Anthropic, which officially filed to go public at a rumored valuation of up to $2 trillion dollars just a week before the elections.
As Bloomberg reports, those prospects are in free fall as the chaos takes hold. Some investors are now weighing bids that would value Anthropic as low as $1.1 trillion. While that’s still a huge valuation for a company that lost over $80 billion in the first half of 2026 alone, it’s a massive drop compared to the previously unbridled enthusiasm surrounding the firm’s IPO.
You must log in or register to comment.
I hope this lack of confidence in AI spreads to openAI too.
It only needs one domino to fall. OpenAI will be toast when Anthropic’s IPO fails.
If it crashes before they even get to IPO, who are the investors I get to point and laugh at?
It’s pretty much all your least-favorite financial entities. You’ve got the megacorp tech firms (Google, Amazon, et al), venture capitol investment institutions (Goldman Sacs, etc), and even some sovereign wealth funds (Qatar, Saudi Arabia primarily).
It’s a real who’s who of institutions that should fuck off and die.
Um, good? The IPO was way overvalued and filing at this time was seen as a desperation move, much like the way Anthropic is trying really hard now to create an agentic model that can commit major crimes and cause large industrial disasters (which it will do by command, maybe by a poorly-written prompt, much less, gaining self awareness and escaping containment permanently).
Anthropic is behaving conspicuously like Enron did before it went bankrupt: It’s getting more and more desperate as it borrows more and more, and continues to hemorrhage funds. It’s now resorting to shenanigans to keep investor confidence up, and those shenanigans will get more and more extreme (and more and more dangerous) until that investor confidence bottoms out and everyone sells.
If we look at Elon Musk’s holdings (Musk is a compulsory liar) we see that investors are gullible, though maybe they’re just gambling that other investors are more gullible than they. That’s always a high-risk wager.
You think this is bad, the guys at OpenAI probably realized they are cooked! IPO is a way to cash out and flee to Argentina, instead they might walk face first into a combine harvester that is the remnants of the AI market. Poor Atlman can sell some RAM.
Have they tried the classical method of adding up what they actually own and what they owe to others, and then factor in what they have earned in the last ten years? That would give a clear and fair evaluation of the company for an IPO.
That’s not how any IPO works.
There was, once upon a time, some limit to the market valuation of a company that approximated the most money they could earn if they came to dominate their entire market, times the inverse of the rate of return for highly secure (like AAA) bonds.
Now speculation assumes that companies will grow to exceed the size of their whole industry, perhaps in the near future, and returns from highly secure investments were really low post covid (resulting in pretty much a 40x valuation multiplier), which has just… held on since then.
It’s weird, and I don’t like it. Why should a company ever be worth more than all the money it could ever make?
Not Spacex, the company that is 80% an ai company?
What?
On Thursday, a massive sell off wiped out a record $2.3 trillion in market value, highlighting major volatility and uncertainty.
Oh no, they lost a bunch of pretend speculation value that wasn’t based on reality!
Fuck the stock market.
You’re going to find out just how real stock market money is when the AI bubble implodes.
Buffett Index: total stock valuation vs. GDP
1929: 130%
2007: 130%
2025: 220%
Oh no, they lost a bunch of pretend speculation value that wasn’t based on reality!
The problem is that the pretended value will likely be paid from people’s pensions savings.
Don’t invest in AI companies for your pension then?
don’t think you fully understand how retirement savings work…
Or I live in a different country with different pensions. Here the government pays for it from general taxation and its pretty generous.
Honestly could argue its too much, if you have used your working years to buy a house you are probably going to be better off than a minimum wage worker. Calculated it based on my mortgage rate, minimum wage and pension a couple of years ago anyway.
Even if you have that choice, it’ll only stop some of the damage. The fact is that every sector will feel this collapse. It won’t only effect AI companies.
People whose pension savings were based on pretend speculation value will have less pretend speculation value.
$3.50, final offer. You could bribe me with a good sandwich though too i suppose.
$3.50? Done!
Congratulations, you now owe $80 billion.
But it has to be like a reeeeaaally good sandwich
It is the BROODWICH.
Who ate the sun-dried tomato!?
Voice: “…they’re good!”
Sandwich-heavy portfolio made the hungry investor happy.
Ooh, it must’ve been about seven, eight years ago. Me and the little lady was out on this boat, you see, all alone at night, when all of a sudden this huge creature, this giant crustacean from the paleolithic era, comes out of the water and I yelled. I said, “What do you want from us, monster?!” And the monster bent down and said, “…Uh I need about tree-fitty.”
You give them a dollar they’re gonna assume you got more!
they do own some gpu’s. I bet you could hack one to run a server for somethingm
As “low” as 1.1T
I guess thats priming in effect? My guess is original target was 1T, so they primed public with 2T…
If they are 80 billion in the hole shouldn’t the valuation be negative 80 billion?
On one hand you need to consider future growth so the value is high because growth is high.
On the other hand the path to profitability is going to be brutal as they need to compete with Chinese models which are often on par or sometimes better (kling v suno) than the yank models. Often the Chinese models are open source too so that helps consumers, competitors and self hosters.
On the other other hand the us government could just outlaw competition in the us market and create a regional monopoly/ oligopoly for whichever companies it wants.
On the other other other hand (I am an octopus, please don’t turn me into calamari, I have a family and friends) early movers in new industries are rarely the ones who make it big. If you invested in MySpace or AOL having seen a vision from the future where social media took over the internet you would be both correct and broke.
TLDR: traditional financial modeling has been thrown out the window years ago for tech companies. SpaceX is basically guaranteed to lose money for IPO buyers if you account for the time value of money. https://futurism.com/space/math-spacex-ipo-virtually-impossible
If they stopped now yeah, but those are rookie numbers. They’re gonna evaporate way more money than that before they finally crash and burn.
No, you have to consider future valuation. They just need to run a loss for as long as it takes to eliminate their competition (humans maintaining independent thought and learning capability) at which point they can extract whatever price they want.
All true! But they’ll still need customers willing to pay that price, and that’s not lookin’ so good at the moment.
Only if you think they are going to stop all operations right this second and give away all their assets.
However, if you think they are going to make enough money to pay off the debt (or at least make it seem like they can) then their valuation is proportionate to what happens if things go well VS what happens if things go poorly.
I don’t think they’re that clever, I think they missed the wave.
Anthropic is worth ~$150billion.
That’s how much they hold in tangible assets.
Minus debt…
Classical method: what do they own, what do they owe, and how much did they earn in the last ten years. Then you have a realistic base number for an IPO.
So negative 50 billion overall then.
I think it’s negative 80 billion, lmao
I know they paid a lot for it but I wonder how much of that infrastructure is really worth if no one wants to buy it
when they go under I’ll pay 25¢ per GB on DDR4 ram and 35¢ per GB on DDR5.
hell, I’ll even pay up to $3.50 for some of those RTX thingamajigs they got.
Keep in mind that most of the hardware will be obsolete in 3-5. That’s a quickly depreciating asset.
Also, a lot of the hardware they built is specialized for AI server. None of it will be useful to consumers, so they can’t easily pawn it off
Part of my job takes me to data centers, including ones with GPU clusters entirely for LLM workloads, and that isn’t strictly true.
There’s so much hardware sitting around, unused: system RAM, SSDs, HDDs, CPUs, etc that could easily go in consumer hardware, plus some GPUs still have VRAM. Plenty for most gamers. Then there’s always the local LLM scene, and I’m sure there are all other kinds of neat computation you could do with these.
I foresee a big microsoldering scene keeping these babies going long after they’re “obsolete.” Graphics peaked when I was in my 20s anyway, so it’s not like we need to go any more hifi. ;p
Edit: I did some more reading and… I might be wrong on a lot of this. Well at least the storage will hopefully be cheap again! And I’m still hoping for adaptations made for consumer use once everyone stops using SaaS. Any day now…
My son is already drooling at the prospect of getting his hands on some of their RAM.
Hate to break it to you (or your son), but the kind pf RAM theyve got in that datacenter is not compatible with your son’s computer.
It’s essentially worthless for any purpose other than the type of GPUs they put in a LLM data center. That’s part of why their business is so dumb.
The reason RAM prices have skyrocketed is because the limited number of RAM fabrication facilities have mostly converted their setup to produce the type of RAM that only works for slop instead of the consumer grade stuff.
So your son and I are both gonna need to wait for those RAM fabrication facilities to realize that the people ordering years worth of their work don’t actually have the money to pay for their order. At that point they will likely convert the fabrication facilities back to normal RAM production, and a couple of years after thar happens RAM prices will begin to approach sensible levels again.
I am very excited to buy some of the large HDDs currently being hoarded by the slop mongers when they soon eat shit tho… I’ma build me a fat RAID.
Also, the system designs in general expect more power than a residential circuit can provide, some more than an EVSE circuit. So folks imagining buying a system that does have the memory are also going to be disappointed.
The storage is the most likely to come over, with the caveat of some awkward EDSSF form factor, but enclosures I could see becoming available with a big enough opportunity. E1.L might be the trickiest, but those aren’t too popular in the datacenters either.
I love how you can go to IPO showing financial proof that your business is utter unsustainable and even if that changes would take considerable time to dig out of…
There’s only one appropriate response.
:)
It’s not cratering, the value is still a positive number. Call me when the IPO is valued at something with a minus sign before.
People should get paid to take part of this IPO, not the other way.
