


They have unlimited resources and can keep this floating longer than any of us can imagine.

Softbank doesn’t.

They’ve been failing at literally every step in AI coverage, especially at just parroting whatever the fuck anthropic and openai claim to be their newest breakthrough without questioning anything. Failing basic math is just icing on the cake.
What’s fun is now you can stick it to a bank easily by taking out all your money and investing it yourself. Make money off your own money. Cuz they sure do when they have it. (Hence the wait time when withdrawing large amounts)
Lots of trade apps out there now. You only need banks for holding small transaction amounts for paying bill. Soon they won’t even be required for that.
Like the article says, their internal risk management models are telling them to get the risk off their books.
Taking on the debt in the first place is like agreeing to store a gallon of milk in a fridge for an undetermined amount of time. The milk is going to go bad, and when it does the jug (or bag) might explode forcing a cleanout of the fridge, throwing out other things in the fridge, and maybe replacing the fridge altogether if the milk leaked down into a lot of crevices. Everyone knows the milk will go bad, but the big banks will know how old the fridge is, what coolant it uses, how old the compressor is, how many times a day the fridge has been opened, the average time it stays open, the dairy the cow who produced the milk, and the name of the cow if they can get it. The milk is going bad quicker than expected, and they’re finding out they don’t want to be the ones to guess when the jug will explode, and they’re also finding out that nobody else wants to buy the jug.
Oracle pulling a “god did it” was probably a part of this, but nobody involved is going to give a real answer.
Uh brother! (or sister on non binary sibling).
You just sell the entire fridge a sub prime fridge and watch the number go up.
No they can’t do it forever. Exponentially growing demands outrun everything eventually. The onl, way out of this are either controlled deflation of the bubble or a hard crash. The former is possible but unlikely.
“Markets can remain irrational longer than you can remain solvent.” - John Maynard Keynes
But fact of the matter is that the people who run the market are the ones that decide who wins and loses. And they never lose.
Is this similar to the Dot.Com bubble burst?

I suspect this might be worse due to the amount of money being thrown around. As much as I want the companies responsible for the stupid bubble to go bankrupt, there’s a lot of money from pension and investment funds, so that’s going to hurt a lot of people who had nothing to do with that.
My youtube algorthim is filled with AI-skeptics (Eli the Computer guy, Ed Zitron, Prof G Markets) so this has been the slowest car crash I’ve ever gotten to watch happen. If Donald had actually focused on inflation all of these debt-sensitive AI plays could’ve continued to whistle through the graveyard. The skeptics have called out coreweave as being potentially the first domino to topple and their 6 year debt is trading at 15% right now. Thanks to plummeting token costs, they need to double their token sales every 90 days.

My youtube algorithm
Why are you deliberately living in a bubble that shows you what you want to hear? Like I have a cousin convinced for years that either the US or China is >6 months from total economic collapse, because that’s their algo is showing them reinforcing narratives they want to hear.
Using youtube logged out gives you the general algorithm filled with crap I would never consider watching, at least mine’s curated. It gave me a little too much hope in Kamala but it’s not like I fought any less hard for her. I aggressively cull my feed and it’s pretty much just the creators I subscribe to. If I want “discovery” I filter reddit by youtu.be or youtube.com.
Stop using YouTube. The algorithm is not good for your mental health, nor to learn anything of substance.
I watch more perfect union and the good liars. They’re not good at pointing out major issues? They bring receipts.
It’s true, disable YouTube feed and only look at subscription.
You can do it at myactivity.google.com/myactivity

I have to second that, yeah, YouTube’s Algo is insidious and drives my binge-watching, but I’ve been able to curate a subscription list of solid content creators.
The algorithm is dog ass. YouTube is still full of plenty smart people with something to say, though. It’s a video sharing platform; that doesn’t inherently mean every content creator who uses it is also dog ass.
The part nobody prices in is the stranded-asset scenario. These underwritings assume a demand curve that keeps rising for a decade. If inference efficiency improves the way training efficiency did — and there’s real pressure in that direction now — a lot of this capacity gets repriced hard.
The bank behavior is the tell: debt that gets packaged and sold on right after underwriting is the classic ‘we have concerns but we’re not holding the bag’ move.
I would say there are definite advances for inference. Custom backends built by AI for specific models, such as Strata, makes efficient leveraging of hardware much easier. Engrams, such as those in Qwen Next-Flash, allows for using more parameters via SSD, rather than memory alone.
These and other things over the past few years have allowed me to use bigger models, faster. Given another year of such progress, I think the value of subscribing to big American companies would start drying up.

Sorry, ignorant question from someone who didn’t understand much of your comment - do you mean that AI models are writing code that improves them? Cos that would be interesting.

Banks are in the business of underwriting debt and reselling it to private investors.
We’ve been reading this “End is Near” shit since COVID. Put something more on the table.
I’ve been reading, “the end is near,” since HIV/Aids, y2k, 9/11, the SARS outbreak, the anthrax scare, the swine flu/h1n1 and the several evils outbreaks. Oh yeah all those wmd scares and the potential for a world War after 9/11.
The end is always fuckin’ near
Bad tech rapidly became a central pillar to that buying/selling and everyone is realizing it may be worthless.

everyone is realizing it may be worthless
Idk who you think “everyone” is
Owning a debt from the AI industry is a fairly insane amount of risk

Is it? They’ve been outperforming every other debt class for five years running.

So how does the government bail them out without triggering it’s own treasury yield default?
Quantitative easing.
The same shit that put us right here after the GFC

Ooo I know this one! I won’t give out any spoilers.
Apparently they’ve spent 10 trillion in commitments that soan to 2032… And have to get AI earning to 3 trillion a year to avoid a crash.
…and there’s 8 billion people on earth. So I assume they’re hoping for some lucrative corporate/government contracts and bailouts.