TRASHFUTURE - *PREVIEW* Son of Sam Altman ft. Ed Zitron

Episode Date: September 5, 2025

  Ed Zitron returns to talk about the AI bubble, market concentration, and popular backlash. We also look ahead at what would happen when the AI bubble begins to pop. Check out Ed's work here! Get the whole episode on Patreon here! MILO ALERT: Check out Milo's upcoming live shows here! MERCH ALERT TF Merch is still available here! Trashfuture are: Riley (@raaleh), Milo (@Milo_Edwards), Hussein (@HKesvani), Nate (@inthesedeserts), and November (@postoctobrist)

Open AI, turns out it's gone public and its revenues rock solid. Yep. Not losing anything. Not losing a dime. No, no, no. It's Ed. We're back to do a big mega roundup of developments in the AI economy. I try to sort of round some of those up for a while, save them for Ed, because it's just, it's good to look at them all at once.

And before we get into it, I mean, I'm going to say before we get into it and another thing, like 50 times this episode, because I have so much I want to talk about. But Sam Altman's newest disclosure on the previous Open AI investment round, which keep me honest on my numbers here, Ed, I believe has valued it at $400 and $600 billion simultaneously depending on who's buying and selling the shares. No, no, it's actually better than that. It was $300 billion in April. And now they are doing a second employee share sale that has grown from $5 to $8 to $10.3 billion for a valuation of $500 billion, which is around the market cap of Netflix. But the best part is, is I generally think that when you think a company's going to survive, you do like a $10.3 billion employee share sale. I don't think that anyone has ever sold this many shares before. And it's just like, just like two weeks after Samuel, it was like, yeah, it's a bubble.

Yeah. Check out this cool bag holding opportunity. And what's great is they're selling shares. This is what really melts my brain. So they did an $8.3 billion round like a month ago. And Dragon Air involved themselves in it. But now SoftBank and Dragonair are investing at $500 billion, despite already investing at $300 billion like a month ago. It's just pumping their assets.

It's insane. And I will write about this and I will feel crazy every time I do because everyone in the tech media is like, yeah, this looks great. There are retail investors currently holding a bag so suspicious that it's not even worth podcasting about anymore. I'm texting British Transport Police. There's no retail money even in this. It's just venture capital eating itself. Oh, good. Actually, no.

Good. Yeah, that is good. It's good when those people lose money. And now the negotiations between Microsoft and OpenAI so that they can actually convert to a for-profit going into next year, which means that SoftBank will cut their round in half, which is all good. This is all fine. Everyone's doing well.

Is there anything built on top of this that we need to worry about? I hope there isn't. Nah. It's fine. Good. Some stuff. No one is using the economy. So this is the funny thing on this latest sale.

The prospectus says, Investing in Open AI Global LLC is a high-risk investment. Investors could lose their capital contribution and not see any return. That's standard, standard practice. It would be wise to view any investment in Open AI Global LLC in the spirit of a donation with the understanding that it may be difficult to know what money will play in a post-AGII world. However, now we do need some American money, please. Just trust us.

Universal basic income. In fact, universal remarkably high income once we source out everything. You're not going to need this start where we're going. Can Rocco's basilisk could get like a fiver off you? It says, so this is, as you mentioned, Sam's saying,

AI is a bubble. He specifically said, are we in a phase where investors as a whole are over-excited about AI? My opinion is yes. I'm sure someone's going to write some sensational headline about that.

I wish he wouldn't, but that's fine. But the models are already saturated the chat use case. They're not going to get much better. And in fact, they might get worse. Which I think he said because he realized he did an oopsie

because GPT-5 is bad. They were like, it's going to be great and then it's bad. And then what he did was he was like, but GPT6 is going to be incredible. Which I love, I genuinely like, clammy Sam Hortman. I love him, clammy will. And he's he goes out there, he's just like, well,

GPT5 sucks, so GPT6 will be better. And the fucking, anyone who reported that without any, without, just completely seriously, should just quit. Just you can't fucking do journalism anymore. This guy is going to fucking wallet inspect you every year or so and you will just accept it.

But what was funny was I was reporting out GPD5 because I got one of the infrastructure providers to tell me how it works. And it's fucking hilarious. What they've done is they were saying, oh, yeah, it's a super smart router to choose the right model. They chose an LLM to choose which LLM to use. It fucking rocks. It's so cool. And it's less efficient. It's significantly less efficient.

I'm not going to go into tech details. They're very boring. But they specifically made it so you can't cash a certain part of each prompt. So they just do a bunch of extra work for no real return. It doesn't appear to save the money. In fact, it might lose them extra money. And this is GPD5, the thing they've been talking about for two years. It's meant to be the biggest thing ever.

It's so cool. And people are still taking this shit seriously. Well, what I'd like to remember as well, remember when we talked about Deepseek, and we said, what's going to happen in the end of Deepseek is that the American AI sort of superscalers, what they're going to do is they're going to go all in on size, and they're getting it way more inefficient. They're going to do way more token burn.

They're going to build way more server. farms and it's not going to produce, and it's going to be diminishing marginal returns. So I think what we can do is put a big X in the spot next to, we were right about that. That's what happened. 100%. Yeah. I hate that the name of the podcast means that any time we were right about something, it's not good. Yeah.

Well, there you go. I mean, it marked that on your big trash chute your wall charts at home. The Trash Rucho World Cup of everything is fucked. But I also, I want to do a quick temperature check as well, right? Right. This temperature check is of a CEO I follow. Vlad, I like follow what he says and thinks. This is the CEO of Robin Hood.

I think he's a good barometer for what the very wealthy think of the soccer economy. This is the show from Newlingham? So, Ten have said, I do worry generally about the short-term impacts of AI on the labor market. I think there's going to be a lot of labor force disruption. I think that's actually a huge

tailwind for the importance of retail investing. So I think it's going to be much more important to invest because if you can't rely on labor to generate money to make a living, capital becomes much more important. Um, reading reading my copy of like capital volume one upside down with a bunch of the pages falling out and I'm just like is this good does this hold up it's so cool being like normal people should just consider living off their investment yeah what he means is gambling what he means is if no one is working we're going to Vegas you know like instead of all this economy shit right we're cashing in the podcast and we are going to Vegas we're going to put it all on red you know That's what Marxism is about.

That's why it's called the All In podcast. So basically what Ted have said is the government needs to get everyone a Robin Hood account so kids can start investing in a young age. Because the concept of working for living is going to be gone. But because we're still capitalists, we still need a world of scarcity.

Instead, what he sees is a world where AI has all the jobs. Everyone is entrepreneurs. And you're sort of just trying to gamble on capital allocations. which I guess is this is what Hayek wanted, basically. I feel like it's not even a dystopia that makes sense, though. Like it's like in the real dystopian future with like AI, I mean, this is making a lot of assumptions about AI being able to do any of this shit, right?

But let's take them at their word and assume the fantasy world where AI is doing everyone's job. I feel like what will actually happen is like there's a sort of class of people who own all of the AI who are like incredibly rich and everyone else is living in like the waste. Yeah, well, this is the vision they see for themselves. That's why Mark Zuckerberg keeps building castles. He's like, well, yeah. But the other thing about this is, other than the insane proposition of we need everyone to start investing in the stock market so that they can protect themselves from losing

their jobs, just one of them, how, I don't know how I miss this. This is, I'm going to be thinking about this all day because this is an insane thing. But the brain madness on these fucking people, these models can't do any jobs. They're replacing translators, transcribers, like real. Like, those are jobs, but the way these people are talking about it is like every knowledge work, every worker in like a shop is going away and it's happening now. I think it keeps stuttering because I'm just trying to find a nice way of saying, are we ruled by fuckwits?

Is everyone, are they all fucking stupid? And the answer is yes. I wrote this thing called the era of the business idiot a few months ago because it's like, what if they're all just fucking morons? Because in that future where they own all the AIs, they probably would still fuck it up too. you'd still have all these people having all this money you'd have just more Elon Musks

you'd have more people are the woke people in the wasteland are eating each other when they should be making more babies for me My Tesla robot plumber has electrocuted himself and drowned in my toilet Optimus plumber

Yeah it's really what they're talking about is the vision of the Vlad Tenev has which I think is the vision of your standard business idiot on the grift side of the economy is the only things that are going to be left is gambling and hype. That's it.

Everything else, you're trying to hype stuff that you've gambled on and that's it. That's what humanity is to be reduced to. Yeah. So really what you want to be doing

is gambling with a hype man and then you're like, right, then you've got the economy by the years. You're like playing blackjack and you'll like hit me and a guy's behind you like, yes.

So let's go. I want to talk about some numbers. I want to get, I'm going to say like let's get into it like 30 times. But let's get into how open AI and anthropic have sort of invented new kinds of math

to make themselves look profitable. This is Anthropic, I believe, is not counting. What is it? Because Anthropic released a revenue statement recently where they decided just not to count one major portion of their costs and be like, we're profitable now. Yeah, they're right. They leak their gross profit margins.

And Anthropics gross profit margins. I forget exactly what they are, but they don't include training costs for their models, which are most of their costs. And they're still not profitable. They're still going to lose $3 billion. That's the other thing. They're like, oh, yeah, we're going to lose $3 billion.

this year, which is less than the $5 billion we lost last year. And it's like, no, you're not. You're going to lose more. There's no, you had, they also, and here's a really, okay, so as you can tell this stuff slightly driving me insane, because Anthropic claims that they went from making $72 million a month in January to $400 million a month in July. And I just don't believe them. I don't believe them at all.

And they also released this thing called Claude Code, which is just, you don't need to, it's a coding LLM thing. But through it, you're able to work out how much token burn you're doing by using it. You pay $200 a month at max. There are people who have leaderboards of how much money they're spending on this platform, how much they're costing Anthropic. There's someone who spent $10,000 in a month. For $200.