[00:00:00] Andrew: Okay, we're live on YouTube and Twitch. Hello, friends
[00:00:04] Nathan: Yes, we're just a [00:00:05] little late today.
[00:00:06] Andrew: We're actually we're on time, if anything, right? We- we're [00:00:10] usually super,
[00:00:11] Nathan: know
[00:00:12] Nathan: super lateYes, a- and yes [00:00:15] relatively we're on time.
[00:00:17] Andrew: Yeah.
[00:00:18] Nathan: going? How are you doing?
[00:00:19] Andrew: I'm good. [00:00:20] How are you?
[00:00:21] Nathan: I, I'm doing well. Yeah, we were talking before. It seems [00:00:25] like this week has blown by and yeah, we're both just ranting on random things.[00:00:30]
[00:00:30] Andrew: Yeah, and I'm highly motivated. So the United States plays, [00:00:35] US soccer, world football
[00:00:38] Nathan: Yeah, football. Yeah,
[00:00:39] Andrew: [00:00:40] tonight at 7:00 PM Pacific, so 10 o'clock [00:00:45] Eastern. And the thing that's cool about that is one, we're already going forward, so we [00:00:50] win or lose, doesn't matter. But two, we have the opportunity to be the first [00:00:55] team in world, in our World Cup history, so for America, that could win [00:01:00] all three games in the group.
[00:01:02] Andrew: So for folks who aren't familiar, you play [00:01:05] we're playing three different teams in a group. There's four teams in a group. [00:01:10] Whoever wins in that group, you're the group winner, you move on. If you don't win or come in [00:01:15] second, there's some opportunities to move on. It's a whole thing. But it, it's pretty rare to [00:01:20] win all three of your games.
[00:01:21] Andrew: U- usually there's a draw or a loss. [00:01:25] So it'd be really cool. It's a cool milestone. And it should be a fun game I think, 'cause it, it's on one [00:01:30] hand, like it doesn't matter 'cause we're moving forward, so you wanna balance that strategy with [00:01:35] we don't wanna tire everyone out or get people hurt.
[00:01:37] Andrew: On the flip side, anyone who's gonna be on the pitch is [00:01:40] gonna be like, "I want a record," right? "I wanna clinch the like three [00:01:45] wins i- in the US team history franchise." so it should [00:01:50] be a fun game. It's against Turkey, who, sorry Turkey, you're not very good. Better luck [00:01:55] next time. But you never know how these things go, right?
[00:01:57] Andrew: There's a lot of games that have been on [00:02:00] that, not really sure how it happened, but it [00:02:05] did
[00:02:05] Nathan: Yeah, yeah. I, I, I really don't have any-- I'm not a sports person. Although
[00:02:08] Andrew: Yeah.
[00:02:09] Nathan: are interesting, [00:02:10] so the, the
[00:02:10] Andrew: Yeah
[00:02:10] Nathan: made about people, like the game result doesn't really matter, but [00:02:15] there's something bigger here at, at stake, which is potentially being record holders,
[00:02:19] Andrew: in [00:02:20] the North America's hosting it, right? So we have, I think, four sites of the [00:02:25] seven or six or something, and there's a couple in Mexico, maybe a couple in Canada, and then there's three or [00:02:30] four in the US. So that's cool. We're the host country, I think.
[00:02:34] Nathan: [00:02:35] Right
[00:02:35] Andrew: and so it's a big deal, right?
[00:02:36] Andrew: You wanna win at home. D- not to get anyone's hopes up or to crush [00:02:40] the team, but, the US very rarely makes it anywhere near the [00:02:45] final just historically. That could be very different this year. I hope we do good. But [00:02:50] just for folks who might not know anything about soccer or what the rest [00:02:55] of the world calls football we're not generally very good on the world stage.
[00:02:58] Andrew: So we'll [00:03:00] see.
[00:03:00] Nathan: Yes. Yeah
[00:03:01] Andrew: So yeah, I'm mo- I'm motivated for that. I'm excited. I, I [00:03:05] love, I love football. I'm really... Okay, I got a gripe for anybody who's a [00:03:10] footballer. These fucking hydration breaks are killing me
[00:03:14] Nathan: What can you
[00:03:14] Nathan: [00:03:15] allow me? What's
[00:03:15] Andrew: so historically and generally in most [00:03:20] matches, you play two 45-minute halves, plus or minus a few minutes for [00:03:25] overage or stoppage.
[00:03:26] Andrew: If you're in some kind of a competitive championship, usually there's like a [00:03:30] shootout at the end where you just try to take goals to determine the winner if you're, you're [00:03:35] tied or you just draw. And, I believe the, [00:03:40] the FIFA World Cup precedent got set when they were in Qatar and the, [00:03:45] the UAE, I think was last year's...
[00:03:48] Andrew: I forget if it was in... [00:03:50] somewhere in the Middle East. I thought it was in Qatar, but it could've been other places too. But it was so hot and [00:03:55] so dry, right? 'Cause it's, it's hot and dry there. They instituted hydration breaks, and [00:04:00] they do that in the States too. In our soc- our soccer football league's a little [00:04:05] bit different than Europe, but like we'll do those too 'cause we have like places like Arizona and like even LA [00:04:10] sometimes just gets wretched, and the idea is like player safety.
[00:04:14] Andrew: The, [00:04:15] the problem I have with them is that they're like now pretty much a given, like roughly [00:04:20] half of h- roughly the half time in the [00:04:25] half, so about, what? 25, 30 minutes into the half, and then [00:04:30] again in about 60 minutes in the second half. And so this, the part of [00:04:35] the strategy is endurance, right? And so it really breaks the endurance game, and then if there's a [00:04:40] lot of momentum with one team either doing really well or starting to [00:04:45] recover, it breaks up that momentum.
[00:04:47] Andrew: Yeah. And so what I'd love to see is that [00:04:50] they be optional and they be official's, discretion. And so it's like you may or may not get one, so you can't [00:04:55] really plan for it. But if it is really hot and shitty or people are like passing out or [00:05:00] whatever, the official has an opportunity to like give a break
[00:05:04] Nathan: when [00:05:05] people are passing out.
[00:05:06] Andrew: Yeah.
[00:05:06] Andrew: And, and and then the players, so you know, like [00:05:10] hydration, so water, it's not water. They're drinking Gatorade and like salt, potassium drinks, [00:05:15] all that kind of stuff, which whatever, that makes sense 'cause you don't wanna cramp up. But like they have ice [00:05:20] packs they're using a lot of the players...
[00:05:21] Andrew: And so it's it's gotten to be this more of And like for anyone [00:05:25] who's never ran like 12 miles in an hour and a half, that little [00:05:30] bit of cooling or that little bit of liquid can really, really change. You [00:05:35] wanna run a little bit harder now, or there's a couple big plays and you don't get as exhausted or as [00:05:40] tired, or your body has more time to recover, so you have more performance [00:05:45] over the, the whole period of, of the game instead of getting tired towards the end or trying to save yourself to the [00:05:50] end, which again, it's a strategy, right?
[00:05:52] Andrew: I don't know. I, I don't want people to get hurt and I don't want people to get [00:05:55] dehydrated problems, but it's, it's also really, really fundamentally changed the game and I, I [00:06:00] just, I gotta gripe publicly for all the Americans who have no idea what I'm talking [00:06:05] about.
[00:06:05] Nathan: Got
[00:06:05] Andrew: it, it sucks. I don't like it.[00:06:10]
[00:06:10] Nathan: Sorry, to be clear, you don't like it because it changes the nature of the game, not so much that they're trying to protect the players[00:06:15]
[00:06:15] Andrew: Yeah, and
[00:06:15] Nathan: to
[00:06:15] Andrew: it's bull- it, it's bullshit, right? So it's And because it's every [00:06:20] game, a game in Seattle, what is it? I think it's [00:06:25] Friday or Saturday, like we're supposed to have like unprecedented cold temperatures. So
[00:06:29] Nathan: [00:06:30] there's still gonna have a hydration
[00:06:30] Nathan: break.
[00:06:31] Andrew: but I think the reason they did that is so then every match is equal.
[00:06:34] Andrew: So [00:06:35] every match has two hydration breaks. But it's it totally changes the nature of [00:06:40] the game and, and I would like to see it, or if the, the temperature is above 32 or [00:06:45] something like that where it starts to get really warm. 32 degrees Celsius for folks [00:06:50] is approaching like 80
[00:06:52] Nathan: Hmm.
[00:06:52] Andrew: I think if my math's right
[00:06:53] Nathan: 80. Yeah, no, it's, [00:06:55] that is, that's right. Yeah.
[00:06:56] Andrew: Yeah.
[00:06:56] Andrew: So it, it's like
[00:06:57] Nathan: I, I don't know too much about [00:07:00] the sports industry, but i- isn't there something with this idea of maybe this is [00:07:05] something that, that we get into, but is it FIFA or the... [00:07:10] Some sort of organization has like full control over how these [00:07:15] games are
[00:07:16] Nathan: set up and played and...
[00:07:17] Nathan: Okay. Yeah, that's what I thought.
[00:07:18] Andrew: Yeah, and it's one of the most [00:07:20] corrupt organizations in the world, so like
[00:07:21] Nathan: that's what
[00:07:22] Andrew: that's...
[00:07:22] Nathan: didn't wanna
[00:07:23] Andrew: Yeah.
[00:07:23] Nathan: by going [00:07:25] yeah.
[00:07:25] Andrew: Yeah. I'm just-- I guess I probably got banned for life, but whatever, I'm not gonna pay seven grand [00:07:30] to go to a football game. Sorry. I, I, I'll gripe on [00:07:35] that too. I think it's insane the prices for these tickets and like the idea that you [00:07:40] can live in like a state or even a city where the match is and it's not [00:07:45] anywhere close to being approachably affordable for folks is, is crazy.
[00:07:49] Andrew: [00:07:50] Like it's
[00:07:50] Nathan: that, that happens with con- music concerts too, right?
[00:07:54] Andrew: Yeah.
[00:07:54] Nathan: people [00:07:55] from,
[00:07:55] Nathan: yeah
[00:07:55] Andrew: And, the, the FIFA this year I think is, is different than the last World [00:08:00] Cup, but they control the sale cycle, so you never take custody of the ticket. [00:08:05] And so then they get buyer and seller fees on both sides of it too, which is insane. I don't want [00:08:10] people to get ripped off, so I think having that system makes sense, but they should not be making [00:08:15] massive fees.
[00:08:16] Andrew: I was gonna... I was look- I, I found last-minute tickets to the game in [00:08:20] Seattle, which is the closest ma- mega city I'm, I'm, I'm by, and [00:08:25] they were $750 tickets, which were already bonkers for a sporting event. And [00:08:30] I was like, it's w- first time in 30 years we're hosting it. We might not [00:08:35] ever host it again.
[00:08:36] Andrew: Probably won't be another 30 years. It's in my [00:08:40] backyard. Fuck it, right? And then I went to check out, and it was somehow $7,000[00:08:45]
[00:08:49] Andrew: And I was [00:08:50] like, "I don't know how this got 10X more expensive in two clicks. I don't even care to try to [00:08:55] figure out why." I am like, "I am totally done with this." So if, if [00:09:00] you're not at the game or think the pricing is outrageous, you're not alone. And I would say [00:09:05] definitely find a watch party. It's way more fun to watch with a bunch of soccer fans, [00:09:10] in a public setting than at home, it takes a little bit of getting used to for a lot of American [00:09:15] folks. A very different game. It's, the action comes in spurts. It's not continual. [00:09:20] It's very dissimilar to most American sports. It's more probably similar to baseball in that regard, where it's [00:09:25] like a lot of not a lot happens, and then all of a sudden something happens and the game can [00:09:30] radically shift.
[00:09:30] Andrew: So I don't know. Now it's the sports stream. Throw a little box in the [00:09:35] corner and we can
[00:09:36] Nathan: sports in general, I, I think there are [00:09:40] questionable... I don't wanna say economics because I, I don't know if you can call them economics, but
[00:09:43] Nathan: the way that a lot of these sports, [00:09:45] major sports events are structured, because most of them are typically monopolies, right?
[00:09:49] Nathan: [00:09:50] They are
[00:09:50] Andrew: Yeah.
[00:09:50] Nathan: they're held and
[00:09:52] Andrew: have a monopoly on the stadium or the [00:09:55] venue, and it's owner-operated.
[00:09:57] Nathan: Yeah
[00:09:58] Andrew: there is no com- c- you can't [00:10:00] just spin up a sports team and I, I don't know if people even know this, like the [00:10:05] NFL, right? The quintessential American Football League, you cannot act- It's a [00:10:10] private company,
[00:10:11] Nathan: Yeah,
[00:10:11] Andrew: and I can't just make Andrew's awesome football [00:10:15] team and compete.
[00:10:16] Andrew: They have to vote to approve me. But we give them all these [00:10:20] public subsidies to build stadiums and to move players and all this kind of shit, so that, that's a whole [00:10:25] thing. It's interesting 'cause I think this actually segues really well into what, what [00:10:30] we were gonna actually talk about.
[00:10:31] Nathan: caught, you caught
[00:10:32] Andrew: I know. You just threw me this softball[00:10:35]
[00:10:35] Nathan: I was trying to alley-oop it to you.
[00:10:36] Andrew: Yeah.
[00:10:37] Andrew: But, and so that's a whole thing. But I do [00:10:40] think, there was an antitrust thing about sports tickets or Ticketmaster, I think specifically. [00:10:45] And Ticketmaster does the same thing. They either own the venues or own controlling interest in all the venues, [00:10:50] the big venues in America, and then you have to get a Ticketmaster ticket, and then they do surge pricing and [00:10:55] all this crazy shit to extract the m- And th- this money does not go to the artist.
[00:10:58] Andrew: This money does not go to the [00:11:00] performers. Most of it goes to the company. It's just like Uber and Lyft. Most of the [00:11:05] money you pay doesn't actually go to the driver. It goes to the company so that they can have a really cool [00:11:10] headquarters and cater lunch for their employees. It doesn't go to the actual person doing the work.[00:11:15]
[00:11:15] Andrew: Exactly.
[00:11:16] Nathan: Yeah
[00:11:16] Andrew: So I'm not, I am not against if that money is, is [00:11:20] given disproportionately to the people who are actually creating the value. [00:11:25] Awesome. And I don't think facilitating a digital ticket is creating a ton of value in [00:11:30] 2026. The thing that I don't like about it, and I'll call out SeatGeek on this too, 'cause this is who I have to deal with [00:11:35] most of the time, they forbid you from taking a ticket.
[00:11:39] Andrew: I can't even, [00:11:40] I guess I can technically send it to somebody, but I-- it's almost impossible to actually have a [00:11:45] private sale of the ticket, so I have to list it on their platform. They charge me a fee [00:11:50] to sell it, and they charge a buyer a fee to buy it. And [00:11:55] that is the most scammy, anti-competition, [00:12:00] monopolistic feature I have ever encountered in this country, and it's it's just getting worse like that.[00:12:05]
[00:12:05] Andrew: But that's the-- like, that's a huge problem. And it's, again, if it was going to the [00:12:10] players, maybe okay. But it's not.[00:12:15]
[00:12:15] Nathan: Yeah
[00:12:15] Andrew: I think there should be a cap on what people make, too. Not necessarily an income [00:12:20] cap, but the idea that you can just charge whatever you want is insane, that there's market economics, 'cause there [00:12:25] really isn't.
[00:12:26] Andrew: But we haven't been able to solve it and grapple with [00:12:30] this, and it's directly related to all this tech AI hysteria that we're [00:12:35] seeing. And how is it related, you ask? Great question. [00:12:40] The headline this morning was Apple's prices went up 20% on [00:12:45] MacBooks and iPads. Interesting it didn't hit phones, which is their number one cash
[00:12:49] Nathan: Yeah, not [00:12:50] yet. Yeah
[00:12:51] Andrew: Well, I think they're gonna raise them a little bit. I think they're gonna really cut [00:12:55] what you get up, and they're not probably gonna have a massive hardware bump to try to keep the prices [00:13:00] stable, or they're gonna really, really bump the pro models to outrageous prices and [00:13:05] they'll keep the other models more affordable.
[00:13:07] Andrew: But the, the purported reason for [00:13:10] this was like Apple was trying to shield us poor customers from these [00:13:15] high prices that AI is causing, and they just couldn't bear it anymore. And [00:13:20] their, 37 point something percent margin went to [00:13:25] 39% margin last quarter. And so while they were shielding us, they made 2% [00:13:30] more margin on their products.
[00:13:31] Andrew: Poor them. Don't quote me on the numbers, but it was [00:13:35] about that. So that's nonsense. It's PR. It's bullshit. Apple's a [00:13:40] multi-trillion dollar company, I think, in market cap at this point, or at least well [00:13:45] healthy trillions.
[00:13:46] Andrew: They're fine. They can eat a 20% hike for a [00:13:50] long time and be totally fine. But I, I think w- so 20%'s a lot, [00:13:55] and I think that was my number one shock was that a few percent here, a few percent [00:14:00] there.
[00:14:00] Andrew: Some companies have ratcheted stuff up gradually and there, there's pros and cons for both [00:14:05] pricing strategies. I've been in these conversations. I'm sure you have too. But I think it's better to [00:14:10] do smaller bumps more frequently than a massive bump like that, because it [00:14:15] does... It creates its own shock, right?
[00:14:17] Andrew: You're like, "Oh, shit, this thing that used to be 20% less." [00:14:20] And, and, and for folks like Apple hardware, you're talking a few hundred dollars [00:14:25] to a few thousand dollars of, of increase And a lot of [00:14:30] people have opinions on Apple products that they're already overpriced. They're actually pretty-- [00:14:35] They were pretty fairly valued for what you get.
[00:14:39] Andrew: [00:14:40] They're definitely on the more premium side, but in compared to the performance, the engineering [00:14:45] objectively they're actually pretty comparable to what you could get in the market. A comparable [00:14:50] chip would be, like, the, was it the AMD AI [00:14:55] 395 Max Plus or something?
[00:14:57] Andrew: System on a chip.
[00:14:59] Andrew: It's [00:15:00] similar design. It's not as good, frankly, but it's a, it's a similar, [00:15:05] system architecture, and it's roughly, similarly priced, I would say. I wouldn't say [00:15:10] roughly the same, but similarly priced. And that's where we're going, right? We're going with CPUs and memory [00:15:15] being embedded and even storage embedded on the same controller because the only way to [00:15:20] get high speed throughput and low power is to make things really close [00:15:25] and,
[00:15:25] Nathan: Yep
[00:15:26] Andrew: And not plug into anything because the, the socket [00:15:30] systems have a lot of electronic interference basically, and there's a whole other bunch of shit, [00:15:35] right?
[00:15:35] Andrew: And so the reason then that these, this stuff's getting more expensive is because [00:15:40] they're more complex to design. It's a bigger actual physical die. [00:15:45] So the, the thing that, you know... How do you explain dies for folks? The, that [00:15:50] used to just be the CPU, the RAM, and the memory, and the storage were all kind of separate components made [00:15:55] of chips.
[00:15:55] Andrew: You plug them into a big motherboard and, and that, that was that. You might be able [00:16:00] to upgrade them, you might not be able to. Putting all those components on the same [00:16:05] actual circuit board per se, is
[00:16:08] Nathan: Physical form factor or [00:16:10] something, yeah
[00:16:10] Andrew: Yeah, and it's much more expensive. But the-- when you build an [00:16:15] actual chip, there's only so many chips you can build at a time.
[00:16:18] Andrew: If you think of like a pizza, right? [00:16:20] There's only so many toppings you can put on that pizza. You can stack 'em, but even then they start falling [00:16:25] over and falling off, right? You can do that with chips too. And so you can make the [00:16:30] pizza bigger. The problem with doing that with chips is that you get more errors and you get [00:16:35] less good chips.
[00:16:35] Andrew: So there's a physical size problem with reliability. [00:16:40] And so you're kind-- and you're also just, there's also like other reasons, but you're stuck kind
[00:16:44] Nathan: There's, [00:16:45] there
[00:16:45] Andrew: constraints. Yeah. And so you now, instead of being able to get [00:16:50] 100 memory chips on your pizza, now you can get 30 CPU [00:16:55] and memory chips on your pizza.
[00:16:57] Andrew: And so there's always trade-offs. And so [00:17:00] Apple's kind of moved that direction. Other companies are moving that. There's pros and cons. And, and so why [00:17:05] is this expensive? Why is AI the cause? I'm not... AI has been [00:17:10] purported to be the cause.
[00:17:12] Nathan: You gonna corroborate that?
[00:17:13] Andrew: I'm gonna qualify it in a [00:17:15] second. But AI is being blamed because of the massive amounts of memory [00:17:20] that GPUs are needed to train inference systems [00:17:25] to serve up the model, to store memories, to run all the tools [00:17:30] and shit in the background.
[00:17:31] Andrew: And the same memory chips that are [00:17:35] in those GPUs and those servers are the same memory chips that Apple would use in their system, are [00:17:40] the same memory chips that Sony uses in the PlayStation, that Xbox uses, that [00:17:45] Nintendo,
[00:17:45] Nathan: all your hardware that has a
[00:17:47] Andrew: Yep.
[00:17:47] Nathan: has the same, yeah.
[00:17:48] Andrew: And there's different kinds [00:17:50] of chips, but the reality is there's really only two places or two factories [00:17:55] or two, I guess two companies in the world and a handful of factories that actually make the, the [00:18:00] core components to make a chip.
[00:18:03] Andrew: And that's the [00:18:05] problem, right? The problem is not necessarily AI. Like they've put a bunch of futures on [00:18:10] purchasing. They've done a lot of guaranteed purchases. They've done a lot of promissory purchases, have driven the cost [00:18:15] of these components sky high. But the problem is it's a monopoly, right? There's [00:18:20] r- there's, there's more than two memory companies, but there's two main ones that are, I think it's 97% or [00:18:25] 96% of the entire world's supply.
[00:18:28] Andrew: And they're booked out for [00:18:30] three years to companies like...
[00:18:32] Nathan: TSMC and Nemes,
[00:18:33] Nathan: Uh
[00:18:34] Andrew: When it-- [00:18:35] they they make the, the,
[00:18:37] Nathan: Chicken general
[00:18:38] Andrew: They make [00:18:40] the, the dies.
[00:18:41] Andrew: But the, the full package memory assembly is, is [00:18:45] Hynix, XK Hynix and, Micron.
[00:18:47] Nathan: oh, you're talking memory. Yeah, Micron.
[00:18:49] Andrew: [00:18:50] Yeah. So yeah, you have to make the chips and then you assemble them, and then if you have to put them in a die because [00:18:55] you're doing these CPU memory packages, that's a whole 'nother [00:19:00] process.
[00:19:00] Andrew: But yeah, so there, there's two companies that kind of make the... Samsung's the other big one, but I think they're like 2% like [00:19:05] 1%. It's just some minuscule amount. And they sell to themselves. They sell to other companies too, but they're [00:19:10] like, it's vertical. So the, the, the fundamental problem is that [00:19:15] there is a supply constraint, but it's not a supply constraint brought on by demand per se.
[00:19:19] Andrew: It's [00:19:20] a supply constraint because the monopolies are not [00:19:25] building more productions because they know that these things are really... [00:19:30] Fabs are really expensive to build. Assembly plants are very expensive to build. They take forever. And [00:19:35] the cyc- cyclical nature of technology is, is boom and bust, up and down.
[00:19:39] Andrew: And so they don't [00:19:40] necessarily want to invest in it. But the reality is there's no little mom-and-pop fab [00:19:45] that could be like pumping out memory chips for Apple or you get 20 [00:19:50] mom-and-pop fabs to, to, to bring in that supply or change their production. It's just really two big [00:19:55] companies. They're doing massive scale Apple's kind of cornered the market on [00:20:00] these for a long time.
[00:20:00] Andrew: They've guaranteed purchases in the future, that kind of stuff. But now NVIDIA [00:20:05] and other companies are, are, are just throwing money at them. And [00:20:10] so they're going with the highest bidder to some extent, and it's really just upset the entire [00:20:15] dynamics. And so if you don't use Apple, you're probably like, "I don't really care."
[00:20:18] Andrew: The problem is [00:20:20] memory chips are literally in every digital device.
[00:20:23] Andrew: Your alarm clock probably has a [00:20:25] memory chip
[00:20:26] Nathan: For flash memory, sure. Yeah.
[00:20:27] Andrew: Right?
[00:20:27] Nathan: I think that was like the, [00:20:30] like you said, it's not necessarily demand per se, but I, I think the more interesting, I don't know, [00:20:35] maybe business angle is it, it really was what you said they're, they're going with the [00:20:40] highest bidder. So if, if
[00:20:41] Andrew: Yeah
[00:20:41] Nathan: produce X many chips in your consumer market, [00:20:45] which is like your average person here and there buying a laptop or some hardware device versus [00:20:50] those hardware device manufacturers, like the big names like Apple, why am I brain farting on the-- Big [00:20:55] companies that, that are trying to build this hardware, right?
[00:20:57] Andrew: Dell, HP, NVIDIA
[00:20:59] Nathan: Yeah, I [00:21:00] was, I was hesitant on, on mentioning like Dell, 'cause I didn't know if they're still, they're still in the game, but I
[00:21:04] Andrew: Yeah, no, they're [00:21:05] still in the game. They're, they've doubled down on AI stuff. I, I can't really speak to [00:21:10] any of it. I know they're in it. I
[00:21:12] Nathan: mean, I think the s- the split is like consumer [00:21:15] hardware versus the big companies that are building the consumer hardware, but also have other [00:21:20] motives for buying that RAM, which
[00:21:21] Andrew: Yeah
[00:21:21] Nathan: going towards AI compute or AI inference or something [00:21:25] like that,
[00:21:25] Nathan: right?
[00:21:25] Andrew: yeah. W- I hear what you're saying. Normally I, I [00:21:30] wouldn't disagree, but I think for technology, there isn't [00:21:35] really a difference between professional and consumer for [00:21:40] most stuff. Yes, servers have different processors and memory, but a lot of the core components are [00:21:45] very similar. And even, like, when you look at GPUs, up until the [00:21:50] HBs and s- in the, the, what, three generations ago [00:21:55] Nvidia chips that were designed to be either, rack, [00:22:00] racked themselves or part of integrated chassis.[00:22:05]
[00:22:05] Andrew: Those chips that you were getting in a high-end server [00:22:10] GPU were a higher memory bandwidth or a higher memory version or a higher [00:22:15] core count of a consumer. And so they were really just selling the consumers the shit that [00:22:20] wasn't good enough to make the cut for the server market for the enterprise.
[00:22:24] Andrew: So it was kinda [00:22:25] discard in a sense. And like it was a lit- not totally discard, but it was kinda that way. [00:22:30] And same with Intel for the longest time, a lot of the, the Pentium, I [00:22:35] think it was the Pentium 4s
[00:22:36] Nathan: chips that
[00:22:37] Andrew: actually just Xeon [00:22:40] rejects.
[00:22:40] Andrew: Uh, but
[00:22:42] Nathan: that's what I meant, like the
[00:22:43] Andrew: yeah
[00:22:43] Nathan: hardware, which was like if [00:22:45] you, if you can only manufacture so many chips, you're going to allocate more of that resource to building the [00:22:50] chips that net you a higher
[00:22:51] Nathan: return
[00:22:51] Andrew: and I think the the, the additional rub on this though too is [00:22:55] not all memory is the same, right? And like the HB memory, the high bandwidth stuff, which is what [00:23:00] the GPUs really need, is I think effectively [00:23:05] four RAM chips, DRAM chips stacked together [00:23:10] essentially
[00:23:10] Nathan: I need to do
[00:23:10] Andrew: to get the...
[00:23:11] Nathan: this. Yeah,
[00:23:11] Andrew: Yeah.
[00:23:12] Nathan: process for like
[00:23:13] Andrew: Yeah.
[00:23:14] Nathan: versus
[00:23:14] Andrew: [00:23:15] quote me on it, but...
[00:23:16] Andrew: And it's, it's, yeah, it's a little different, but my understanding is like each [00:23:20] module that sits on an HB chip is like a stacked [00:23:25] chip design for two memory modules, and then they, logical them, logical [00:23:30] them, logic them together to create the high bandwidth for, so you have a, I think [00:23:35] it's a four-channel read/write versus a single double [00:23:40] with the DDR spec.
[00:23:40] Andrew: I, I don't get into hardware that much anymore as I don't have to deal with it. I [00:23:45] just AI stuff. But, so i- it's, it's a totally, it's more [00:23:50] intensive raw materials, I guess is how I would probably put it. So that's part [00:23:55] of the problem too And so like instead of being able to make three [00:24:00] iPhones out of those chips, now you can make one GPU and no iPhones.
[00:24:04] Andrew: And so it's [00:24:05] a resource problem, but the reason that we're here is because we don't [00:24:10] have a diversity of manufacturing for memory or chips. We don't have [00:24:15] any, supply, slack, right? These things have been operating at [00:24:20] massive... And like even during COVID, like there was a massive dip in it, and then everyone started buying TVs and [00:24:25] computers and video game systems and everything else, and there was a huge demand.
[00:24:28] Andrew: So instead of trying to [00:24:30] build to level out the demand And now we're here, right? We have zero [00:24:35] slack capacity. It takes seven or eight years to build a new fab. It, it's crazy. And there's [00:24:40] no sign of anybody even caring at this point, 'cause if you look at the numbers from SK [00:24:45] Hynix specifically they're not really investing in new f- new [00:24:50] facilities.
[00:24:50] Andrew: I think they're building one in the States. But it's been kinda planned for a [00:24:55] while. And again, seven years from now, great. One-- if we're really gonna AI the [00:25:00] world, and I'm not saying we are, but if we believe the hype of these companies, we're gonna need [00:25:05] several orders of magnitudes more production to meet the [00:25:10] supply need.
[00:25:10] Andrew: And it-- Again, people getting all pissed off about AI. It's not real because we can't [00:25:15] get the electrons for the data centers to operate, so just cool down. It's not [00:25:20] real. But if it were real, this con- this supply [00:25:25] constraint also is a linchpin, right? Even if we could figure out how to build all the transformers and bullshit to get all [00:25:30] the electrons in the data center, which we can't, it's another similar problem.
[00:25:34] Andrew: The [00:25:35] memory stuff the CPU stu- It's, it's just the, none of the supply chains are there. And so there's [00:25:40] a massive push to try to, to fix all this stuff, but it, [00:25:45] it
[00:25:45] Andrew: I'll pause. I feel like you have something to say
[00:25:48] Nathan: I, I think you made good points and, and these [00:25:50] were things that I wanted to bring up as, as well because you mentioned it takes seven or eight years [00:25:55] to build one of these fabs, right? So there is this element of pace layering where
[00:25:59] Nathan: the [00:26:00] supply for things we want to, to do right now versus the time it actually [00:26:05] takes to get there.
[00:26:06] Nathan: There, there is a timeline because, again, we live in a
[00:26:08] Andrew: Yeah.
[00:26:08] Nathan: world while these concepts [00:26:10] of AI demand or what have you are, are almost instant. [00:26:15] So people don't necessarily have that capability to plan quite correctly when it comes [00:26:20] to how things are gonna change versus what we currently have in, as infrastructure. And if you [00:26:25] want another example of this, I, I work in energy, so you look at the energy [00:26:30] grid. The energy grid is, is so old, but all this demand is shooting up now, and [00:26:35] it takes decades to build the infrastructure that we need to move the energy around. So [00:26:40] it, it's not something new. But one other thing I wanted to mention, we do [00:26:45] this on 3/8 is, is we don't undercut the difficulty of things. [00:26:50] Building a fab is difficult, and I think there's a reason why it takes so long to [00:26:55] build one up is because it's, it's very, very hard and it's very costly to build these types of [00:27:00] things. So
[00:27:01] Andrew: Yeah
[00:27:01] Nathan: the upfront effort and someone wanting to think about how [00:27:05] to actually get this off the ground is already a barrier enough to not wanna do it. So that, that's [00:27:10] an additional flaming hoop to jump through when we're talking about building supply chain for [00:27:15] these types of,
[00:27:15] Andrew: Yeah. The--
[00:27:16] Nathan: want now.
[00:27:17] Andrew: That's, a good comment 'cause the point I was gonna [00:27:20] steer us towards is that it is hard, [00:27:25] but we're not, we're not building for the future. [00:27:30] And I think that's the biggest problem I have with this, in that instead of building more [00:27:35] capacity to build like the nuts and bolts, we've decided we're gonna build like [00:27:40] assembled beams.
[00:27:41] Andrew: I don't know what the, the analogy is.
[00:27:44] Nathan: Yeah
[00:27:44] Andrew: [00:27:45] And what I'm really concerned about is that this push now will [00:27:50] be that compute is relatively affordable [00:27:55] today or recently. I'd still say it's pretty affordable, but it, it's getting to be [00:28:00] not affordable. That the, the concern I have is that compute is going [00:28:05] to become a, a...
[00:28:07] Andrew: no longer a, a commodity or on the commodity [00:28:10] track, and it's gonna become a scarce resource that only a few people with money can [00:28:15] interact with. And that's a huge danger. And so folks have made this [00:28:20] argument with AI, and I think it's hyperbolic and overblown, and this [00:28:25] stuff's gonna get cheaper, and it's gonna be more readily accessible.
[00:28:28] Andrew: But if you can't run the [00:28:30] model or you don't have the hardware to compute the model, doesn't matter. [00:28:35] And then the, the folks who do have the compute have no pressure to [00:28:40] do anything, right? And so it, it, it's kind of like what Apple's done on the iPhone is that they've [00:28:45] closed it down and made it a lot harder for people to develop software for.[00:28:50]
[00:28:50] Andrew: So that ecosystem now requires a level of specialization that you used to not have to [00:28:55] have. You used to be able to just tinker around with software and everybody I know, I'm old now, but everyone that [00:29:00] I know, l- I learned how to code, with the exception of a very few amount of people, were just like, "I [00:29:05] was on my computer," and I was like, "I wanted to do this, and I figured out how to make it happen."
[00:29:09] Andrew: Can't do that on the [00:29:10] iPhone. And and you-- there's trade-offs, right? It's more secure. It runs [00:29:15] better like better battery life, whatever you wanna call it.
[00:29:18] Nathan: guess.
[00:29:19] Andrew: But the, but it's [00:29:20] less open, and the more abstractions you get away from that, and it's less [00:29:25] accessible. And I'm very concerned with the...
[00:29:28] Andrew: I'm not concerned with [00:29:30] AI becoming a scarce resource at this point. But if [00:29:35] hardware continues to become ridiculously expensive and unobtainable [00:29:40] by average people, or just people without money, frankly, like I guess if you have money, it doesn't [00:29:45] really matter. Now we've lost an entire pressure mechanism to keep [00:29:50] the, the AI companies that are well-capitalized in check or provide [00:29:55] alternatives for people who don't want a cloud.
[00:29:57] Andrew: And there's a lot of-- if you are full AI [00:30:00] company, it's fine, but there's a lot of instances where running a model in a cloud makes zero sense. [00:30:05] And like you-- edge compute, as we call it, is gonna [00:30:10] be completely and totally unobtainable if the trajectory we're on keeps going.
[00:30:14] Nathan: Yeah. [00:30:15] Yeah.
[00:30:15] Andrew: And, and the, the other thing, if, if you still don't give a [00:30:20] shit, anything that runs software off your computer, [00:30:25] AKA the cloud or SaaS, is going to become more expensive.
[00:30:29] Andrew: The [00:30:30] hyperscalers, as we call them, Amazon Google, Microsoft, [00:30:35] there's a couple other players, they need to buy servers, and they can buy them at large [00:30:40] scales, but everybody's so focused on AI that that means the website that [00:30:45] run, your company runs is gonna get more expensive. The website that [00:30:50] your favorite product runs on is gonna get more expensive.
[00:30:52] Andrew: That service that moves the data from your phone [00:30:55] to another phone or allows you to connect with other people runs on a server, it's gonna get more [00:31:00] expensive. And so we're gonna see an inflation in all digital goods. We're already starting [00:31:05] to see a little bit of it, but it hasn't hit anywhere near where it's gonna be yet.
[00:31:09] Andrew: And I'm not trying to like [00:31:10] doom and gloom, like this is real, it's gonna happen. We can check back in a couple months. But the [00:31:15] concern I have there is that software has been largely deflationary, meaning it's gotten cheaper and [00:31:20] it's made things more accessible for the most part. I think we're gonna see a massive spike the other way, [00:31:25] and it-- what I hope is that it's just a spike and then it becomes deflationary [00:31:30] again.
[00:31:30] Andrew: Observing what I'm seeing and just the nature of the world right now and the [00:31:35] winner-take-all capitalism is all that matters, it seems like it's just gonna be an excuse to [00:31:40] continue to charge more and further monopolize tech [00:31:45] stacks and compute as a resource, and that means it's less accessible to everybody and just [00:31:50] gets more expensive
[00:31:51] Nathan: Yeah. The optimistic lens here is, is [00:31:55] scarcity is often the driver of innovation, right? So maybe there's, I don't know if you'd call it like a revolution of [00:32:00] sorts, but oftentimes we use more compute than we actually really need, right? So if I, if I go back to [00:32:05] the, the days of the early stage video games, they had limited hardware, which meant they, [00:32:10] they can do limited things.
[00:32:10] Nathan: And, and out of it came Mario and these great works [00:32:15] of art because they were resource constrained. So maybe
[00:32:18] Andrew: Yeah.
[00:32:19] Nathan: also head
[00:32:19] Nathan: [00:32:20] or one
[00:32:20] Andrew: maybe...
[00:32:20] Nathan: that we go in
[00:32:21] Andrew: But like The reason the personal computer revolution and [00:32:25] video games, all this stuff happened was because you had these, mostly Japanese [00:32:30] manufacturers of chips and diodes and capacitors, and they just flooded the market [00:32:35] with this shit, made it really cheap, and you could go to an electronics store and build your own board [00:32:40] to do stuff, right?
[00:32:40] Andrew: Yeah, literally RadioShack, right?
[00:32:42] Nathan: Yeah.
[00:32:43] Andrew: I don't even think they sell [00:32:45] anything that doesn't come in a box now. And for f- those are the basic [00:32:50] components of electronics, and you used to be able to literally just go to a store and buy them or [00:32:55] get a catalog. And you can still today, but the thing I'm afraid of is that everyone [00:33:00] will be chasing these massive money deals, and then they'll either spin down those [00:33:05] businesses or they won't be able to compete, or it just won't make sense for them to even [00:33:10] supply, and then they're just gonna push it all to the big suppliers, and then we lose, a massive [00:33:15] chunk of the electronics market.
[00:33:16] Andrew: We've already lost a massive amount of the electronics market. And [00:33:20] for folks to say it just gets better," maybe, but I also don't [00:33:25] want companies making decisions about what I can and cannot compute, how I can and [00:33:30] can do what I want. That's my take on the FLOSS, free and [00:33:35] open source model, open computing standards.
[00:33:38] Andrew: The reason we have the internet is 'cause [00:33:40] people were able to get under the hood and be like, "I wanna make it go faster," or, "I wanna make it do [00:33:45] this thing," right? "I wanna add six wheels instead of four." And y- n- now it's getting to the [00:33:50] point where you're not gonna be able to do any of that, and we're gonna lose innovation, and we're gonna [00:33:55] continue the monopolistic track.
[00:33:56] Andrew: And this is a very well-studied phenomenon in business. This is not take Andrew's word [00:34:00] for it. We know monopolies are really bad for everything, and they eventually [00:34:05] fail. It just takes a long time sometimes. And I'd rather not s-see [00:34:10] that happen. But here we are on the track, right? Yeah, no, it, it's...
[00:34:14] Andrew: If we don't [00:34:15] learn the, from the history, it's really, the last time we had massive consolidation like this was the oil industry, [00:34:20] and it was Standard Oil and Co. became massive monopolies on energy distribution, and [00:34:25] people are like energy's too expensive because there's a couple companies that run it all, so we're gonna change [00:34:30] that."
[00:34:30] Andrew: But we ha- well, now we're celebrating that there's only a couple companies that run AI and run technology, and that's a [00:34:35] good thing somehow. And it's, I don't think that's a good thing. If you wanna use those companies, great. But I [00:34:40] think you should have the option to not use those companies And I'm, I'm-- AI is [00:34:45] already making technology stuff, I think, hard to get into because it's so easy to be able to do [00:34:50] stuff that, to your point with constraints if you can just have AI do [00:34:55] the stuff, you don't put in the work to try to really figure out that problem to make it better.
[00:34:59] Andrew: Maybe if [00:35:00] AI is more expensive or less usable, then you will. But if everybody's renting their compute from the [00:35:05] cloud provider and they don't allow that or they don't allow you to modify stuff, then we're not gonna get the next level of [00:35:10] innovation.
[00:35:11] Nathan: Hmm.
[00:35:11] Andrew: All right. Happy Fri or Thursday
[00:35:14] Nathan: I was trying to [00:35:15] think of an optimistic note to end it on, but yeah, we're living in some
[00:35:18] Nathan: crazy
[00:35:18] Andrew: I don't think it's optimistic. [00:35:20] It, it and I, I think there's a big push in our society for positivity. We've talked about [00:35:25] pos- toxic positivity, and it's I don't think we need to spin it. I think it's really dangerous, and [00:35:30] I think we need to make corrective action, and we need our representatives and senators here in the [00:35:35] States to have policies f- that enforce antitrust, that enforce [00:35:40] diversity of thought [00:35:45] and, ways of doing things, even if it's not the most efficient.
[00:35:48] Andrew: We have this push for [00:35:50] all efficiency, but sometimes non-efficient ways yield different results in ways that we couldn't [00:35:55] predict. And we need to invest in, in those types of things and encourage [00:36:00] people to start up s- businesses and stuff, [00:36:05] and I don't-- Without doing that, it's just going to be a slow slide.
[00:36:08] Andrew: It's just already a slow [00:36:10] slide, in my opinion, and it's only gonna get worse. And so this is maybe a wake-up call to folks [00:36:15] to be like, I understand AI is like a love or hate thing for a lot of people. That's a huge [00:36:20] problem because it's like a binary. It's like following American politics.
[00:36:24] Nathan: Yeah [00:36:25]
[00:36:25] Andrew: it, it's not good or bad. It's how we use it. Like everything else, you have to [00:36:30] get out of that mindset, and you need to be pushing your representatives and, and folks in [00:36:35] Congress to enforce antitrust laws. This is the same thing for social media. [00:36:40] This is the same thing now for these AI companies are starting to suck up all these other competitive companies that [00:36:45] are either improving ways to make models more efficient 'cause they need it.
[00:36:48] Andrew: And so then that takes away [00:36:50] the ability for someone to self-fund a lab or to, to tinker with your own... You can train a [00:36:55] pretty decent sized model on, half a million bucks, which sounds like a lot of [00:37:00] money, but in
[00:37:01] Nathan: I was waiting for, I was, I was waiting for either the hardware spec or the amount of
[00:37:04] Andrew: I know, [00:37:05] right? You can train a decently sized [00:37:10] usable model on a GPU in your computer, depending on what you're trying to do and [00:37:15] how specified you're, you're trying to get. But these general purpose models [00:37:20] half a million bucks, a million bucks gets you a really, really long way, and [00:37:25] that's only possible if that compute is available or people have the [00:37:30] ability
[00:37:31] Nathan: right?
[00:37:31] Andrew: to leverage other technologies.
[00:37:33] Andrew: And so i- if, [00:37:35] if the companies that are building the AI are sucking up all of the other companies that are contributing to that [00:37:40] ecosystem as well, then there isn't an ecosystem for everyone else to build and learn and, and [00:37:45] build on. It's the same thing with internet providers, right? If, if there's one company in your town who [00:37:50] owns all of the lines in the ground, then they set whatever price they want, and you don't...
[00:37:54] Andrew: You get [00:37:55] crappy speed, or you get crappy service, or the internet goes out all the time. If you live in an area like I do, I have four [00:38:00] ISPs to my door.
[00:38:02] Nathan: Wow
[00:38:02] Andrew: I have gigabit fiber, and [00:38:05] I think that's 45 bucks a month. That might sound like a lot for some people, but [00:38:10] that is... I know people who pay hundreds of dollars a month for the exact same service.
[00:38:14] Nathan: For fun. Yeah, [00:38:15] yeah
[00:38:15] Andrew: And I-- And if my provider of choice here decides to, [00:38:20] to, raise their prices, I can go to
[00:38:21] Nathan: you can go to,
[00:38:22] Andrew: three other ones if I want. But I guess, [00:38:25] plus cellular, so I probably have seven or eight ISPs. That's very rare for [00:38:30] mo- most people have one or two options, right? But, but the point is that if you don't have a bunch [00:38:35] of different options and people trying different things, so it's okay, everybody is an ISP.
[00:38:39] Andrew: Great. Now [00:38:40] what? You figure out a way to bundle email, and that was, like, the big thing back in the day, right? You get an email account, [00:38:45] you get 50 gigs of space or 10 megs or whatever. And then filtering and, like, all these... And [00:38:50] then some people, especially more technical people, are like, "Yeah, it's all bullshit."
[00:38:53] Andrew: Yeah, but for some people, that's [00:38:55] valuable. But that's competition, right? It's breeding other ideas, and those ideas [00:39:00] clearly didn't pan out 'cause I don't think any ISP offers email, or it's like a Gmail or something now, right? [00:39:05] But and things change, and that's okay. But we would've never gotten [00:39:10] the fiber technologies we have if we wouldn't have had competition and, the [00:39:15] cable companies for a long time, and this is more, we're, like, way out, but we're in c- in, in [00:39:20] competition land.
[00:39:21] Andrew: The cable companies for a long time were locked in physics. You can only transmit so [00:39:25] much information on a copper wire, and the f-
[00:39:28] Nathan: yeah, right
[00:39:28] Andrew: light pulses, you can [00:39:30] transmit more. But then somebody figured out we can do all sorts of [00:39:35] different weird things with frequencies and transmit, I think it was, like, [00:39:40] eight X more information on the same copper wire.
[00:39:43] Andrew: So there was in- there was an innovation [00:39:45] spike where it was like this is unacceptable. We have to be able to compete with fiber. We can't just rerun [00:39:50] cable everywhere. And ta-da, right? So I think the, the way out of it and the [00:39:55] optimistic way is to enforce competition and enforce different ideas [00:40:00] and support different ideas.
[00:40:01] Andrew: So again, if you're mad that Apple raised their prices 20%, there's all sorts of [00:40:05] companies you can buy computers from that aren't Apple Right now with AI, there's [00:40:10] there's the big three American, and then there's what the big three Chinese companies. [00:40:15] France has one. There's a lot of interesting projects that are starting to [00:40:20] crop out out of Scandinavia specifically,
[00:40:22] Nathan: Yeah, yeah. Uh-huh
[00:40:23] Andrew: uh, where people are starting to [00:40:25] get on this idea of we can't just let Google and, and Anthropic and the other [00:40:30] company make all the decisions and the advancements, and we need to invest in this too, which is great.
[00:40:34] Andrew: But [00:40:35] if we don't have access to run those models ourselves or in, [00:40:40] communal ways where we could all get together and, and start a company to, to build some [00:40:45] AI thing, then it, it's all moot in my mind too, 'cause you're still at this high dollar [00:40:50] institutional investor. And then for people listening yeah, it's a n- new technology.
[00:40:53] Andrew: That's how it is. [00:40:55] Great. But the cost or the means of production for this are [00:41:00] being artificially inflated in cost, and that's the fundamental problem I have.
[00:41:04] Nathan: [00:41:05] Interesting. I don't know if I'd say artificial, but they're definitely being inflated
[00:41:08] Andrew: It, it's artificial [00:41:10] because there's a, a monopolistic control over supply [00:41:15] and there's an artificial demand spike. The AI companies have committed to [00:41:20] build billions of dollars and billions of chips and all this nonsense, but the reality is they can't even rack it in [00:41:25] a data center. It's all fake
[00:41:27] Nathan: It's,
[00:41:28] Nathan: it's
[00:41:28] Andrew: it's 90% fake. They're, [00:41:30] they're building some data centers, but the amount...
[00:41:32] Nathan: is still there. The demand is not artificial,
[00:41:33] Andrew: Is it?
[00:41:34] Andrew: [00:41:35] It's an artificial demand if you can't actualize it, right? It, it's a non-- It's [00:41:40] a, it's a non-committed demand So they have a promissory note to say, [00:41:45] "I'll probably buy this amount of stuff from you given all of these conditions being true."
[00:41:49] Andrew: That's not [00:41:50] demand, that's fake, because Apple probably could buy all the chips to [00:41:55] make a bigger computer. They've actually had to limit their hardware options because they don't have enough memory
[00:41:59] Andrew: [00:42:00] So like they're actually selling products. The AI people are selling a, an AI [00:42:05] product and they're having to create servers and data centers and shit. [00:42:10] Those aren't real because we can't power them, we can't build them fast enough, we [00:42:15] can't connect them. There's all sorts of other problems. We're solving them, but like [00:42:20] I think that there's been like 300 and some data center proposals and the [00:42:25] reality is we can probably build like 20 of them Because of just like you [00:42:30] literally, if you could build the building and put the servers in, which you can't you can't hook them up to the [00:42:35] grid for power.
[00:42:35] Andrew: So okay, great, you have a building sitting full of Nvidia boxes that [00:42:40] don't turn on. That's not demand, right? 'Cause they're not gonna buy those boxes if they [00:42:45] can't turn them on within a reasonable amount of time because they're just gonna buy the next one that comes out. So they've [00:42:50] created by bas- basically like promissory notes of purchase, [00:42:55] artificial demand spike on the, the market, and some of it, the demand is real, but...
[00:42:59] Andrew: And some of it is, [00:43:00] is probabilistic demand, if we could put these all in data centers, we would, but there's so [00:43:05] many other dependencies that it's not real.
[00:43:07] Nathan: Yeah. I, I guess I, I'm getting semantic because the [00:43:10] demand I think is real. Whether you can actually deliver on whatever you're [00:43:15] going to do with the demand is, is not-- That's a
[00:43:17] Andrew: but it's not even--
[00:43:18] Nathan: right?
[00:43:18] Andrew: it's like a futures market [00:43:20] problem. So if I'm, if I'm trading future contracts, which, oof [00:43:25] really quick and simple is the... Yeah, but no, this is-- But this, [00:43:30] this futures markets were designed for this problem, and they're not being applied to [00:43:35] this problem because they're, I don't know, there's a whole, there's a whole, that's a whole podcast to get into.
[00:43:39] Andrew: [00:43:40] But for a futures mark-
[00:43:40] Nathan: physical goods, right? They don't deal with delivery on future-- on, on
[00:43:43] Nathan: physical
[00:43:44] Andrew: what's a [00:43:45] chip? What's a server? What's a power unit? That those are all physical [00:43:50] things that they've promised to buy with a bunch of conditions attached, right? That's, that's-- So for the [00:43:55] futures market, so for folks to maybe get in the frame here, the-- when you [00:44:00] have crops and you're growing stuff, you might, you, you grow an acre of crops and you expect to get an [00:44:05] acre of corn, let's say.
[00:44:06] Andrew: You may or may not get an acre of corn because the flood or [00:44:10] burning crops or whatever, right? Too much rain yields, [00:44:15] yeah. You, you decide not to sell half of it, right? And so people can [00:44:20] place futures on those to say, "In six months, I will pay this much money [00:44:25] for an acre of corn,"
[00:44:27] Nathan: X amount
[00:44:27] Andrew: 10 bushels or whatever, right?
[00:44:29] Andrew: The thing with [00:44:30] that is, is that futures contract obligates you to take receipt of [00:44:35] what you purchase. It's a, there's, there's more stipulations. There's some other things in there, and [00:44:40] value can materialize, immaterialize, but effectively that's what it is. This is not that, and [00:44:45] the, the AI companies are trying to sell everybody that that's what this is.
[00:44:48] Andrew: They are saying that we will buy [00:44:50] 10 million GPUs if we can build these six data centers, and these six [00:44:55] states give us the power and environmental waivers, and, and, and, and, and, and, and. There's all of [00:45:00] these dependencies that sit on these promises. But what's happened [00:45:05] then is that these companies are booking out all of their production, the future [00:45:10] production, based on fake promises that probably won't [00:45:15] materialize.
[00:45:15] Andrew: And will they-- Maybe they will, maybe they won't. Andrew says they won't to anywhere near the [00:45:20] capacity that, that people are, are speculating on. Maybe I'm wrong. But the reality [00:45:25] is the, and the market-based solution for that is [00:45:30] futures, and that's not what's happening. And it's, it's a little weird in most manufacturing circles to, to [00:45:35] book out production that far in advance.
[00:45:38] Andrew: And there's-- It's a contract. Like [00:45:40] everything else, I can say, "Okay, I'm not gonna buy it And that's why, that's a l- that [00:45:45] is the primary reason that these memory manufacturers don't build excess capacity because they know there's a [00:45:50] dip coming. They don't know when it is, and they'll just make the chips anyways 'cause there's people to sell [00:45:55] them to.
[00:45:55] Andrew: And because there's an artificially short supply because they refuse to build [00:46:00] more capacity and there's no other players in the market, they can do that. That's the problem I [00:46:05] have
[00:46:05] Nathan: Hmm. So wait, so where, where does it go from, from there though? So [00:46:10] even if, if you can't deliver on these, on, on this artificial demand or the whatever you wanna use [00:46:15] as artificial demand for, the- there will, like you said, naturally be a point where it being stabilized in terms [00:46:20] of prices for
[00:46:21] Andrew: but, I,
[00:46:22] Nathan: November
[00:46:22] Andrew: the prices for stuff have been bid [00:46:25] up artificially so high that I don't think there [00:46:30] is
[00:46:30] Andrew: Because the prices are artificially so high and there's no [00:46:35] demand curve correction because there's no new supply being [00:46:40] built, that I don't think the, the prices will correct [00:46:45] anytime soon because the primary bidders for the, the chips [00:46:50] are competing with other people, and there's, there's secondary contracts that say if if you end up with [00:46:55] extra capacity, we'll buy it at this price.
[00:46:57] Andrew: While that price used to [00:47:00] be I don't know, 20% of what the actual contract would cost, and it's a [00:47:05] futures contract, and it, it's almost an option to, to [00:47:10] purchase chips if somebody else pulls out, and that used to be not really a [00:47:15] premium. Now it's a premium product, and that the, the cost of those secondaries, if you [00:47:20] would, are now also astronomical.
[00:47:23] Andrew: So when it-- [00:47:25] let's say a RAM chip costs 100 bucks. I contract you for two years at [00:47:30] 100 bucks a chips, and I'll buy X amount of chips. You come to that same supplier and say "I'll, [00:47:35] I'll do a contract for 90 bucks a chip, and I'll pay you [00:47:40] $10 up front for that, that ability to, if Andrew can't [00:47:45] make them or, or can't get rid of them or can't sell them, or maybe Andrew then will sell them to me," and there's all this [00:47:50] weird shit, right?
[00:47:51] Andrew: But the problem is because this is a private market, [00:47:55] because there's no market economics at play, and there's no com-- more importantly, there's [00:48:00] no external competition, they've bid up these prices [00:48:05] totally out of control. And this sort of stuff used to happen in agricultural and, and other [00:48:10] commodities markets, which is why we have futures and why we have all these other market-based...
[00:48:13] Andrew: It's I guess maybe [00:48:15] one of the solutions would be to have the price of chips traded. I [00:48:20] don't know what that would look like. I have not actually thought anything about it. But it's like, why not, if I can trade [00:48:25] grain, why can't I trade DRAM?
[00:48:26] Nathan: That would be an interesting world.
[00:48:28] Andrew: But I [00:48:30] think the reason that doesn't work is because there's two companies that produce 98% of the supply. And if there were more companies, then you [00:48:35] could get a consortium of smaller companies together and they could say, "Look, we're gonna go to the market and we're [00:48:40] gonna, have a futures price for the product that we're producing.
[00:48:44] Andrew: And it's risky [00:48:45] for us 'cause we might lose some money, but realistically we probably won't." And then everything [00:48:50] would commoditize. Maybe. I don't know. We're in market theory now, and I guess I haven't really thought [00:48:55] about it, so like don't, don't... It's a thought experiment
[00:48:58] Nathan: all, all I was gonna say was, "Yep [00:49:00] we know what your next business venture is."
[00:49:02] Andrew: I know, right? Big DRAM. [00:49:05] But anyway and I don't know, th-this gotten quite technical for folks, [00:49:10] but fundamentally, again, it, it's the, the lack of competition is yielding [00:49:15] higher prices, which is always what it does. But we're in such an artificially inflated [00:49:20] market space right now, and these AI companies have come in so well-capitalized [00:49:25] and bid up stuff so expensively that, it's the same kind of [00:49:30] thing that's happened to the housing market is it's one person comes in the neighborhood and pays a million [00:49:35] dollars for a $100,000 house.
[00:49:36] Andrew: Now all of a sudden everybody's $100,000 house is worth [00:49:40] 180. And then someone else comes in and pays 220 or 250 for it 'cause they really want a house, so they really... [00:49:45] So now everybody's goes up, and eventually it gets to the point now where no one can really afford [00:49:50] to buy a house because it's just too expensive, and nobody can afford to sell their house [00:49:55] because there's nowhere else to move because you need to sell your expensive house in order to buy a-another expensive [00:50:00] house.
[00:50:00] Andrew: And any profit that you would've gotten in that ownership
[00:50:04] Nathan: It goes
[00:50:04] Andrew: is [00:50:05] gone, right? 'Cause you had to buy it at a higher rate. And we're-- this is the same kind of pr- we're stuck in, right? And [00:50:10] that, the whole, the housing thing is a whole 'nother thing, too, but it's the same type of idea. And what we've [00:50:15] s- kinda like what we talked about last stream is we've seen the previous generation take [00:50:20] every little bit of slack out of the market and create, and allow for the creation of all these [00:50:25] monopolies to happen.
[00:50:26] Andrew: And now we have to go through and de-Rockefeller ourselves like we [00:50:30] did back in the day and make rules about antitrust and break up Standard [00:50:35] Oil, AKA big AI, into smaller companies so they, they don't have an outsized [00:50:40] market position. And for all the free market people listening, it's not a free market if you [00:50:45] have a monopoly
[00:50:46] Andrew: You can argue about who changes it, if it's the government or a [00:50:50] private corporation like the SEC spun out like the FCC or something. But it-- Once [00:50:55] you've accumulated so much capital in a capitalist system, it doesn't [00:51:00] self-correct. It runs away. And we used to understand that problem, and we used to curtail it, [00:51:05] and now we've, we've just been like if it's gonna run away, why don't we all get a little cut in it, and then we [00:51:10] can all run away together?"
[00:51:10] Andrew: And it's that's not how it works either
[00:51:12] Nathan: Yeah, yeah, ride the waves.
[00:51:14] Andrew: All right. [00:51:15] That's my rant.
[00:51:16] Nathan: It was a good one. It was, yeah, I think there were three points where, where we were supposed to end [00:51:20] the stream, but
[00:51:21] Andrew: Yeah
[00:51:21] Nathan: it's a good one. You're passionate about the topic um
[00:51:24] Andrew: [00:51:25] Well, yeah, and I, I-- The headline I saw it was just, poor Apple [00:51:30] trying to protect us poor consumers from these massive price...
[00:51:34] Nathan: that's what
[00:51:34] Andrew: It's [00:51:35] multi-- It's an over trillion-dollar company. And I, I get it from their business. [00:51:40] They can't just take a hit either because then their share price goes to the tank [00:51:45] because they lower their margin.
[00:51:48] Andrew: And nobody knows-- And I think part of the problem [00:51:50] with this problem too is nobody knows when it's done. Usually the, the last time we had a [00:51:55] massive electronics supply strike was, the floods, I think it was in [00:52:00] Korea and Indonesia, where all the hard drives were made. There's this one [00:52:05] valley, and it's 80% of the hard drives in the world were made there, so all the storage couple [00:52:10] factories got blown out or whatever, and everyone was like, "Oh, shit."
[00:52:13] Andrew: And then prices [00:52:15] tripled over a couple weeks, and then it took three years for them to normalize.
[00:52:19] Nathan: [00:52:20] Hmm
[00:52:20] Andrew: If you could say "Oh no, the factory got wiped out, and it takes five years to build a factory," or [00:52:25] whatever the math is, you can be like, "Okay, in a couple years this will probably..." But there's no-- there's [00:52:30] seemingly no end in this race because it's [00:52:35] not grounded on anything fundamental.
[00:52:36] Andrew: It's grounded on everybody's speculation and hype, and so it's [00:52:40] impossible to predict when it will crap, but it will. It will 100%, [00:52:45] right? It's 100% a bubble, but nobody knows when it... Nobody ever knows when a market [00:52:50] bubble or a correction happens, but when there's-- But if there's fundamentals,
[00:52:54] Nathan: Yeah. [00:52:55]
[00:52:55] Andrew: Understand.
[00:52:55] Andrew: You're like, "Ah, this feels like a bubble." But the- when there's no fundamentals, there's no facts, there's no [00:53:00] data that anyone's driving on it, it's just all made up and no one knows. But [00:53:05] physics is physics. The laws of reality are eventually gonna... And we're talking about physical goods. We're talking about [00:53:10] manufacturing.
[00:53:10] Andrew: We're talking about physics, moving atoms and electrons. It will [00:53:15] correct. It has to. It's the law of the universe.
[00:53:18] Nathan: What a place to end. It's a lot of[00:53:20]
[00:53:20] Andrew: But is that, is that happier for you?
[00:53:22] Nathan: I think we took a rollercoaster ride, but [00:53:25] we got somewhere.
[00:53:27] Andrew: Y-you, you really wanted to end on a positive note. I [00:53:30] think that's as, as a positive as I can get
[00:53:32] Nathan: No, I was looking for a positive note to end on. Didn't [00:53:35] necessarily mean we would find one, but All right.
[00:53:37] Nathan: got somewhere.
[00:53:38] Andrew: Happy Thursday
[00:53:39] Nathan: [00:53:40] Yes. See y'all next week.
[00:53:41] Andrew: See ya