Leaked IPO docs: Anthropic tempts investors with existential risk warning

After an unusual pitch to investors, Anthropic has laid bare its challenges in becoming anything like a profitable, normal business in its IPO prospectus that was leaked to the FT on Tuesday. Posing an “existential risk to humanity” was among the messages reportedly intended to allure investors, as if to say, “We could make you …

[Tron]

AI is not too big to fail.

Most of the investment is circular, and most of the projects are 'ahead'.

AI functions in a bubble of wealthy folk, who are all wealthy enough to survive the bubble bursting.

Without AI we would be in a bog standard recession caused by poor government choices globally. And that is what we would flip back to.

Better political decisions would see the global economy pick up again. And the rich folks would soon find a new scam to artificially boost markets. Probably with fewer dead people than Trump's on/off bombing of Iran, used as a joystick for oil futures to benefit his buddies.

xPUs have the life expectancy of a hamster, so you are not really investing, you are subscribing. Which is most definitely not the same thing.

The traditional post IPO sell off may well be the start of the bubble bursting. When the value maxes, everyone may sell and consider themselves well out of it.

It may be worth getting out, because the legal and financial culpability for what this junk does will eventually land at the vendor's door. Maybe not until Trump is out of power, but if what we've seen is a taste of what is to come, the compensation demands will be astronomical.

[1752]

Too big to save

Take into account the derivatives market (there are different estimates but 'commonly estimated in the hundreds of trillions of U.S. dollars globally.') then what I would say if bubble starts the dominos toppling when it goes, too big to save. That being the global economy.

[T. F. M. Reader]

Re: Too big to save

There is probably an unknown amount of CDS (Credit Default Swaps) and similar instruments around the bubble. As on earlier occasions, it may not be easy to find out who is exposed. That's in support of the derivatives bit above.

What quite worries me is that private equity shops that hold a lot of AI debt also own insurance companies. If bad debt is opaquely offloaded to said insurance companies and the bubble goes pop we'll all find out that insurance companies don't go bankrupt. There is secondary and tertiary insurance and, ultimately, government guarantees. This is good for all the policy holders, but may be really bad for taxpayers, if governments have to make good on those guarantees.

[jdiebdhidbsusbvwbsidnsoskebid]

Re: Too big to save

Is this going to be bad for real people though?

When the dot-com bubble burst, lots of startup companies (even quite big ones) collapsed and a lot of people lost their jobs. In 2008, some banks collapsed and some people lost their jobs, but the worse thing was that lots of people couldn't afford their homes anymore.

If the AI bubble bursts, yeah a lot of investors' risks are going to come home to roost, but will it affect real people very much? I guess anyone working in AI infrastructure night be affected (data centre construction and running etc.)

[Herring`]

Re: Too big to save

It always affects the real people most. The people at the top will have squirreled away enough cash to buy the dip and end up even richer. Sure, a lot of this is funded by private capital, but a lot of that is borrowed from actual banks. In every crash, the people who caused it walk away with a fucktonne of money while the general public gets austerity.

Have a nice day.

[Snake]

Re: austerity

Ain't that the truth. It is so entombed in the social construct of "capitalism" that even the World Bank, if asked to assist, 'favors' you with the premise.

"We need help"

"We can lend you money to get yourself back on your feet. All we expect is "austerity", but not for the capitalists, they should expect healthy returns and favorable interest rates."

"We can't do that, what about the people?"

...

"Markets help us decide our policies. I'm sure you understand. Please keep your trades high and your public help low, we'd be happy to help after you do that."

[Anonymous Coward]

Re: Too big to save

There's also the fact that abuse gets covered up.

The whole Credit Swap (CDO) affair resulted in effectively one person ending in jail, and he was just a small fish effectively doing what the big fish got away with. And we paid these people via government handouts because otherwise the whole artifice would have taken down more.

[cmdrklarg]

Re: AI is not too big to fail.

**** Probably with fewer dead people than Trump's on/off bombing of Iran, used as a joystick for oil futures to benefit his buddies.

The Orange God Emperor does not have "buddies". There is himself, and then there are others (everyone else) who are not important to him. Any "friendly" interaction with him is a transaction, nothing more. And like any interaction involving him, you can't rely on the OGE to follow through on his promises, especially when it has to do with him giving up part of his money stash.

His "buddies" only benefit because the same things that benefits the OGE benefit them.

[DS999]

Re: AI is not too big to fail.

who are all wealthy enough to survive the bubble bursting

Doesn't matter, they are the ones who always get bailed out when anything goes wrong.

I don't think it happens this time - not even if the real economy is suffering - because the massive backlash against AI datacenters pretty much guarantees any politician who supported a bill to bailout AI datacenter owners would get voted out if not lynched.

Fortunately most of the money actually going into datacenter construction is coming out of cash flow so the impact on at least Microsoft, Google, Amazon and Meta would be minimal, other than a big haircut on stock price. The only big company it would kill is Oracle (which would concern almost no one) though SpaceX would be severely wounded. Whoever is loaning OpenAI cash would be hurt, but they can afford it. And reportedly a lot of that money is coming from the Middle East so as with Oracle no one will care that they get hurt. They certainly won't deserve taxpayer funded bailouts!

[Anonymous Coward]

Re: AI is not too big to fail.

Root Cause of Force Majeure.

See Fallout, Battlestar Galactica, Star Trek Picard, Foundation, Dune etc ….

Problem solved.

[Anonymous Coward]

$2 trillion value on revenue of $4.6 billion with losses of >$8 billion.

Sorry but did I miss the punchline on that comedic valuation? That's higher than the GDP of South Korea.

Then it spends $518 billion on compute etc... which puts money into Nvidia etc... bumping up their price who then put money back in bumping up the price again. Till pop. It's like a Ponzi scheme. I don't know enough about stocks and shares to be fair but this seems like madness.

[Philo T Farnsworth]

> It's like a Ponzi scheme.

No like about it. The similarity is known in mathematical circles as identity.

Though an acquaintence of mine claims it actually an "Iznop scheme," which is like a Ponzi scheme except nobody gets paid.

Pick yer poison.

[powershift]

err

Iznop is ponzi spelled backwards.

[Philo T Farnsworth]

Re: err

Uh, yeah.

[Joe User]

Re: err

"Iznop is ponzi spelled backwards."

That whooshing sound is the point going over your head.

[Steve K]

Jensen gets paid

Jensen gets paid….

[Blazde]

The madness isn't even really about AI, it pre-dates it.

Tesla still has an insane valuation for a modest-sized auto-mobile maker.

Meta has an insane valuation considering it's enormous profitability yet almost complete aversion to paying dividends due to a hostile share structure that allows Zuckerberg to piss away dozens (now probably hundreds) of billions on pet projects without challenge.

Everyone thought Elon messed up buying Twitter for so much but hah you reckoned without him just buying himself out of that mess using a space company funded by US tax-payers, hype, and a similarly lopsided share structure that probably isn't even needed given the size of his personality cult. Buying himself out and still being able to control a platform most of the world's media fawn over, natch.

Welcome to the world of celebrity corporatism, the best reality show ever conceived. Pull up a seat, pick your hero, shout 'take my money' and watch them take it. This season's big story-line is a classic will they or won't they (Humanity. Survive). Suspend your disbelief and join the fun today.

Personally I'm 100% #TeamJensenHuang. You can't argue with that sick leather jacket and the visionary guru stare. I'm long 300% of my net worth. Nothing can go wrong.

[Filippo]

This. The sheer misallocation of resources. The completely screwed up rewards for doing a good job, and punishments for doing a bad job. The strong disincentives against our best and brightest doing anything that's actually useful. Productivity only growing a little bit, in spite of massive improvements in automation. I could go on. Capitalism showed signs of insanity well before LLMs. It's still the least bad system we've figured out so far, because every other one we've tried had the same problems and then some... but I'm absolutely certain that at some point in the far future, historians will look at it in the same way we now look at the feudal system.

[Blazde]

Capitalism isn't really to blame, it's merely a risk factor (probably a least-bad risk factor as you say).

What's needed is robust governance of capitalism. It's long been known that capitalism's strength is in allowing economic autonomy, motivated to efficiency by free-market competition, and when allowed to do so is supremely good at allocating a wide range of resources control-type economies can't even dream of being aware existing. It therefore can't function well without a justice system intent on (relatively minimally at least) imposing that autonomy, fairness and competition, and curbing the ever-present tendency of concentrations of economic power to exploit and corrupt. Of course this never properly happens in practice but..

The US is now hideously bad at it. It's long nurtured and protected an array of tech companies whose business models simply revolve around establishing and maintaining a monopoly, but they've now gotten much bigger and more numerous. Monopoly is anathema to free-market capitalism, which US leaders continue bizarrely to claim adherence to. The US allows wealthy corporations and individuals to buy politicians with almost no limits, even as the wealth concentration keeps increasing and the federal government is increasingly indebted. It has poor worker rights, even in industries where there is no real overseas competition. This damages the labour market high-quality corporations used to invest in. It allows dual-class share structures banned or heavily discouraged in many capitalistic liberal democracies, and all those opaque finances that hide and glorify the mess long enough for it to be passed-off to the little people before the music stops.

Many of these factors haven't yet so severely distorted most other capitalistic economies, but there's certainly been general sliding and country-specific problem-areas, and more importantly it becomes increasingly difficult for others to manage rules-based capitalism in a global economy when the world's biggest economy cares nothing for those vital rules that make capitalism so powerful.

[Geoff Campbell]

Robust governance

Ah, yes. That's where the problem lies, I think.

Capitalism is, at root, a religious worship of money in large lumps. If that is robustly and fairly governed, no problem.

However, the governing is always done by other people. Almost exclusively capitalists. Who worship money in large lumps.

And then <*BOOM*>, your economic system is doomed.

The USA is showing us just how badly this all ends, but sadly the rest of the world seem not to be learning the lesson.

GJC

[Clausewitz4.1]

> Sorry but did I miss the punchline on that comedic valuation?

No, you didnt. Anyone with half a brain can see through it.

[VoiceOfTruth]

>> $2 trillion value on revenue of $4.6 billion with losses of >$8 billion.

I wonder where and when this insanity will end. It seems clear to many people, but I'm not a professional 'investor'.

I am more worried that average Joe Bloggs' pension scheme buys into this and ends up with nothing.

[Tron]

Pensions schemes should be OK.

They don't chase short term gains (BitCoin, AI etc) and tend to be naturally conservative in the stocks they buy, planning decades ahead. There will be a dip in all stock prices when the AI sector collapses, but nothing more than a typical reset.

Anyone who goes all in on AI stocks and doesn't get out quickly enough will be screwed. Oh dear, how sad, never mind.

A lot of the big numbers in the AI sector are promises that can be withdrawn - it's not real money. GAFA has insulated itself from the fallout: There is an AI sector rather than GAFA doing AI itself. They can feed on and artificially pump up the expansion, lose little when it collapses, and then buy up the residual IP for peanuts in the firesale. Some bits - the natural language interface, and the data scraped from anyone mug enough to use AI will be of value.

Anyone can make money from a bubble, but you absolutely have to get out before it pops. Some folk are going to find that their app locks up or the stock market suspends trading, before they do.

[Cubbie Roo]

Don't mention the vast swathes of GPUs sitting in storage, or Anthropic's run-rate plateauing since June . . . . . . no wonder they want to IPO at any cost, bag-holders are desperate to exit their positions before the house of cards comes down.

[Dan 55]

Free bunker if you invest $1bn or more

What they're saying is not a reason for investing in it, it's a reason for staying the hell away. Is our fate in the hands of raving lunatics, but as they're rich they get the red carpet rolled out for them?

[Juha Meriluoto]

Wish El Reg had a switch to filter out 'AI'-related content... I'm just fundamentally fed up with it...

[Ken Hagan]

I'm fed up too, but the thing that needs the switch is The World, not The Register.

[amanfromMars 1]

How very strange that so many might think they have a choice in such a novel matter.

Posing an “existential risk to humanity” was among the messages reportedly intended to allure investors, as if to say, “We could make you rich, but also you might not live to spend it.”

There is one thing for sure, and be in no doubt, ...... If you're not in, you can't dare who win wins. And it's not ever going away, you know, nor ever is it going to be put back into any kind of secure and locked box ..... this Universatile Virtual Force of Immaculately Resourced Assets.

And please, don't waste your time and efforts on believing any proposed rules and planned regulations will either halt or hinder the scale and alarming speed of AI progress and remote human instruction. That's a tale for the fairies.

[Anonymous Coward]

Re: How very strange that so many might think they have a choice in such a novel matter.

Are you confusing AI with ultrasound attacks or mind reading? They are different things

[amanfromMars 1]

Re: How very strange that so many might think they have a choice in such a novel matter.

Are you confusing AI with ultrasound attacks or mind reading? They are different things. ....... Anonymous Coward

No.

[Anonymous Coward]

Re: How very strange that so many might think they have a choice in such a novel matter.

> No. … amanfromMars 1

Yes, AI and ultrasound / remote mind reading are different technologies.

AI as in text processed by GPUs to do inference, may very much be a bubble, from:

‘Big Short’ investor Michael Burry is betting against some key AI names on a new, shorter time horizon.

Ultrasound / Remote Mind reading are military / intelligence technology, being used to target some intelligent people.

If you sell what you are collecting to an audience or AI companies, its a different business model from AI.

[amanfromMars 1]

Re: How very strange that so many might think they have a choice in such a novel matter.

Those military/intelligence technologies, being used to target some intelligent people, sound far too much like AI pronouncements/hallucinations in Neuro-Linguistic ProgramMING* resulting in activities attempting to alter and redirect the future course and plights and blights on humanity with alternative and competing renderings of ideally anonymous and virtually autonomous remotely controlled and commanded possible existences presenting unfolding realisations for engagement and/or employment and/or enjoyment, AC, to be considered as different in anything other than its flavour to favour and savour.

And if the master plan is, and please excuse the crudity, to polish a turd and maintain an ignorant and arrogant corrupt and belligerent status quo rather than embrace and explore and experience the brave new strange progressive worlds of Live Operational Virtual Environments, SMARTR** IntelAIgents are deployed to lead and prevail and ensure such nonsense always catastrophically fails, implodes and self-destructs.

* ......Neuro-Linguistic Program Mined IntelAIgent Network Games

** ....SMARTR Mentoring Analysis Reporting Titanic Research

That’s where y’all are presently at, is it not, whether you like it or not if the truth of the terminal decline and rotten decay in your systems be told.

[amanfromMars 1]

Re: How very strange that so many might think they have a choice in such a novel matter.

And please, don't waste your time and efforts on believing any proposed rules and planned regulations will either halt or hinder the scale and alarming speed of AI progress and remote human instruction. That's a tale for the fairies.

That’s where y’all are presently at, is it not, whether you like it or not if the truth of the terminal decline and rotten decay in your systems be told.

And how very strange, but not necessarily unexpected and unfortunately so extremely disappointing, that such is not welcomed and recognised as more than just good news in worlds of pain being rent asunder by mountains of bad and fake news

[Anonymous Coward]

Re: How very strange that so many might think they have a choice in such a novel matter.

> just good news in worlds of pain … amanfromMars 1

You are just whitewashing remote torture (sleep deprivation), nightmares, spying as AI.

Its technological superior spying? Yes. But selling data to AI companies and/or producing news (or fake news), sorry, its not AI. Its plain and simple IP theft and torture.

[amanfromMars 1]

Re: How very strange that so many might think they have a choice in such a novel matter.

But selling data to AI companies and/or producing news (or fake news), sorry, its not AI. Its plain and simple IP theft and torture. .... Anonymous Coward

You might like to reconfigure those opinions, AC, for selling data to AI companies and/or producing news (or fake news) is certainly never ever plain and simple IP theft and torture ...... although one has to admit and accept, whenever one knows what one is doing and what needs to be done, it is not difficult.

[Uh, Mike]

It's a New Arms Race

American companies can wring their hands all they want,

but they are not the only game on the planet.

Do you think nation states that don't have free press

are throttling their development in any way?

In the nuclear arms race, at least it took huge investments to move forward.

In an/the AI arms race, all you need is a data center.

And nobody can sniff the air or watch their seismographs to detect your progress.

[Anonymous Coward]

A marriage between Las Vegas and Russian roulette?

"Posing an “existential risk to humanity” was among the messages reportedly intended to allure investors"

What are they trying to advertise? A marriage between Las Vegas and Russian roulette? A chance to get rich if you roll the dices and put a gun to your head? With still no clear path to profit nor answers how this would make you rich?

[Anonymous Coward]

Re: A marriage between Las Vegas and Russian roulette?

They are likely aiming contracts in military or intelligence sectors

[stiine]

Re: A marriage between Las Vegas and Russian roulette?

I bet it was intended to excite Peter Thiel's wallet.

[Joe User]

They're like carnival barkers

Step right up, folks! Lay your money down, spin the wheel, and see if yours is the contribution that helps us destroy humanity!

[AbeSapian]

With Apologies to Norman Whitfield and Barrett Strong

(AI, huh) Yeah

(What is it good for?) Absolutely nothin'

Uh-huh, uh-huh

(AI, huh) Yeah

(What is it good for?) Absolutely nothin'

Say it again, y'all

(AI, huh) Huh, lookout

(What is it good for?) Absolutely nothin'

Listen to me, aww!