hunky men like Dario Amadei.

32:53Right, right. Yes. Yeah. I've never, by the way, I don't think I've ever seen someone spread so much fear so anxiously as him. It's just pathetic. Yeah. Also, here's a random thing as well. A question to the listeners and to you as well, Steve. you ever see anyone on you've seen this weird habit of silicon valley people of using ai now to add muscles to guys yes i have seen this at all what's that about one of the weirdest things i've seen in the tech industry's history and there's a lot of weird shit i don't know if it's just well they're they're not used to working for intellectual property at this point they just harvest it so why work for the muscles either well it's just it's really strange like i like I think the TBPN, the Blowjob Brothers Network, they do it with Dylan Patel, who's just a regular-sized guy.

33:50And it's just... Sorry, I've just realized that's weird. But then I saw someone do it for Dario Amadei the other day. And this is like, your body's your body. I don't judge people for how they look. But at the same time, Dario Amadei is not like a fifth course of steroids guy. Like, it's just the weirdest thing I've ever seen. Did you see the story just recently of the copper thieves stealing shit from data centers, too? No. That was good. Average better offline listener there. Yep. Yeah, they're probably wearing one of your t-shirts. Yeah. They're going to get a summon. Oh, God, no. I disavow them.

34:31yeah the uh i saw there was a title of um of thieves stealing copper from data centers and it's like oh okay i mean it's really i guess there's no honor among thieves you know it's just thieves stealing from thieves it's also inevitable at this point because you make these giant things full of earth metals and then you put them in neighborhoods usually with like like disadvantaged farm neighborhoods or like like distressed towns in texas you're like oh i wonder what will happen with the other business in the area yeah i wonder what will happen when we screw over all these people and and like mock them via the media constantly anyway back to back to ram so is this something that's going to start affecting like storage so hard yeah yeah yeah so you mean like actual uh yeah like like solid state drives and the like yes so it's already affecting it um storage is made at the same fabs for the most part as memory with basically the same tools and mostly by the same companies although there's there's like one or two differences but for the most part it's the same people and tools and so definitely one affects the other anyway um but storage in general has it hasn't gone up as violently as memory but it has gone up uh some of the ssds we had bought previously i was just checking this actually for like 90 bucks maybe about a year ago are now closer to like 180 190 and jesus christ yeah uh let me pull up a chart here that's this is genuinely like i think i don't know if i'd say i do anything that's activism but like i think people just need to know this is ai's fault yes and the worst part is it's not even ai's fault because they need these data centers because they're taking forever to build it's like you can access clawed and chat gpt you don't there's no one no one's like unable to log on because there's not enough data centers nothing's being stopped they're just like no we need to do we need to buy buy the chips need to build the data sense for some reason yeah it's i mean yeah there's not like a queue i mean i guess they're maybe they make the argument of like we're trying to make it better and the models will be more complicated and that requires when's that gonna happen but yeah the uh storage though so looking at an average price chart for four terabytes of solid state drive storage and vme on pc part picker which averages it um they have the pricing at around it looks like maybe about 270 dollars in january last year and now it's somewhere around somewhere around 800 and 900 and that's for what kind of storage that is for four terabytes uh so relatively large capacity which is going to be affected the most jesus christ yeah and this is where this is interesting too because one of the ways you'll see consumers getting shafted that's not as obvious is um some of the modern devices that should be your next sort of you know improvement upgrade on past generations they're actually coming down in capacity so i haven't looked at the exact device names for a while we ran a report a while ago talking about some of the phones and laptops that have shifted from say a prior model of 12 or 16 gigabytes down towards eight for memory there's decreasing storage capacity as well yeah that's fucking terrible so everyone just suffers here and it's pretty like there's microns sk hynix samsung they could just they could have not done this just to be clear just so they could have sat with like 70 % margins.

38:29They're already colluding anyway, in my opinion. So yeah, they could collude in a different way. It's just, it's so fucking straight. Well, I mean, it's actually, no, that's not, that's not fair. This isn't strange at all. This is exactly like, this is exactly what companies like this do. Yeah. Like this, it's this kind of vulgar kind of chancerism. And their argument from what I understand is, well, memory is cyclical. So we need to fill our boots when we can. and if you're calling something vulgar then it's pretty vulgar yes just to be clear as the vulgarist man in tech and that is a word now uh but the thing is though just to be clear that's their argument though though because things are good now they can charge what they want so they can fill their coffers for the future yeah i think so and i mean you see it affecting i think that's that's probably their argument i think on the consumer side too even a company like lenovo which is not a manufacturer of memory but they buy it and they put it in stuff uh i'm not sure if there may be this i don't know for fact but there may be a benefit for some of these companies as well for a few reasons one is uh there's a lot fewer it kind of reminds me a little bit of COVID when all the stores that were 24-7, all the grocery stores and whatever, suddenly stopped being 24-7.