AI Investment Is Starting to Look Like a Slush Fund
nymag.com
nymag.com
Everybody's valuation depends on everybody else's valuation and a bad revelation takes everything down.
It's always the same idea, people just find different contortions to make it seem legit this time.
To be clear: I’m not calling AI a Ponzi scheme. I genuinely value this technology. But nearly everything else is: the markets, the policies, the promises.
Both things can be true. The technology may be valuable, and the investment in the technology may make zero sense from a financial perspective.
Very tangential rant and not an accusation: Popular misuse of the term pretty rampant these days, it's just like if everybody started saying "Buffer Overflow" for every kind of security risk.
To be specific, "Ponzi Scheme" is getting treated like a pejorative term for any kind of unsustainable profit, but it's a specific situation where an agent is lying to you about how much stuff exists and how much of it is yours.
At this point it's just been who has been paying attention. Look at gold's breakout this year, it's not happening in a vacuum.
Otherwise what is inflated away will be recouped by the interest rates when forced to keep borrowing.
everyone was crying how unsustainable these real estate prices are ... and then they kept going up still.
because supply fucking stagnates, yet demand in cities rises with the force of an angry god (cross-financed by the whole economy, of course)
https://www.fullstackeconomics.com/p/the-2000s-housing-bubbl...
Almost every podcast nowadays is just "Where is all this money coming from?", "This is just infinite money and you are just out of the loop." Feels so weird to hear.
I didn't expect LMs to be able to attack IMO problems this soon. There's also this video stuff recently discussed here, where it was explained that video models can be prompted to segment, solve blind deconvolution problems, solve maces, etc.
I'm not sure it's going to work out economically-- I think a crash is unavoidable because P/E ratios for firms like Tesla of 200 or whatever it is are silly, but I think that technology is going to keep evolving throughout and I don't see why some kind of mild not-quite-AGI-but-beating-humans-on-what-we-pay-humans-to-do-for-their-jobs type thing is not achievable.
If these video models get better and more available we might actually be able to use them to spot patterns in the natural world and in humans.
The dot com bubble came crashing down because we were still not there yet, at the turn of the millenium. We lacked the logistics to make large scale ecommerce viable, and we also lacked the smartphones to make it practical.
I generally feel we are in a very similar position right now. LLMs shows real promise, but also real limitations. I'm very skeptical it will bring significant utility in the next few years. Maybe in a decade or more, when other pieces of the puzzle are in place.
LLMs are already bringing significant utility to me. I don't think they're a good investment though: they're very expensive to train and somewhat expensive to run. And I'm extremely fickle and will jump ship to the next one without a second thought.
a) it isn't creating that value but taking it from somewhere else (past IME contestants and test preparers, StackOverflow, the job market, etc)
b) users fight to bid up the value of a single set of fixed capabilities, like rats fighting over the same scrap of leftovers. The market floods with generic copies of whatever stuff it can do.
That is to say it's utterly, humorlessly laughable to think that we're trending towards these things doing a whole economy worth of work.
Imagine the constraint on every breaking change going forward being 'can AI be trained to handle these changes against the pre-AI training dataset'. Because AI is an inertia machine. It takes all the momentum from everything it touches. Next step after it takes the devs jobs is to freeze out all breaking changes that could make the AI training set useless (and we don't have a new one yet, because it's a new breaking change, and also because, you know, no one is posting knowledge for free AI scrapping anymore, yay, huge step forward for humankind!!!). Inertia machine sucking the momentum out of everything.
They tried working with a movie studio for a video tool. They needed more movies than the output of the studio. How do you train in new things if you just happen to need 5 studios worth of 'new thing' but output is all from 'old thing' and AI generated? Inertia machine.
I feel like this tbh:
AI Skeptic: Sure its profitable but its a bubble...
AI guru: Sure it may be a bubble but its profitable...
Pick your sides logically because bubbles (by definition I think, are profitable till they aren't)
As software engineer, that means there are reasons to avoid investing in AI stuff, regardless of what I think about its trajectory.
1. If you become unemployed because it "succeeds", then a diversified portfolio probably does well as other industries reap the benefits, and if you have to cover lost wages you're at least "selling high."
2. If you become unemployed because it "fails", then there's a crash/recession where everyone is spooked and you get screwed by being forced to "sell low" during the transition.
I suppose there's also the chance you're put out of a job by powerful AI, but the "winners" are all lesser companies you never invested in, so you lose both your job and investments.
That would be such a radical societal transformation that I’m not sure we would come out of the other side even having a capitalist society.
And if generating those messages are cheap, well there might be competition as price maters more than quality to enough of users.
See, I am not an AI skeptic in the sense that I don't think the tech is great, I mean its kinda great but its also a lot overhyped tbh, but I mean its definitely a mix bag of opinions for me imo
But there are some people who think that just because a tech is great fundamentally makes it oh so lucrative and profitable.
That is the one thing that I really really took away from intelligent investor. And I think he even paraphrased it that this should be the one thing you should take away from the book iirc but I am not sure
Its that, just as how historically airplane companies have been kinda bad in profit even though they are technologically marvels. It doesn't matter the tech if there is a P/E mis-match imo / stock price is overvalued.
The same goes for tech.
You can appreciate tech while still thinking that its sort of being a bubble/shit's not good.
What are your thoughts?
Everybody has starry dollar signs looking at last part. CEOs are really bunch of pretentious little boys (and few girls) cargo culting each other, and showing very little innovation or originality... Clearly enough for markets, till its not.
These companies cannot turn on a dime. I would be very surprised if Microsoft isnt secretly developing a doomsday recovery plan right now. They got very deep into Narnia this time. I have a feeling this particular game of musical chairs will end much more abruptly than the prior ones.
Windows, despite the irrefutable enshittification, is still orders of magnitude more productive in a UI/UX/human/team setting. Anything linux that has to touch a non technical human needs a billion dollar effort to smooth over the rough edges. Even in the case of Steam Deck, I had trouble getting the machine going for the first hour.
Is it really, though? Most of the work in the Business Factory is done using cloud software these days. The OS itself has become mostly irrelevant to the end user and has been reduced to the web browser's runtime.
The rest is going to depend heavily on your industry, with some tooling just being better suited to a specific platform. Creative industries? MacOS. Web development? Linux. CAD? Windows. Use the wrong platform, and even if it technically works, you could still end up in a world of pain rather than having it Just Work.
Trying to hammer a square peg into a round hole is always going to lead to friction, that is not a unique Linux-only problem.
In the 2008 housing bust, there were NINJA loans (no income, no job, no assets). This time around, each GW ai-datacenter costs 40B of which say 20B is debt. The only profits coming out of these data centers are through the ad business models for meta and google, and the cloud business models for the hyperscalers. The rest is funded by equity or debt or capex. It seems likely that the tax-payer will be on the hook for the debt down the road given how ai datacenters are being projected as critical infrastructure - because - the profitability of the new investments are yet unproven and speculative.
Already happening. The tax/rate-payers are already on the hook for the cost of building new electric power generation [1]
[1] https://www.axios.com/2025/08/04/electricity-costs-bills-dat...
Sounds about right to me. Jensen Huang is playing well not as a business but as a global actor.
I wonder if OpenAI starts getting looked at for antitrust at some point. They now have significant ownership by Microsoft and Nvidia, 2 major parts of the AI supply chain.
Unless you're talking about employee or consumer taxes, but that's unrelated to businesses passing around money.
While I wouldn't invest in Nvidia (not now with the AI bubble liable to pop at any moment) the reason many people have is because they couldn't invest in Open AI (aside from via MSFT which is a less pure AI stock).
So Nvidia investors now get what they wanted all along: exposure to Open AI.
> Intertwined
And that is the _point_, this same pattern existed in all those previous instances too. OpenAI/Stargate and so on are just "fronts" for a team effort investment into building wildly unprofitable infrastructure costing upwards of 1T. e.g Saudi does not give a hoot about LLM technology itself, its just that there is infra being built, and once it is done, it will lock in energy demand in a massive way, and so it makes sense to help build this infra. Similarly for nvidia/gpus, which is why they invest in openai and such.
All the money people are dumping into this now, will be made back on the millions of small million dollar companies that will build products on top of this infrastructure, once it is built (which causes the money to move out of this and onto those smaller product companies - this is the market crashing/bubble bursting). I don't need to point you at the various SaaS/whatever companies making a killing on top of the extremely unprofitable cloud investments of the past, or the absolutely insane economic reach telecom has enabled today.
During this push to build the infrastructure, there will be a lot of short-term investors trying to make easy cash by greater-fool investing -- they have no intention of staying and building product companies after the crash. A portion of these that don't anticipate the crash's timing will lose lots of money.
Unfortunately, if Wall St. and such package this debt along with other debt that is bought by institutional investors/bonds/pensions, or if these guys and other wide-reaching funds buy the volatile debt themselves, then everyone who is connected to that (including grandma, including aunt's house) is exposed to that risk. This is what will decide whether it's going to be a tech bubble problem or a financial crisis.
And almost surely, the job market will dive for a while [1], since jobs depend on cash flow, and cash flow will dry up in the time period between the "infra building" state, and the "now we have high-margin products on this infra" state.
[1] The key thing is, and this is the primary problem, what "a while" is, is dependent on broader factors. The ability to pick up pace again after a bubble bursts will be dependent on the general economic health of involved countries as a whole. This is because job market diving leads to consumer/real estate/etc etc falling i.e wide reaching negative feedback loop = "macro slowdown"