Oversættelse og strukturering af YouTube-transcript om AI-boblen og datacenter-gælden.(Mistral)
Indledning: Boblen er tæt på at briste
Elon Musk har netop indrømmet, at opbygningen af AI-datacentre er én kæmpe boble. Uh-oh. Det er et kæmpe problem. Og oven i det har flere finansielle institutioner angiveligt stoppet med at yde lån til AI-virksomheder.
Smart. Det betyder, at boblen er få centimeter fra at briste. Men vi har endnu mere deprimerende tal til jer, for når først I ser størrelsen på faren her i denne AI-boble, er det super skræmmende, og I må tage forbehold. Anna har hele historien.
Musks advarsel: Gældsmarkederne lukker
Først og fremmest et stort shout out til Taylor Ellis, som har produceret dette indslag og gjort det let at forstå for dem, der ikke rigtig er en del af tech-verdenen. Men lad mig fortælle dig, hvad Elon Musk specifikt havde at sige. I går skrev han dette på X:
I de seneste par uger har gældsmarkederne næsten helt lukket adgangen til penge for enhver virksomhed, der ikke har betydelig omsætning uafhængigt af AI-datacentre. Det burde egentlig have været tilfældet hele tiden, men okay. SpaceX lukkede en gæld på 40 milliarder dollars til grafikprocessorer, men vi har investeringskarakter og har omsætning fra Starlink og opsendelser. Det betyder, at de 200+ neoclouds kommer til at se død og konsolidering på guppy-niveau. Nogle kommer igennem, men ikke mange.
Hvad er en "neocloud"?
Okay, så tænker du måske: hvad fanden taler han om? Lad mig bryde det ned, så det er let at forstå. Først: hvad fanden er en neocloud? Tja, en neocloud er en virksomhed, der i det væsentlige bygger beregningsinfrastrukturen til AI-virksomheder. Og Coreweave er blot ét eksempel på en neocloud-virksomhed. Husk navnet – neocloud – for det dukker op igen senere i indslaget.
Neocloud-virksomhederne optager nu gæld for at bygge deres infrastruktur ud. Men ifølge Musk er gældsmarkederne ikke længere interesserede i at yde lånene til disse virksomheder. De er ikke længere enige i – eller interesserede i – denne forretningsmodel. Og mens Meta og Alphabet, som bygger deres egne datacentre, tilbyder andre produkter, der skaber omsætning – som Facebook og Google – har andre neocloud-virksomheder måske ikke den samme opbygning eller samme type indtægtsstrøm fra andre diversificerede dele af deres forretningsmodel.
Nvidias rolle: En gensidig fordel – men manglende kapital
Der er desuden en temmelig magtfuld figur involveret, som ikke ønsker at se neocloud-opbygningen kollapse: Jensen Huang, direktøren for Nvidia. Nvidia har investeret i neocloud-virksomhederne, og neocloud-faciliteterne bruger Nvidias produkter – så der er en slags gensidig fordelagtig situation her.
Men her er problemet: Selv Nvidias direktør har indrømmet, at der ikke er penge nok at låne ud til disse neocloud-virksomheder. I et interview med CNBC sagde Jensen Huang:
Efterspørgslen på neoclouds er helt vild. Det, der mangler, er arbejdskapitalen til at brænde den vækst, for der er simpelthen så meget efterspørgsel på tværs af så mange sektorer af kunstig intelligens. Der er en mangel på kapital. Efterspørgslen på kapitalen er meget større end tilgængeligheden af kapitalen. Det kommer til at blive løst.
Han tror, det bliver løst. Kun tiden vil vise, om han har ret.
Coreweave: Et eksempel på krisen
Coreweave er et godt eksempel på, at långiverne ikke længere rigtig er interesserede i at yde den gearing, der er nødvendig for disse opbygninger.
Coreweave har set sine aktier falde 41 % siden 6. oktober frem til fredag sidst. Aktier i andre neoclouds har også haft tocifrede fald i samme periode. Coreweave er nødt til at udstede obligationer – lidt ligesom den amerikanske stat gør med statsobligationer – for at finansiere deres opbygning af beregningskraft. Men der er et problem: Tidligere i denne måned faldt prisen på deres obligationer så meget – og deres afkast steg tilsvarende så meget – at deres obligationer er klassificeret som distressed (truede). Så obligationsvejen fungerer heller ikke rigtig for Coreweave.
Analyst Gil Loria udtalte, at det at låne til meget høje renter for at bygge datacentre, som i stigende grad er commodity-produkter uden for hyperscalers, måske ikke ender med at være en god investering: »Jo mere gearing vi bruger på at bygge datacentre, jo mere risiko bringer vi ind i systemet.«
Jeg kan ikke understrege nok, hvor vigtigt det punkt er, for husk: disse overgærede virksomheder bliver securitiseret.
3,7 billioner dollars i gæld – fordelt på ni virksomheder
Ifølge Americans for Financial Reform har blot ni virksomheder – Alphabet, Amazon, AMD, Broadcom, Meta, Microsoft, Nvidia, Oracle og SpaceX – hele 3,7 billioner dollars i gæld.
Nvidia er desuden investor i Coreweave. Analytikere forventer, at virksomheder i S&P 500 rapporterer et spring på 30 % i tredjekvartalsindtjegninger i forhold til året før, men kun to virksomheder – Nvidia og Micron – står for 34 % af den vækst ifølge en analyse af Dow Jones-markedsdata.
Endnu farligere: Gælden havner i dine pensionsopsparing
En ny rapport fra Americans for Financial Reform viser følgende:
Spekulativ datacentergæld bliver pakket om og solgt til pensions- og aldersopsparingsfonde, universiteternes kapitalfonde (endowments) og andre institutionelle investorer på en måde, der er uhyggeligt – uhyggeligt – mindende om subprime-boblen.
Subprime-boblen handlede om rovdyragtige realkreditlån. Grundlæggende taler vi her om overgærede AI-virksomheder, som meget sandsynligt kommer til at fejle – og de er involveret i din portefølje, din pensionsportefølje, din livsforsikringsportefølje. Det er en katastrofe af episk proportioner.
»Når det havner i dine pensionsopsparing, er du den sukkerklods ved bordet«
Når de begynder at putte det ind i dine pensioner, er det, hvad de kalder de sukkere ved bordet. De tjener penge på de lån, de har givet – betydelige penge på grund af de meget høje renter. Og når de har taget en bunke af renten, vil de gerne dumpe lånene over på jer, så når de eksploderer, siger de: »Åh, nå, jeg gætter på, at de eksploderede i jeres pensionsfonde og ikke på vores budget. Så sørgelig dag for dig.« Det er enormt problematisk.
Ligesom i subprime-lånekatastrofen er pistolen overgearingen, og aftrækkeren kan være hvad som helst. Dengang bundtede de realkreditlånene og gambled på dem. Her kan det være datacentrene eller AI-virksomhederne selv. Når de går under, udløser de et sæt dominoer. Det er aftrækkeren. Men pistolen er den absurde mængde gæld, de har påtaget sig – meget spekulative finansielle risici.
Grunden til, at det er meget spekulative, er, at de siger: »Åh, nej, det er okay. De kan låne 800 milliarder dollars, for AI kommer til at styre verden.« Hvad hvis AI ikke styrer verden – eller det tager 20 år? Først og fremmest ønsker jeg ikke, at den skal styre verden. Det gør de fleste heller ikke. Men selv hvis de har ret forretningsmæssigt: hvad nu hvis det tager for lang tid, og du sidder med høje renter?
Taj Mahal-lessonen: Høj rente dræber selv vindere
Vi talte om dette før i forbindelse med Trumps ejendomme. Grunden til, at han gik igennem en spektakulær serie af konkurser i Atlantic City, var, at hans tredje og sidste casino – Taj Mahal – havde fået renter på omkring 14 % på de penge, de brugte til at købe og bygge Taj Mahal. Alle på det tidspunkt troede – og havde ret – at det er en alt for høj rente, fuldstændig uholdbar. »Huset vinder altid« – medmindre det drives af Donald Trump, som tog et lån med 14 % i rente.
Da Taj Mahal kollapsede, gjorde de to andre casinoer det også. Det skabte en dominoeffekt. Høje renter er en super risikabel måde at bygge en virksomhed på. Og hvis dine profiler ikke kommer – glem helt at de slet ikke kommer – hvis de kommer for sent, begynder renterne at hobe sig op så meget, at du umuligt kan betale dem. Kun på timingen, for slet ikke at tale om substansen i dit produkt.
Vi er i en verden af problemer. De kan ikke betale de renter hurtigt nok givet det, der sker på markedet. Coreweave er allerede en distressed ejendom. Når du er en distressed ejendom, er du få centimeter fra at gå under. Og så kommer du ind i en ond cirkel.
Dominoeffekten: Når virksomhederne har investeret i hinanden
Når én af dem falder, har andre virksomheder investeret i hinanden. AI-virksomhederne har investeret i hinanden. Nvidia har investeret i AI-virksomhederne og neocloud-virksomhederne. Så får de det som et problematisk aktiv i deres bøger. De bliver nødt til at nedskrive det. Deres indtjegninger lider, og så straffer markedet dem. Og så går kapløbet i gang. Dominoerne falder alle sammen. Boblen brister, og vi er i massive problemer.
Bankernes engagement: 450 milliarder dollars i åbne forpligtelser
Hvad sker der, når bankerne beder om redningspakker, som de gjorde i den globale finanskrise i 2008? Bankerne havde 450 milliarder dollars i åbne forpligtelser over for AI-relaterede låntagere – primært kreditlinjer – ved udgangen af 2025, hvoraf 150 milliarder var udtrukket.
Det store spørgsmål: Kommer markedet til at tage alt denne indlysende urimelige risiko alvorligt før eller efter Anthropics forestående børsnotering? Anthropics omsætning var 16,3 milliarder dollars i første halvdel af året. Det er imponerende. Men der er en hage: Anthropics nettounderskud var 80,7 milliarder dollars, primært drevet af en engangsudgift. Mest bemærkelsesværdigt har virksomheden forpligtet sig til op til 87 milliarder dollars i kontrakter på datacentre, chips og beregningskraft – mange af hvilke ikke kan annulleres over de næste 10 år.
Så igen: de hentede 16,3 milliarder dollars i omsætning, men tabte 80,7 milliarder – i første halvdel af året.
Anthropic: 16 milliarder i omsætning, 80 milliarder i tab
De håber at rejse hundrede milliarder i deres børsnotering, men analytikere siger allerede nu, at det ikke vil være nok til at betjene de 87 milliarder i kontrakter. De får brug for mere straks bagefter. Men det hele er spekulativt. Vi har ingen idé om, hvorvidt de 87 milliarder i kontrakter kommer til at virke for Anthropic, og hvor lang tid der går. Dette er en episk katastrofe.
Hvordan kunne disse mennesker ikke se det? Det er vanvid. Men på dette tidspunkt ser de det faktisk. Det er derfor, Elon giver den advarsel. Det er derfor, Elon efter min mening forsøger at fusionere alle sine virksomheder, så han kan sige, at den er »for stor til at fejle«. Det er derfor, OpenAI siger, at dette grundlæggende er et offentligt gode, så regeringen burde være forsikrer i sidste instans – hvilket betyder, at når vi går under, bliver I nødt til at redde os.
Og hvis du ikke kan redde dem med hundrede milliarder – hvilket er en spektakulær børsnotering – hvor meget bliver der så brug for? Det er fuldstændig uholdbart. Og kontrakterne kan ikke annulleres? Det er en ekstra byrde på balancesedlen – et enormt finansiel problem. Hvis du ikke kan annullere de kontrakter, skylder du de penge uanset hvad. Selv en let forkert drej, og al den gæld og alle de forpligtelser i bøgerne kommer til at trække virksomheden ned. Der er næsten ingen tvivl om det.
Hvorfor det trækker hele markedet ned
En virksomhed går under – så hvad? Det er kapitalisme, ikke? Nej, der er to kæmpe problemer. Nummer ét: næsten al væksten på markederne kommer fra bare et par virksomheder. 34 % af væksten kommer kun fra to AI-relaterede virksomheder. Hvis de to får problemer, trækker det hele markedet ned, hvilket skaber endnu en ond cirkel.
Den vigtigste onde cirkel af alle er det samme, der skabte panik i 2008: Bankerne er gearet op til attenhjulingerne, og som Musk sagde, giver de ikke mere, fordi de nu indser, at deres gearing til disse virksomheder er farlig.
Hvis du låner en lille smule fra en bank, ejer banken dig. Hvis du låner meget fra en bank, ejer du banken. Det er den gamle model, Trump brugte til ikke at betale sin gæld. Men bankerne giver alle kortfristede lån til alle de andre virksomheder. Så når bankerne går under, fordi gælden ikke bliver tilbagebetalt, har alle andre virksomheder slet ingen penge til at drive deres forretninger – og så smelter hele økonomien sammen.
Konklusion: Spænd sikkerhedsselen
Vi er faretruende tæt på det. Næsten sikkert kommer det til at ske. Så spænd sikkerhedsselen og gør dig klar til nedslag. Dette kommer til at blive et gigantisk crash.
Hver gang du ringer med klokken herunder, får en engel sine vinger. Helt forkert, men det holder dig opdateret på vores live-shows.
original transcript
Elon Musk just admitted that the AI data center buildout is a giant bubble. Uh-oh. That's a huge problem. And then on top of that, apparently some financial institutions have stopped giving out loans to AI companies. >> Smart. >> That means the bubble is inches away from bursting. But we have even more depressing numbers for you guys because once you see the size of the danger here in this AI bubble. H it is super scary and you got to take precautions. Anna's got the whole story. >> First of all, big shout out to Taylor Ellis who produced this segment and made it easy to understand for those who aren't really in the tech world. But let me tell you what Elon Musk had to say specifically. Just yesterday he wrote this on X. As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn't have substantial revenue independent of AI data centers. I mean, that should have been the case all along, but okay. He continues, SpaceX closed $40 billion of debt for graphics processing units, uh, but were investment grade and have Starlink and launch revenue. That means the 200 plus neoclouds are going to see guppy level death and consolidation. Some will make it through but not many. Okay, so you might be thinking what the hell is he talking about? Let me break it down to make it easy to understand. So first what the hell is a neocloud? You might be wondering, well, a Neocloud is a company that essentially builds the computing structure for AI companies. And Coreeave is just one example of a Neocloud company. And I want you to remember the name of that company, Neocloud, because it's going to come back up later in this segment. Uh, now the Neocloud companies are taking on debt to build out their infrastructure. But according to Musk, debt markets are no longer interested in um providing the loans to these companies. They no longer agree with or are interested in this business model. And while Meta or Alphabet, which are building out their own data centers, offer other products that provide a revenue stream for their companies, you know, like Facebook and Google. other Neocloud companies might not really have the same setup or the same type of revenue stream from other diversified aspects of their business model. And by the way, we should also note that there is um a pretty powerful figure involved in all of this who does not want to see the Neocloud buildout go bust. And that person is of course Jensen Wong, the CEO of Nvidia. Nvidia has invested into the Neocloud companies and the Neocloud facilities use Nvidia's products. Okay, so there's a bit of a mutual benefit type situation going on here. But here's the catch. Even the CEO of Nvidia has admitted that there's not enough money to loan to these NeoCloud companies. And so in an interview with CNBC, Nvidia CEO Jensen Wong said this. The demand for neoclouds is off the charts. What is missing is the working capital to fuel that growth because there's just so much demand across so many sectors of artificial intelligence. There's a scarcity of capital. The demand of the capital is much greater than the availability of the capital that's going to get solved. He thinks it's going to get solved. Only time will tell if he's correct about that. But um let's circle back to what I mentioned earlier. Coreweave, right? one of the neocloud companies I was referring to. Well, Coreweave is a good example of how um well, the lenders are no longer really interested in, you know, providing the the loans, the leverage necessary for these buildouts. Uh Coree saw it shares drop 41% since October 6th as of last Friday's close. Shares of other NEOClouds have also posted double-digit declines during that time. So Cororeweave needs to issue bonds kind of like the US government does right with our treasuries. Uh Cororeweave needs to issue bonds to basically finance their computing buildout. But there's a problem with that. So earlier this month the price of their bonds fell so much and inversely their yields spiked so much that their bonds are classified as distressed. So going the bonds route is not really working out for core either. Analyst Gil Laura stated that borrowing at very high interest rates in order to build data centers that are increasingly commoditized outside of the hyperscalers may not end up being a good investment. The more leverage we use to build data centers, the more risk we're bringing into the system. And I can't emphasize enough how important that point is because remember these overleveraged companies are being securitized, right? And so we'll get to that in just a minute, but let's look at this graphic first. So according to Americans for Financial uh reform, just nine companies, guys, that's Alphabet, Amazon, AMD, Broadcom, Meta, Microsoft, Nvidia, Oracle, and SpaceX have a whopping 3.7 trillion with a t trillion dollars in debt. And remember how Nvidia is an investor in um Cororeweave? Let's talk about that just briefly. Analysts expect companies in the S&P 500 to report a 30% jump in third quarter earnings from a year ago, but just two companies, Nvidia and Micron, account for 34% of that growth according to a Dow Jones market data analysis. And if you're ready for things to get even riskier, I'm not. Doesn't matter. It's coming anyway. A new report from Americans for Financial Reform indicates the following. Speculative data center debt is being repackaged and sold to pension and retirement funds, university endowments, and other institutional investors in a manner eerily, eerily reminiscent of the subprime bubble. Now, the subprime bubble had to do with these predatory loans for, you know, for mortgages. Basically, in this case, we're talking about overleveraged AI companies and um they're very likely going to fail and they're involved in your in your portfolio, your retirement portfolio, your pension portfolio. I mean, this is a disaster of epic proportions. Back to you, Jen. >> So, we've got more disastrous numbers in a second, believe it or not, but let me go through the ones that Anna mentioned. So guys, once [clears throat] they start putting it into your pensions, that's what they call the suckers at the table. So they make money from the loans that they've given. And they've made substantial money because of the very high interest rates that Anna mentioned earlier, [cough] excuse me. Um, and once they've taken a bunch of the interest, they'd love to dump those loans onto you guys so that when they blow up, they go, "Oh, well, golly gee, I guess they blew up in your pension funds and not on our, you know, uh, budget. So, sad day for you." And, um, and so that's hugely problematic. And we told you earlier too that just like in the subprime loan disaster the gun is the overleveraging the trigger could be any given thing. In that case they were bundling the home mortgages and then gambling on them. In this case it might be the data centers or the AI companies themselves. Once they go under they trigger a set of dominoes. That's the trigger. But the gun is this absurd amount of debt that they have taken on very speculative uh financial risks. Now, the reason I say it's very speculative is because they're like, "Oh, no, it's okay. They could borrow $800 billion because uh AI is going to control the world." What if AI doesn't control the world or it takes 20 years for it to control the world? First of all, I don't want it to control the world. Neither do most Americans or or world citizens. But even if they're right from a business perspective, what if it takes too long and you're holding high interest rates? So, uh you know, we talked to you about this before in the context of Trump's properties. The reason that he went into spectacular series of bankruptcies in Atlantic City is because his third and last casino, the the Taj Mahal, had gotten interest rates at around 14% on the money that they used to buy and build the Taj Mahal. And everybody at the time thought, and they were correct, that that is way too high an interest rate, totally unsustainable. you know, the the house always wins asterisk unless it's run by Donald Trump who took out a 14% interest rate. So once the Taj Mahal collapsed, so did the other two casinos. It created a domino effect. [snorts] So these high interest rates are super super risky way to build a business. And if you're uh profits don't come, forget not coming in at all. If they come in too late, the interest starts to pile up so much you can't possibly service it. just on the timing issues, let alone the substance of your product. So, we're in a world of trouble here, guys. They cannot pay those interest rates quickly enough given what's happening in this market. And you could tell because corre is already a distressed property. Once you're distressed property, you're inches away from going under. And so, and you get into a vicious cycle. So, and remember once one of these fall, other companies have invested into one another. The AI companies invested into one another. Nvidia invested into the AI companies and the Neocloud companies. And then they're going to have that as a problematic asset asset in their books. They're going to have to write it down. Their earnings will suffer and then the market will punish them. Oh, and off we go to the races. The dominoes all fall. This bubble's going to burst and we are in massive trouble and we're going to show you more terrible numbers in a second. But Anna, what were you gonna say? >> More terrible numbers. I mean, who cares what I have to say? Like the facts are the facts and you guys need to know how disastrous these numbers are. So, I want to talk about the banks uh which of course provided loans to these AI companies. What how much money are we talking about here? Because remember, you and I have been talking about this a lot on the show. uh what happens when the banks ask for bailouts like they did in the global financial crisis of 2008 because banks also had $450 billion in open commitments to AI linked borrowers primarily lines of credit as of the end of last year 2025 of which 150 billion was outstanding and here's the big question guys this is the real big question here is the market going to get serious about all of this obvious unreasonable risk before or after Anthropic's impending IPO because look, we've talked about this before. Anthropics revenue was $16.3 billion in the first half of the year. That's impressive. That's impressive, right? But there's that catch. What am I talking about? Anthropic's net loss was $80.7 billion driven largely by a one-time charge. Stunningly, the company has committed up to $87 billion in contracts for data centers, chips, and computing power, many of which cannot be cancelled over the next 10 years. So again, they brought in $16.3 billion in revenue, but lost 80.7 billion >> in the first half of the year. >> In the first half of the year, yes. >> Okay. So these losses as far as the eye can see, how are you going to get return on investment on your massive interest rates at least for the Neocloud companies? And so let let me uh flesh out some of the other problems here. They're hoping to raise hundred billion in their IPO when they launch in the stock market, but already analysts are saying that won't be enough to service the $87 billion in contracts. They're going to need more immediately afterwards. But this is all speculative. We have no idea if the $87 billion in contracts are going to work out for Anthropic and how long before they work out. So, no. No. This is an epic disaster. How could these people not see it? It's insanity. And I think at this point, Anna, they do see it. And that that's why Elon is giving that warning. That's why Elon I in my opinion is trying to merge all of his companies so he can say it's too big to fail. That's why Open AI is saying well you know this is basically a public good. So you know the government should be an insurer of last resort meaning when we go under you're going to need to bail us out. And so now to the point of needing to bail them out if you can't do it with hundred billion which is a spectacular IPO when you launch how much is it going to need? No, this is totally unsustainable. And did you hear that part where the contracts are not cancelellable? Oh my god, that's an extra burden on your balance sheet. That is an enormous financial problem. If you can't cancel those contracts, you owe that money no matter what. No matter what happens to you in business. So, even a slightly wrong turn and you've got all of this debt and all of these obligations on your books, they're going to drag the company down. There's almost no question about it. Then, okay, a company goes under, you can say, so what? That's capitalism, right? No, guys, there's two giant problems. Number one, almost all the growth in the markets are from just a couple of companies. As Anna pointed out there, 34% of the growth is just from two companies related to AI. So, if those two companies have problems, that drags down the entire market, which then creates another vicious cycle. Uh but the most important vicious cycle of all is the same uh thing that made everyone panic so much about in 2008 and they never fully really clarified that for you guys. Uh so I'm not sure you understand why the panic will be so large from the people in power once this bubble bursts because the banks are all you know levered up to the gears and as uh as Musk said they're not giving anymore because they realize now they've uh they their leverage to these companies are are dangerous. If you borrow a little bit from a bank the bank owns you. If you borrow a lot from a bank you own the bank. Okay? And so that's the old model that Trump used to use for not paying his debts. But the banks all give shortterm loans to all the other corporations. So once the banks go under because of all this debt is not going to get repaid, then all the other corporations have no money at all to operate their businesses and that's when the entire economy melts down. And we are perilously close to that. almost certain to happen. So, buckle up and brace for impact. This is going to be a gigantic crash. Every time you ring the bell below, an angel gets its wings. Totally not true, but it does keep you updated on our live shows.