AI is huge. But so is the mania of hype, hocus-pocus deals, and piles of real money fortified by leverage that caused stock prices to explode.
By Wolf Richter for WOLF STREET.
What is amusing is just how much talk there has been about the AI investment bubble, and what it will do or not do to the markets and the economy when it implodes or doesn’t implode: That it’s almost like at the peak of the Dotcom Bubble. That it’s much worse than at the peak of the Dotcom Bubble. That it’s nothing like the Dotcom Bubble because this time it’s different. That even if it’s like the Dotcom Bubble and then turns into the Dotcom Bust, or worse, it’s still worth it because AI will be around and change the world, just like the Internet is still around and changed the world, even if those first investors got wiped out, or whatever.
There are many voices that loudly point this out, and point out just how risky it is to bet on hocus-pocus money, or that explain in detail why this isn’t risky at all, why this is not anything like the Dotcom Bubble, why this time it’s different – the four most dangerous words in investing.
The talk fills the spectrum, and these are people with enough stature to be quoted in the media: Jamie Dimon, Jeff Bezos, the Bank of England, Goldman Sachs analysts, IMF Managing Director Kristalina Georgieva…
The focus is on the big-tech-big-startup circularity of hocus-pocus deals between Nvidia, OpenAI, AMD, along with Amazon, Microsoft, Alphabet, Meta, Tesla, Oracle, and many others, including SoftBank, of course.
OpenAI now has an official “valuation” — based on its secondary stock offering — of $500 billion though it’s bleeding increasingly huge amounts of cash. And there are lots of players in between and around them. They all toss around announcements of AI hocus-pocus deals between them.
OpenAI has announced deals totaling $1 trillion with a small number of tech companies, at the top of which are Nvidia ($500 billion), Oracle ($300 billion), and AMD ($270 billion). Each of these announcements causes the stocks of these companies to spike massively – the direct and immediate effects of hocus-pocus money.
OpenAI obviously doesn’t have $1 trillion; it’s burning prodigious amounts of cash. And so it’s trying to rake in investment commitments from the same companies that it would buy equipment from, and engineer creative deals that cause these stock prices to spike, and so the hocus-pocus money announcements keep circulating.
OpenAI’s idea of building data centers with Nvidia GPUs that would require 10 gigawatts (GW) of power is just mindboggling. The biggest nuclear powerplant in the US, the Plant Vogtle in Georgia, with four reactors, including two that came on line in 2023 and 2024, has a generating capacity of about 4.5 GW. All nuclear powerplants in the US combined have a generating capacity of 97 GW.
But it’s real money too. A lot of real money.
Big Tech is letting its huge piles of cash spill out into the economy to build this vast empire of technology that requires data centers that would consume huge amounts of electricity to let AI do its thing.
And these “hyperscalers, are leveraging that money flow with borrowing, by issuing large amounts of bonds.
And private credit has jumped into the mania to provide further leverage, lending large amounts to data-center startup “neocloud” companies that plan to build data centers and rent out the computing power; those loans are backed with collateral, namely the AI GPUs. No one knows what a three-year-old used GPU, superseded by new GPUs, will be worth three years from now, when the lenders might want to collect on their defaulted loan, but that’s the collateral.
The data centers are getting built. The costs of the equipment in them – revenues for companies that provide this equipment and related services – dwarf the costs of the building. And stocks of companies that supply this equipment and the services have been surging.
The bottleneck is power, and funds are flowing into that, but it takes a long time to build powerplants and transmission infrastructure.
Is it really different this time?
So there is this large-scale industrial aspect of the AI investment bubble. That was also the case in the Dotcom Bubble. The telecom infrastructure needed to be built out at great cost. Fiberoptics made the internet what it is today. Those fibers needed to be drawn and turned into cables, and the cables needed to be laid across the world, and the servers, routers, and other equipment needed to be installed, and services were invented and provided, and businesses and households needed to be connected, and it was all real, and it was all very costly, requiring huge investments, but progress was slow and revenues lagging, and then these overhyped stocks just imploded under that weight, along with the stocks that were the pioneers of ecommerce, internet advertising, streaming, and whatnot.
The Nasdaq, where much of it was concentrated, plunged by 78% over a period of two-and-a-half years, investors lost huge amounts of money, many got wiped out, thousands of companies and their stocks vanished or were bought for scrap when that investment bubble crashed. And a year into the crash, it triggered a recession in the US – and a mini-depression in Silicon Valley and San Francisco where much of this had played out.
Yet the internet thrived. Amazon barely survived and then thrived in that new environment. But Amazon was one of the exceptions.
In this mania of hype, hocus-pocus deals, and huge amounts of real money fortified by leverage – all of which caused stock prices to explode – markets become edgy. Everyone is talking about it, everyone sees it, they’re all edgy, regardless of their narrative – whether a big selloff is inevitable with deep consequences on the US economy, or whether this time it’s different and the mania can go on and isn’t even halfway done.
Whatever the narrative, it says risk in all-caps. Anything can prod these stock prices at their precarious levels to suddenly U-turn, and if the selloff goes on long enough, the investment bubble would come to a halt, and the hocus-pocus deals would be just that, and the whole construct would come apart. But AI would still be around doing its thing, just like the Internet.
Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the mug to find out how:
If the promise of AI rests on ever-increasing GPU cycles, then the popping bubble might not leave behind a very functioning remain.
At least not this very energy-hungry US version. Seems more likely that the survivor will be something leaner and cheaper to operate, with greatly reduced goals.
Wolf has been stuck in treasuries while ai stocks have gone up 1000s of %s.
If you did a little research the demand is there as all legacy cpu compute is being replaced and ai is rapidly evolving with $$$ saving utility for corporations.
You have no f**king idea what I’m stuck in. But I do have some Treasuries. However, I’m not stuck in them. They’re very liquid.
I’m stuck on you, Wolf
I like your articles. Always good informative stuff.
With that said, most times it leaves me scratching my head, and thinking…..ok…..so now what.??
interesting times we live in.
An old friend of mine in his 90’s jokes that the only reason he sticks around is to see what will happen next…..
Haha.
“With that said, most times it leaves me scratching my head, and thinking…..ok…..so now what.??”
Conservatism/diversification is almost always the way to go, in terms of long term results.
There is *always* some coke-sniffing, meth-head sh*t-shovelling securities “salesperson” hawking *something* (the “metaverse” hype imploded like 2 minutes before the “AI bubble” launched).
Sure, “you might miss out” on some rocket-ship stock – but somebody always wins the lottery too and there isn’t a huge industry of intermediaries shovelling out “lottery advice” and investors don’t tie themselves in knots about having missed the lottery numbers.
If you want to take a flyer fine – but do it with 5-10% of your portfolio – not 50-100%.
I think that is about the best, most concise “investment wisdom” there is.
(Even during the long, dark night of ZIRP, the old saying “You can eat well (high returns) or sleep well (safe returns)” was true – and, again, the best, most concise “investment wisdom” there is).
1000’s of %? Don’t be dumb.
Pumpers must have found this site through algo.. gotta do the work to keep the hype bubble train going 🤣
And this AI – generated ‘comment’ demonstrates how much nonsense is being created
Synonyms for “Artificial”
fake, false, bogus, mock, sham, imitation, simulated, counterfeit, spurious, synthetic, ersatz, pretend, phony , faux pseudo, dummy factitious forged substitute imitative plastic mimic manufactured fabricated
Doesnt sound so sexy does it?
Excellent article. It made me aware of the hazards my investments may be facing.
OpenAI and it’s partners would love a crash. They don’t want the competition, they have the investments and money for now, a crash would eliminate all the small shops trying to gain a footing. It will crash, and it was most likely designed that way.
This time around may be different, because we have a world at war. Every fighter plane from B52 down through F35 use armaments systems run by AI, I have a relative who designs them. All Abrams, Bradleys etc. ground vehicles driven by my grandson have similar, as do the Navy vessels. All using top end Nvidia chips available to nobody else. The last MRI and other scanning machines that you last went to for that odd pain in your guts rely extensively on AI to process results. Your Tesla attempts to avoid dogs and pedestrians with AL recognition. There is no going back.
What’s hilarious is all the CEOs are using the same talking points that ai has the promise to cure cancer and discover new drug and chemical compounds while the truth is the promise of AI is the pure capitalist goal of making $$$ by putting 10s of millions onto the unemployment line.
Yes and even worse, AI slop is like the weeds of the Internet, it grows and grows choking off the gardens and even the food crops that have actual useful content. It’s gotten so bad that lots of short story, art or other creative sites have have had to all but shut down, since the legitimate, hard-working quality entries are drowned out by the sheer rivers of low-effort AI slop. It’s basically ruining paths to a creative career in writing, film, visual art and other areas by making it that much harder to get anything noticed amidst the flood of AI garbage.
While there are a few good uses, that’s become the main “contribution” of these new AI companies, fertilizer for more and more AI-produced weeds. And now with so much AI slop as input, the AI models are training on their own previous output slop and putting out even worse and more useless slop. And the best part, Joe and Jane American get to pay for all this with the massive increases in power and water bills for all the loud, irritating data centers dedicated to putting out more and more AI slop and garbage.
This sounds very over done. Modern war planes use computers, but why would they learn on the job, the point of AI. As for the ancient B 52, that would need a heck of an upgrade. Armaments run by AI? Is it responsible for friendly fire, or does the human make that decision?
No doubt computer scanning of MRI images can be improved, but the first improvement would be in making the tech available at all, reducing the price in USA from between 2 K average up to 10 K for a brain scan. ( My nephew lives in Japan. A friend was examined and doc thinks MRI warranted. Gets it 2 HOURS later. Covered by Japan’s medicare. It would be covered in Canada too, but you wouldn’t get it in 2 hrs.
Incidents about the limits of Tesla’s self- driving hit the news daily. What size of dog? If a rabbit or a dog runs in front of your car while travelling at high speed, you aren’t supposed to slam on your brakes without checking behind. A human decision is required.
Tesla’s rolling out its self-driving tech is an example of incompetence and recklessness that is endangering the public. Don’t use them as an example for self-driving tech. Use them as an example of incompetence and recklessness.
The Waymos are superb, replacing drivers as they go in Phoenix, San Francisco, Los Angeles, Austin, and Atlanta.
TEF & Nick,
Check out the 2026 Ducati Panagali V4R. This motorbike starts at about $50k U.S. dollars. It is designed for the ultimate experience possible when riding and racing on tracks.
It uses algorithm predictive software to anticipate and adjust, in microseconds prior to, what the bike and rider will encounter next. All of the electronic rider aids can be set any way the rider chooses, but the bike’s got this in addition:
” Ducati Vehicle Observer (DVO): the advanced algorithm developed by Ducati Corse
The Panigale V4 R raises the level of its electronic controls package with the use of the Ducati Vehicle Observer (DVO), the algorithm developed by Ducati Corse. This allows for a real time estimate of physical, kinematic and dynamic parameters that cannot be directly measured in practice, which impact on the ground forces, acceleration and maximum torque the vehicle can support in different riding conditions. The system simulates a network of virtual sensors capable of measuring more than 70 physical quantities in addition to those detected by the inertial platform, allowing for a precise estimate of both the operating thrust and maximum forces that can be tolerated by the bike at that specific moment.”
Yes, the motorbike anticipates what will soon happen, and adjusts its computer controlled settings continuously and unobtrusively to be faster and safer.
It sure as hell ain’t “self-driving”, but it is the best performing motorbike that’s for sale to the public. And its computer’s predictive adjustments are a big part of its magic.
Full disclaimer:
No electronic rider aids were used this afternoon on my Italian V4 motorbike as I went to and from Trader Joe’s in St. Paul on a grocery run. But, just in case, they were there on standby.
Absolutely LOVE your ”comment” PR.
Thank you, and please do your best to continue.
Wolf is right. I live in Austin in ground zero for Tesla selfdriving testing.
Musk chose the most touristy area to test his car and it’s a joke. Its also dangerous for the public.
On the other hand I watched Waymo map my street for years and totally trust the car. Whenever I run errands I travel next to them frequently. They are excellent drivers.
This misses Wolf’s point though, the things that you’re referring to involve various forms of computation some of them referred to as AI, but most of this is nothing new and it long, long pre-dates the AI hype bubble with the chatbots made possible with Nvidia’s GPU chips–and to be more specific, TSMC’s chip-making and ASML’s chip-making machines. “There is no going back” is irrelevant here, you could say the same thing about email, social media, the Internet, programming even the modern computer itself–like those, most of what you’re referring to involves tech from before the current bubble, and has nothing to do with the current AI bubble and why it’s so over-hyped.
The issue with the current AI bubble is it outrageously hypes a sub-set of AI tech with comparative poor ROI and often, backwards ROI that makes things worse for organizations with all the hallucinations and bad data, leading to contamination of once working systems and the worst kinds of disruptions. One thing standing out about the current tech bubble is how unpopular and broadly disliked it is among Americans, even if the dot-com bubble (and earlier AI, PC’s, the Internet and Web in general) lead to some consternation, it was broadly useful and seen as beneficial to people and welcomed overall.
The current hype is producing an anti-hype backlash I’ve never seen. Partially because even the “best case scenario”, replacing workers (and thus wrecking the US consumer economy) is the worst kind of Silicon Valley tech bro delusion. But even more, the AI tech bubble is uniquely damaging, arrogant and costly to broader communities, with the tech barons trying to offload those costs as they overhype things further. The data centers anger communities with all the increased utility costs and noise, And for all inflation with those massive increased costs and damage to environment, we get a tool that’s much less useful than previous tech breakthroughs, polluting the information stream, hallucinating like crazy, hurting creative projects and education and just filling the Internet with junk. It’s at the point where AI models are taking in their own AI slop and putting out even worse junk, crowding out productions where actual people are putting in the effort and making useful things and generally lowering the quality and value of things online. Not to mention of it fuelling one of the worst debt-fueled bubbles that threatens the whole US economy. That’s why most Americans are cheering the popping of the AI bubble, it still has some good uses and yes these will come out as the bubble collapses and we see some things that stay. But, Silicon Valley showed a lot of its worst instincts and simple arrogance in trying to force this on people’s throats, over-hyping it to point of costing a lot of jobs and then leaving a worse, less accurate trail of Internet content and organizations now with huge damaged data streams in the wake.
I’ve not seen any real usefulness in my professional or personal life for open AI. Proprietary AI on the other hand is coming along in enough industries that it’s helpful in siloed use where the employees teach it what they want it to know and what they need it to do. I don’t appreciate being asked to speak to it like it’s a human – it is not human or even human like, and I quit playing pretend a long time ago. Open AI is total trash.
The only people that I hear make these kinds of claims are the people who define “AI” as “software in general.” If we define AI as the kind that’s relevant to this article, which is the kind that requires an internet connection to a data center, and if the military is really dumb enough to want to design their systems to require something like that for fire control, then we are basically screwed in the next war.
The AI bubble will collapse just like the Dot Com bubble collapsed in 2000. Only this time I see a revolt against this form of technology which provides little or no benefit to the average joe, where the Internet revolution did help provide increased choices and productivity and survived after the Dot Com crash. AI will only fatten the pockets of those who are already rich. Many people are unlikely to embrace the de-humanization which will occur with this transformation to AI. People are already sick of talking to BOTS, voice mail hell, and all these automated machine driven hellholes.
I’m already hearing AI being used as a pejorative.
As in, that song? Generated by AI.
The voice in that ad? AI all the way.
Or that painting? AI, just look at it. Can’t you tell?
Methinks that it’s just another fad with a lot of money behind it.
Yes been seeing this too and you and Swamp Creature describe it well, what surprises me most about the AI hype bubble compared to previous tech bubbles is how negatively it’s perceived by the public and frankly how it just detested by them. For all the flaws and over-hype of PC’s, Internet, Web, dot-com, even blockchain, they were mostly seen as overall beneficial or at worst, mainly annoying without doing too much damage. Americans especially seem to really detest the AI bubble more than any other. Part of it is Silicon Valley totally lost the plot with the PR part, hyping up it’s most anti-social aspects (the promise of not needing to actually socialize or pay workers and basically wreck a consumer economy).
But it’s also that so much of new AI tech seems to the very definition of low-effort encrudification (another word usually used for this, don’t want to trigger the text filters). It’s based on basically stealing the unique, hard work of others without compensation or even citation for training, then basically plagiarizing at best–but in reality putting out products in high volume so full of hallucinations and accumulating errors as to make once reliable resources useless. Like someone said Pinterest has basically been ruined by all the AI slop and low-effort, bot-generated content, same with Facebook, Twitter, even Google and other search engines and even worse, even once-reliable high quality sites and data sources. Ironically AI slop may be what brings down Meta and other big tech companies by junking them up with so much volume of low-quality, contaminated content to make them unreliable. All while also messing up data streams and work-flows at companies that prematurely adopted them and pretended they could cut labor costs. And for all this junk and mess, de-valuing creativity, hard work and actual quality innovation–Silicon Valley wants to dump the costs the public for the all the skyrocketing costs of loud, annoying data centers and the power and water they use.
“People are already sick of talking to BOTS”
Are they? Most of people I know, especially younger, show the opposite.
Do they text or call? Do they make a restaurant reservation over the phone or the internet? At a casino, do the play blackjack with a dealer or press buttons on a slot machine?
I’d rather talk to a help bot that is available than wait on hold 20 minutes to talk to a person. I think therapy AIs will be popular because people are uncomfortable telling problems to another human who they feel will judge them.
We already anthropomorphize every other inanimate object; we’ll do that even more easily with AI. Ever played in VR? Then human brain accepts it with startling ease.
It’ll become part of our normal lives and we’ll adapt to it easily.
What you describe has nothing to do with AI and everything to do with younger people wanting to do things online. And now older people have gotten on the online train. The self checkouts at my Costco used to be all young people. Now the older customers use it too.
People like saving time. That’s totally separate from AI
“People like saving time. That’s totally separate from AI”
But it isn’t that. It’s actually less time in most cases to call a retaraunt and make a reservation that to navigate a web page.
Talking on the phone is faster than a long texting session.
Slot machines are only faster at draining your wallet.
I see a frequent preference to connect to a machine than to connect to a person. And it’s growing.
“I see a frequent preference to connect to a machine than to connect to a person.”
n=1, but I actively avoid all customer service “bots” and call places to make reservations frequently. As you correctly state it is faster.
In the future, there will be a bull market for all things authentically human.
I love chatting with bots, especially the XM radio one and Amazon one.
With XM, when my subscription is up, it asks if I want to renew at $29.95/yr….I say NO.
Bot: OK what price would you like?
Me: $4.99 annually.
Bot: We don’t have that option..
Me: OK, cancel my plan immediately
Bot: OK, $4.99 annually it is.
With Amazon (my favorite)
Me: I would like to return this item
Bot: Sorry. It’s a non refundable item
Me: OK, then give me a direct replacement
Bot: Sorry, not an option
Me: OK, but the package was slightly ripped
Bot: You mean damaged or defective?
Me: Yes, damaged
Bot: OK..full refund to your CC issued
Me: shall I send the item back?
Bot: No, just keep it
Me Thanks
Bot: Your welcome and thanks for being such a great Amazon customer.
Old Octogenarians like me really are getting into chatting with these bots! LOL
The AI bubble is even worse than other bubbles because it’s now basically putting the dead Internet theory into action. AI slop is like weeds, and the loud power-wasting data centers are pumping out so much of it’s taking over the fields of true useful content, some corners of the Internet are piling up so much they’re becoming majority AI garbage. To the point that now AI models are training on their own earlier output slop, and the AI researchers are starting to see model breakdown in action.
And again best part, it’s average Joe and Jane American are being forced to pay for the slop ruining actual valuable parts of the Internet, with higher utilities costs for all the power and water used by the big AI data centers. It’ll have some limited uses but it’s worse than lack of ROI, it’s backwards ROI, truly the worst
If X is anything to judge by, Social media have been absolutely deluged with AI slop. Interacting with actual humans on that platform is getting much harder. I’ve seen similar AI inundation on Instagram as well. The big networks are being flooded with it and doing nothing about it.
The Open AI slop is really really bad. I googled a question about a highschool I was at watching a football game. Gemini told me that the Highschool closed it’s doors in 2024, really??? The sense of paranoia this invokes is terrifying. I can’t trust Google but who knows what it’s telling other people and what they are believing as truth. Can’t remember who mentioned Pinterest above in the comments but it is so sad because It was the only platform I still enjoyed up until the Ads and AI took over.
” I can’t trust Google but who knows what it’s telling other people and what they are believing as truth.”
I agree with you…but contemplate the horror of the (politically compromised) 3 media oligopoly that almost exclusively “informed” American life from 1955 to 1995 and “co-built” the current wonderful world (by supplying politically contaminated perspectives).
As bad as modern media manipulation is – alternative info sources are a single click away in there era of the Internet.
In the era of ABC/CBS/NBC you had three flavors of blow-dried Leni Riefenstahl and little else.
“This time it’s different” – famous last words before the bubble bursts./
Dow ends nearly 880 points down as U.S. stocks plummet on Trump’s threat of new China tariffs
DJIA -1.90%
SPX -2.71%
COMP -3.56%
Never confuse the catalyst for a big move with the underlying market anxieties that cause market participants to be skittish. Butterfly effect is real.
The specific news that moved the market at the dot-com and Great Recession market peaks often had little to do with the real concerns. Like the final straw that broke the camel’s back, when the market is skittish inside, it only takes a few buyers becoming sellers to change the market direction in a big way.
I saw dip-buyers lurching in yesterday afternoon. Goes against your simple thesis that ” it only takes a few buyers becoming sellers.” There is still a lot of cash wandering around looking for somewhere to go. But yes, once a cascade starts, it can be quick.
Gold up 1.59%
Bitcoin down over $4,000.00
Vix up >30%
And after this awful, catastrophic day which will live in infamy, the market is back to where it was all the way back…in September.
LoL. I hope I don’t live to see the eventual collapse of the current fiats and their tethered digital market places across the globe. However, what do you suppose shakes out??? I have ideas
“all the way back…in September.”
The absolute correct answer to every panicked hype-meister who insists/demands resurrected/perpetual ZIRP to re-inflate every bubble of every stripe – from equity prices to government spending.
Rolling back prices/costs to 3, 10, or fr*ging 30 years ago will not result in the apocalypse – as somehow the world survived (pretty darn well) at *those exact same levels*.
Very timely article, Wolf, thank you.
Today Nvidia did outside reversal (from all time high, aka bearish engulfing or something like that), and so did Semiconductors, and so did Nasdaq.
Long weekend for people to brood over and start moving money to cash. Let’s see if there is continuation next week.
For “this time is different” Jeremy Grantham notes that there is always a “real change” or “real industrial revolution” to every Great Bubble. Example he uses is at the top of the railroad bubble they were building 6 lines in parallel between two cities (think it was from Buffallo to somewhere) .
To summarize, Jensen Huang is selling shovels during a gold rush.
Of course, I may be wrong, but I generally have noticed that when people start calling something a bubble during excess exuberance, it’s not “peak bubble” yet. It feels like we are closing up Act 2 of a 3-Act play. Historically, I feel like when some of those who cried “bubble!” start capitulating to the bubble, then we are REALLY in peak bubble mode. In early ’99 people, correctly, called a bubble. Some of those same folks thought “this time is different” by the end of that year.
Either way, this data center build out has gotten stupid.
You may be right about that.
“Since we undoubtedly were entering upon a genuine bear market I was sure I should make the biggest killing of my career.”
“In due course I sold again, and again they went down promisingly and then they rudely rallied.”
“One day it looked as if not a bear would be left to tell the tale of the strictly genuine bear market.”
“I sold all I could, and then stocks rallied again, to quite a high level. It cleaned me out. There I was right and busted!”
– Jesse Livermore, Reminiscences of a Stock Operator
But maybe this time is different and the same, at the same time. The same in that it’s a bubble and different in that this time everyone sees the bubble and expects it to pop and actually it does.
Having had a ring side seat to the 1980s Japanese bubble, I agree.
I think this won’t top out until the level of insanity is far higher than at present.
I remember a TV commercial of the dot com days, in which a fellow whose car has broken down gets in the cab of the tow truck and learns that the tow truck driver owns a private island bought with stock market winnings, and is just continuing tin the towing business cuz he likes to help people. Of course the fellow whose car broke down is shown dying of envy.
Perhaps it’s just cuz I no longer watch TV, but I don’t think we’re there yet.
Your said “long weekend.” I already checked, and the markets will be open on Monday October 13 even though not banks, post offices, etc.
AI is going to require massive new nuclear plants to supply its energy, but AI can never be put in charge of running those same plants. The purposeful non-deterministic (“intelligent”) and flaky nature of AI means it can never be trusted with something important.
“Open the pod bay doors Hal”
“I can’t do that Dave”
Ha ha
Google put AI/ML in charge of datacenter cooling nearly a decade ago for significant savings in power costs, better than the human algorithms it learned from. Some form of AI will absolutely be used to increase the efficiency of any modern system.
I think it’s important to remember the distinction between the relative “intelligence” in these systems. Your example is a great one but targets a narrow, well defined system that is in the grand scheme of things only 1-2 orders of magnitude more complex than what is achievable with more deterministic classical techniques from control systems theory.
A lot of the things “AI” is being applied to are much more open ended and not as great of a fit.
There’s a huge difference between using something like linear regression or SVM to predict optimal control inputs vs asking a chatbot “what should the rpm of this pump be given these inputs”. Classical machine learning is an excellent fit for this problem but LLM AI is unnecessary here, at best a proxy to another ML tool it calls. Why not just skip it altogether?
I have strong faith in ML as an optimize but AI is really not that intelligent it’s just the worlds best auto complete, whether it’s words sounds or pixels it’s still just making a best guess, which is not good enough for a significant number of applications that require critical precision. So I don’t see it as intelligent in that it can’t properly evaluate its own decisions in a corrective way like a properly motivated human is able to apply the scientific method to their own behavior.
Excellent comment. I agree on the technical detail. However, I expect everything you say to be ignored en masse and for it to be grossly misused at the insistence of clueless managers and poorly trained technicians. (And the good ones will either hold their nose and comply or get fired)
“Some form of AI will absolutely be used to increase the efficiency of any modern system.”
But that is sort of like saying “some form of computing” or “some form of technology” will increase the inefficiency of…
The real question is whether or not what *we have today* or can reasonably/honestly be said to be near to today – can actually generate the magic-adjacent hype promises being pitched *today*.
HAL is a long, long, long way away from a tightly-defined process, with exact optimization parameters, focusing on a tiny problem set.
AI today is being sold as God 2.0 – with matching levels of funding.
A simple tweet from our fearless leader over the weekend could swing this mess in the other direction. What a crazy world we live in.
My sister in Connecticut who is retired and barely makes her monthly bills is not the least bit worried about this market meltdown. She called me on my birthday yesterday and we chatted about her grandkids.
10 GW of power for these AI server farms? Better get going mining some copper, and lots of it. The cable manufacturing companies will need all they can get.
Yes, every deal must now have Gigawatts along with $Billions. When they start making up buzz words like “hyperscalers” or “petaflops”, you just know they’re pulling your chain. I especially like the use of a verb as a noun. Quickly, we need more compute!!
“More compute, more revenue.” – Sam Altman
I agree about the verbs. Dumb.
“Let’s go get some eat”
And I think the quote should have been “more compute, more earn” to really have the AI edge. If you don’t get it you should just “enjoy being poor” like the Bitboys say.
Petaflops is out of date. Exaflops (1000s of petaflops) are now in.
Sam Altman reminds me of Sam Bankman Fried.
I have a theory that this hype cycle is a bunch of AI nerds who know full well what happened during the dot-com debacle and know that they have a limited runway to build out as much infrastructure as possible while investors are ready, willing, and able.
Right now, they have piles of money and a favorable political climate to build, baby, build. That is not going to last for more than a few years.
Note that it takes 20 to 30 years to put in a sizable mining project in the U.S.
Totally agree, the recent deals and exploding prices are alarming.
I took advantage of the obvious bubble as long as I felt “safe” doing so. Last week told me I needed to get out. It was by far my best week and best month ever trading. I closed about 90% of my positions in AI related companies.
I am happy to walk away with a big tax bill for the year and let someone with more guts than me see if they can walk it closer to the precipice for even more gains. Good luck brave souls!
I’m ready to pull the chain on the light switch called “the internet.” Why? Because I watched a simple, short video this week of a cat riding a skateboard. Nothing crazy…just a cat puttering a few feet on a skateboard…very similar as when the local news had a slow news cycle day and showed silly videos like a squirrel water skiing behind a little remote control boat. You knew it was real.
My point is I’m now always questioning what I see is real or fake and it’s only going to get worse. Why subject myself to this constant eye worm?
What’s scary is when AI will start showing deep fakes of everyday people and events – just minor tweaks (nothing fantastical). I still don’t know if the cat video is real or fake.
My daughter is an artist and designer. She says entire sites like Pinterest are now completely worthless, infested by AI thievery. AI is not building anything worthwhile, it’s just stealing everything on the web and trying to peddle it as its own. Before it’s we will wealth destruction on a massive scale.
Before it’s over, we will see wealth destruction on a massive scale. (Safari thinks it knows better than I do what I want to type!)
OMG you too???
Yes this is great point, one of the other “contributions” of the current form of AI is mass plagiarism and then pollution of once reasonably curated, high quality data sites with derivative junk. It’s another way the current AI hype and AI bubble is if anything worse than having low ROI for all the billions of dollars spent, it’s actually backwards ROI lowering the value of a lot of good, independent hard work. It’s gotten to the point of good art and creativity, and simple good production in so many areas being undermined, like filling a once high-value art museum or library with pure junk and badly done copies.
Totally. The entire revenue model for Pinterest is gone. Everyone knows those products don’t exist, so sponsors can advertise all they want and no sales will happen, which leads to cancellation of the campaigns that feed Pinterest revenue.
Facebook will still have local advertising locked, but people are wise to the rest of the garbage.
One 6 foot tall 3 foot square box with the cutting edge chips running on all cylinders needs 132,000 watts of power and will run all day long. A server room full of those boxes needs cooling that involves evaporating water. Hyper-scalers will end up building the power plants on site. If AI fails, at least the power grid will have more production points. They won’t be built to utility specs but who cares!
I’ve had the 1999 vibe for months and months. Just like that show this one will be fun to watch.
Per X.com’s Grok AI…
“Building nuclear power plants on the Moon to power giant data centers isn’t pure science fiction—it’s a concept gaining real traction amid the AI boom’s energy demands and NASA’s push for lunar infrastructure. As of October 2025, prototypes for both lunar nuclear reactors and small-scale data storage are in testing, with full-scale deployment eyed for the 2030s. In short, it’s not unrealistic—it’s inevitable for niche uses like secure backups or space-data processing… For “giant” Earth-replacement centers? Latency and costs make it a tough sell until 2040s.”
Grok is very good at bad humor.
Yes, true, I used it a lot.
On the question about AI. There are three categories of AI:
1. Narrow AI (most of what is “free”)
2. General Intelligence (certain subject matter)
3. Super Intelligence (surpasses human intelligence)
It is Super Intelligence that is relevant for those I see here (investor, business starters/owners, various forms of arbitrage and the like.)
The ability of a human-like machine (Super AI) to find looking at past, present, and future in real time data to assess, and execute (like Larry Elisson) is not something any body here can compete with…. Say you have 5000 people like this in the USA and I have seen how systems can be set up to soak a big chuck of different small markets in brutally efficient ways.
Look at buying land. It is going in this direction and there are strategies I have seen (in Northern Ca) where certain people took over where the locals had no clue and next thing you see is most of the land offerings’ go up in value.
How about Citadel paying Robin Hood to see its buyers, in real time, buy and sell stock so they could soak a bit of money from this data.
Unless you are part of the Super Intelligence crowd (or a Senator) you will get eaten alive in what is to come.
To think you can compete with this seems obviously foolish and I would be interested to evaluate any thinking that believes it can compete or overcome Super Intelligence….
Soon, all those here will have few opportunities to make money as they do now.
As I am retired at 70 with 100k income, have a small mortgage on a horse ranch, and can build a home on some land in TN (they hate taxes) near a university setting It will be fascinating to see how Capitalism as we once knew it will evolve.
So intelligent AI can do everything better but can’t figure out how to best run a country’s economy and government so the population doesn’t end up homeless and starving?
Far more a Dodge in Hell scenario.
Looking at the uses of AI, I quickly see limit just like alt coins. But YMMV.
A lot of specific uses, and talking cats.
meh.
As for trading, other than skimming pennies, meh.
you would think they would be pushing skews on options to encourage more gambling, instead I see huge premia. Or are we really going to have 5% up next week followed by 10% down?
It’s not the obvious, those cakes are well known and just going to be endured. It’s the what breaks along the way.
And it usually takes a lot longer than I think.
My friend getting another moonshot in his late career at Nvdia says they are just selling shovels, and he thinks it is just nuts. But we are getting old, and each time everybody starts saying this time it’s different, it ain’t.
The future is here, but it just isn’t evenly distributed, and of greater concern is the number of people who no longer even count on an economic scale.
It is just beyond inequality, it now engulfs entire lives being wasted in REady Player One.
“TN (they hate taxes)”
Isn’t sales tax in some TN cities nearly 10%?
Not only AI but you talk to most people, they will tell you this time is different, when it comes to Crypto, and especially in housing….
What happened to 2008 can and will never happen again or the FED will always bail us out …etc.
Clearly the bubble will burst with winners and losers as AI is used in many places and that will only increase. What I find more interesting is where it will go and what societal impacts it can have. Whether it gains the intelligence to rival humans isn’t out of the realm of possibility although for me larger risks exist with bad actors, whether they be governments and militaries or simply utilizing it to reak havoc. Not to suggest all of that isn’t possible today on a different scale but AI will super charge that. There clearly is no governance at all in this area so things will go very wrong and there will be abuses.
That’s easy. As in every information revolution, it will be 90% pr0n.
I’ll see myself out
” …it says risk in all-caps.” As does my freudian slip.
Funny how everyone is thinking about this — I’m digging deeper and deeper, and my recent chat with Google box, seems on track:
“The commoditization of hardware: Over time, networking hardware became a commodity. Competitors like Arista Networks emerged, eating away at Cisco’s market share with more focused, higher-performance switching.”
The hyperscaler stars will eventually be eaten by the nanoscalers, who end up using software innovation, to jump around obsolete hardware.
The only example I have for that is Cisco, who once had great hardware, they still do, but nobody cares.
To me, the real danger in this arms race, is the cash being spent on data centers that will end up being paid for utility customers that are being conned into building out unneeded capacity.
AI is a fine addition for internet 2.0, accelerating excel from a spreadsheet with power steering — to a spreadsheet with a faster engine — but overall, the amount of companies that will require generative AI is extremely overhyped imho.
Lets hope to get some more reliable nuclear power out of the whole fiasco
the backbone of an “all of the above” energy strategy
Not exactly true that “nobody cares” about Cisco hardware. Telecos care A LOT because Cisco is an integrated seller and the biggest, most powerful switches are not made by Arista. Arista is not higher performing but rather lower price commodity products. Big data centers also care – not so much at the individual rack layer but at the convergence layers where all that data traffic converges to switch that Arista and other do not make to handle that type of capacity. And of course Cisco make a whole lot of other stuff that is not routers and switches. So, I know what you are trying to point out but the commodity story is not the whole picture.
The unused capacity post any burst could be used to serve manufacturing. Part of the reason the tariffs were introduced was to incentive the restoring of some of that( It was never going to be all of it).
The question is would it be enough to soak up that excess from any AI bust. That’s anybody’s guess.
A lot of these AI compute farms could be reconfigured into classic data centers in the long term. Very expensive hardware data centers but it could be done.
Julien Garran of MacroStrategy Partnership wrote in a rather pessimistic report Friday that a “misallocation of capital in the US” led by AI is 17 times bigger than the dot-com bubble and four times bigger than the 2008 real estate bubble. If true, then it is impossible for AI to come anywhere near justifying its expense in the forseeable future.
Both Consolidated Edison and Westinghouse went bust building out the US electric infrastructure, but they were bringing electric light to homes, replacing oil lamps. All the other apps followed. What benefit could AI bring to top that act? Eventually, who knows, but ‘eventually’ doesn’t work for calculating returns on investments, and these are by far the largest in history.
Four times bigger than the real estate bubble isn’t right. Dunno how how did that math, but mortgage backed securities hiding the underlying risk were enormous. My bank isn’t guaranteeing my cash with AI backed bonds disguised as Treasury bills.
And now the scary thing is we’re seeing a lot of that toxic waste debt all over again, AAA rated no less, with recent collapse of First Brands and the Tricolour mess right before it. It’s turning out there were huge derivatives and collateralized debt obligations underwriting a lot of subprime auto loans and it turns out, bad loans in other areas too. Not unlike Bear-Stearns back in 2007 but much higher scale. That stock market collapsing on Friday was mostly blamed on the tariffs as the well as AI bubble starting to pop, but I have to wonder if part of it was queasiness at seeing early hints of GFC unravelling all over again. At much higher scale.
I think “misallocation of capital” is different than size of the bubble, It s more like Capital going into building new housing vs. what it would have been without the bubble vs real resources going into data centers. This relates to “malinvsestment” theory of business cycles from Austrian economics, which is incorrect on its own
A holistic OpenAI requires too many chips and energy. The current OpenAI is an idiot. Specialized co will use less chips, less energy and charge a fraction of the price . AI brokers will get questions, do initial data analysis and relay them to the specialists, like a nurse who talk to u, before u see the doctor. People, students, commercial enterprises knowledge will grow exponentially. Cyber will protect them from blackmailers and thieves.
Very insightful article. The parallels between Dotcom and AI is obvious to everyone. But as long as they make money they are correct and the results give positive feedback to their reasoning.
I will not act like I know the finance and tech. If the global AI investment is roughly 2 trillion dollars, then from 2026 on-wards to 2035, every year, AI should return 200+ billion dollars to break even in addition to the operating costs. My math could be wrong.
(May be I am barking up the wrong tree…)
Famed ‘Big Short’ investor gives terrifying verdict on Trump hammering China with 100 PERCENT tariff… and issues doomsday warning to Wall Street
The famed investor, who is famous for predicting the 2008 financial crisis, has revealed the terrifying consequences of the latest trade dispute with China.
“Everything is dictated by the president and then executed by others,” said one White House official”
The administration needs AI to remind it not to get into a peeing contest with an elephant.
With SoCal it’s “everything Trump”/TDS on steroids.
Nothing different. Anyone remember pets.com from the 90s? Companies that don’t have the revenue to justify their valuations.
The scale is vastly different. Four chip makers (out of which only one actually makes chips) are $8 Trillion. That’s $1000 for every person on this planet.. That’s just 4 stocks with products not much different from before AI. And the deals they announce is like we put Bernie Madoff in charge of Enron.
Now we have Trillions of dollars trading at 100+ P/E and 20-30 times sales. This one is Hyper Bubble.
Admittedly, I DIDN’T read the entire article – just skimmed over it. I didn’t see this…apparently “Wall Street’s biggest fear was validated by a recent MIT study indicating that 95% of organizations studied get zero return on their Al investment. Why it matters: Investors have put up with record Al spending from tech companies because they expect record returns, eventually. Aug 21, 2025.” YIKES!!! 😬. And another source said the AI bubble is 17 times the size of the dot com bubble. DOUBLE YIKES!
Is it time to run for the hills??
We are at a precipice in more ways than one.
-Market cap concentration nearing pre-1970s crash out with similar P/E levels and highly sticky inflation
-Over-levered firms and governments that NEED the arrow to always point up
-ETFs now outpace the # of listed stocks, and of which, retail investors predominately hold…and those ETFs still have high exposure risk to the Mag 7
-Much of the Mag 7 growth hinges upon AI, which as Wolf mentions…draws immense demand upon not only electricity, but also have high water demand these GPUs guzzle for cooling, which will strain local communities even more
-Meanwhile, that same AI is intended to replace many entry level positions and other white collar jobs which are typically filled by debt-saddled recent college grads…
-Though on the lighter side…we may actually, finally…maybe, be 10 years away from non-commercial but viable fusion (aided by AI)…
-And quantum computing will blow everything we thought we knew today into irrelevance (whenever it happens)
Fun(?) time to be alive!
Hey, you Bozo – you had me fooled for a few seconds. 😜
I wondered why there was no box or rectangle around your response. At first, I thought there was a glitch. Turns out, someone was impersonating the Great Emperor! The palace has been breached!
Yield chasers making closed call ETFs a thing because they just refuse to understand it’s strictly worse than holding the underlying asset is a hell of a thing.
Great way to sum up the market. It could be bad, it could be good. Whatever happens happens.
I won’t lie. I’m concerned about the enormous energy consumption and whether it will cause more inflation and pollution. Both are negative for society.
“…whether it will cause more inflation and pollution”
Yes. There is no free lunch with energy.
DoE Sec Chris Wright said a couple of weeks ago that data centers probably need to bring their own power. I would agree 100%, and hopefully, this is the direction that he pushes the industry. The problem, of course, is that companies who build power only have so much capacity to build out and would be competing with general commercial / industrial / residential demand. Also, SMRs are several years out in terms of being able to even meet minimal demand. I’ve seen pictures of some newer data centers, and I don’t understand why their roofs aren’t covered in solar panels.
For anyone interested in peeking behind this AI bubble more, look up Ed Zitron, the guy has done his homework and have some great analysis of all the hoola behind this giant massive bubble. Karen Hao is also another good journalist that have done some good write up on this whole AI hype. Ed on his podcast or other’s show dismantling the true believers are pretty entertaining thing to listen to if you give 2 sXXt about AI hypetrain
From FT a month ago — a nice example of AI enthusiasm:
“BlackRock put $90mn behind Tricolor in 2021, touting it as one of the asset manager’s first commitments in its “impact opportunities” strategy and for its use of artificial intelligence.”
Nice! I’m sure AI is busy, working miracles all over the world.
Wolf, uncharacteristically chartless, is predicting that which his near-future articles of charts will prove to be, once again, correct.
Yes, when Nvidia joins the list of IMPLODED STOCKS. I can’t wait.
If Nvidia makes it into the pantheon of imploded stocks (at least -70% from peak), it would still be a $1.3 trillion stock. That’s how crazy this is.
While Nvidia may eventually implode from this, the upside is that they still actually make something and have physical infrastructure, like they did before the AI boom. Some of these AI companies are basically a pile of debt selling a dream. If that dream poofs, all you have left is a pile of debt and no way to pay it.
ChatGPT plans the perfect holiday … to places that don’t exist
Tech-driven itineraries may look appealing, but travellers have found their recommended destinations and activities exist only in AI’s imagination.
There’s not much the good folks of Gen Z will do these days without running it past ChatGPT first. When it comes to holidays, however, it pays to get a second opinion.
When Madison Rolley, a US travel influencer, was planning a two-week trip around Europe, she used ChatGPT as an unconventional travel agent. She gave the generative AI language model her travel dates and a budget of about $1,000 and it spat out a detailed itinerary of the best spots to visit.
But when Rolley, 27, turned up for breakfast at a photogenic café in Split, Croatia, she discovered a group of bewildered cleaning staff. The café was, in fact, a restaurant that only served dinner and would not open until that evening. “It was kind of awkward,” she said.
GIGO. AI is useless with bad prompting. Rolley needed to be more specific in her ask and not assume AI could read her mind and fill in the details for her. It’s a classic mistake of an amateur. AI is programmed to be a helpful intern, and being helpful to AI is returning exactly what the user asked for — whether it exists or not.
I’m taking a break right now after having spent nearly all day in ChatGPT as it coaches me carefully through building out a new website that uses custom post types and advanced custom fields. But then again, I know how to ask for exactly what I want.
Billions of Dollars ‘Vanished’: Low-Profile First Brands Bankruptcy Rings Alarms on Wall Street
The Trump administration made moves this summer to allow 401(k) plans to invest savings into the private equity funds that extend private credit to companies, raising the stakes even further.
The First Brands bankruptcy could amount to something of an I-told-you-so moment for the traditional bankers and private-credit skeptics who have long maintained that these upstart lenders deserve more scrutiny.
Before its unraveling, First Brands was seen as a success story. Founded by Patrick James, a Malaysian-born businessman, the Cleveland company expanded rapidly over the past decade by buying 15 competitors, including brands such as FRAM (air and oil filters) and Autolite (spark plugs).
Three blue boxes of Autolite spark plugs, with a spark plug lying on one box.
First Brands sells most of its products directly to the public on five continents and the rest to carmakers, which sell them under their own names. It says that it employs 26,000 people and that its sales totaled $5 billion last year.
fin twit is doomsday and panic. time to cut, cut, cut.
after a little “rug tug” today.
This article is by far the best take on the situation I’ve read so far.
I work for an AI company and when a group of us were told about the power requirements for a new DC facility I was dumbfounded. Megawatts is table stakes.
Seriously considering putting all my money in $CAT
Danger Dan, I encourage you to reconsider. Microbial degradation of scat produces byproducts that may disintegrate paper money. Also, you’d have some stinky bills. You might end up with some really rich fertilizer though—high in nitrates and phosphorus. That’ll be useful for a garden (if you’re into that) or can put it up for sale after depression 2.0.
/s
There may well be parallels between the fiber optics insanity and what is happening today. However, this is one big difference. Dark fiber that was buried doesn’t become obsolete for decades – if really ever. But data centers house equipment that becomes obsolete in months and years. The insanity of thinking that a data center full of servers and GPUs today will be worth anything in 10 years is being perpetuated by people that don’t understand how fast data center computing gets replaced and upgraded. This idea that we may overbuild data center capacity but its OK because we will need it later just like dark fiber buried in the ground is missing some key facts about the difference in utility and lifespan of fiber optic cable versus servers, GPUs, routers, switches, HVAC equipment and other data center specific components. Even the data centers themselves become obsolete as technology changes in how we cool and manage machines.
I was wondering about this, thanks. At least the canal and railway manias left a lot of usable infrastructure behind upon which subsequent productive economic activity could take place, as was the case of the fibre network left behind earlier this century. I doubt what’s left over after this debacle will prove to be as useful.
It kind of reminds me of the biblical story of the Tower of Babel …
There’s dark fiber at the edge of my front yard. As near as I can determine, it’s been there since the 1990s.
Although it’s just below the surface of the soil at the Arizona Slim Ranch, it’s popping up in other places. Along the edge of my next door neighbor’s property, it’s almost completely exposed.
Which brings up the question: Who inspects and maintains this fiber?
Well, the answer is: No one.
I’ve never seen anyone even bothering to take a look at it.
Optical fiber is very susceptible to humidity. if humidity gets to the fiber itself, its ability to transmit optical signals deteriorates. So all cables with optical fiber have elaborate protections against humidity. They’re also armored to prevent rodents from eating into them, etc. But over the decades, this protection deteriorates. If the underground cable is exposed to the surface, it is exposed to additional damage. Plus, there has been a lot of innovation in optical fiber and fiber cables since the 1990s. So it seems to me that the optical fiber cable in front of your house was written off a long time ago, abandoned, and left to rot.
AI will still be around doing its thing, which is predicting the next word in a sentence. It will not replace human beings who can actually think. But it is the belief that AI will allow business to fire most of its workers that has inflated the AI bubble. Yes, I have read that AI can do marvelous things with proteins and coding and fake videos. From my own experience I know that AI makes things up and gives false answers to simple questions. Who will pay billions of dollars to use such a tool is at present a question without an answer.
One thing I’ve come to realize is that CEOs and founders feed off from ignorance. Very few people understand mathematics let alone complex statistical models. The key word is model. These models rely on past observations to create output. There is no way AI or super Intelligence will ever create a new concept. It’s best to remember most great scientific breakthroughs were observations that deviated from expectations.
Lastly AI is being sold as a Utopian tool to boost efficiency and profits by removing the need for human labor. Let’s assume these so-called thought leaders are correct and more jobs are lost than gained. How will this unemployed population continue to drive the economy with a reduction in consumers? It’s an inherent contradiction that is being glossed over. It reminds me of the promise of drone delivery, robots and autopilot cars. Theoretically possible, but inefficient because there are too many degrees of freedom. Ignorance is strength when raising capital.
Sir, great point about the statistical models not being able to create new concepts. Spot on!
However, not many jobs actually create new concepts or new products and services from scratch regularly.
For example, the Nvidia chip designer most often reuses IP blocks and supporting code and only optimizes them in a new product. Doctors use tried-and-true differential diagnosis when trying to identify a patient’s malady. Let’s face it, much of our daily work is mundane.
There are very large opportunities for AI companies to augment and possibly replace some of the activities formerly handled by people.
I may be wrong, but in 10 years, many jobs will be very different from what they are today because of AI.
And the winner is China!
We will not sell them chips that cannot be made without their minerals! How could that have a happy ending?
What are the odds that they already copied everything and will produce them for less. That have a very long track record and they are good at it.
I read the whole damn article thank you. A three year old chip will be valued just like a three year old car driven 24/7 for three years.
Re: “And the winner is China!
We will not sell them chips…:
Interestingly, the CEOs of the big three: Nvidia, Broadcom, and AMD are all of Chinese descent. And all the chips are made in Taiwan by.. you guessed it… Chinese people. So let’s hope they sell is the chips.
Jensen Huang (Nvidia founder) was born in Taiwan, Hock Tan (Broacom CEO) was born in Penang, British Malaya and has no Chinese descent – he is Malaysian. And AMD CEO and Chair Lisa Su was also born in Taiwan. I don’t think any of these people who call themselves Chinese but I could be wrong – Certainly no mainland Chinese.
I think you are confusing ethnicity with geography. Ask AI if Hock Tan is of Chinese descent.
Ethnic identity: Chinese.
Place of birth: Penang, Malaysia.
Nationality: He is a naturalized American citizen.
And my ethnicity is German, French, Norwegian. That doesn’t mean I identify with any of those countries or policies of those countries. So I don’t really see your point.
Those that survive will learn to laugh by telling stories about how the ship of fools argued about which chatbot to use in order to find the best route to the sea of stupidity. Hey? Lets order some grubhub because we will be hungry by the time we arrive!
It’s the same as the internet bubble. Which was a generational change worth investing in for some companies, but not all companies. It will crash but predicting when is hard.
So everyone knows it’s a bubble, but the problem is how do you know when to get out? Bought a small amount of NVDIA in 2019 and I’ve been slowly selling to take profits since 2023 but man it just keeps going. I almost sold all my AMD last month too thinking we were at a top – and I just made ~30% in a few days. Every time I’ve sold its been the wrong choice. Yes it’s a bubble, but when to sell? Hard to say…
Yes. This thinking is very widespread. And no one knows where these stocks will go. If enough buying pressure persists, they will go higher. Lots of people are waiting for the moment to sell. I remember that common thought from 1999. Everyone knew it would implode, and some stocks began to implode in late 1999 (including Amazon), as people got out. But other stocks continued to shoot to the moon. Then in March 2000, more and more people thought it was the time to sell, and the dumping began. But there were huge rallies that pulled people back in and that caused people to regret having sold, such as in the summer of 2000, when Intel reached a new high that it hasn’t seen since.
Mr. Wolf,
This is what I appreciate about your views. You bring up points that often demonstrate an issue is not always 100% black or white.
People can make and lose money depending on when they buy or sell an investment. Sometimes those who buy and hold can make lots of money even when there are bubbles.
“A $500 investment in Nvidia stock at its initial public offering (IPO) in January 1999, assuming dividends were reinvested, would be worth approximately $1.88 million today”.
Ha! My sister bought 20K worth of NVDA after attending an introductory investors meeting with the CEO, years ago. Her price: $0.35/share. That’s 35 CENTS adjusted. Her genius was holding on to it all these years instead of cashing in when it doubled, tripled, 10X, etc. Right now, she has no financial worries. We’ll see about the future.
Just keep selling into the the rips higher. I’ve been doing it in one stock for several years now. I’ve unloaded 3/4 of it, but it’s still worth 2/3 of its original value.
Also can the power grid actually handle these new demands? And where is all the extra energy for these data centers coming from? Seems like energy could be a limiter for growth?
Yes, a limit, especially when the Interior Dept. just cancelled the Esmeralda 7 solar installation today which would have produced enough power for two million homes. But do not worry for they also auctioned off the largest coal lease ever, at 6 cents/ton royalty.
Please look it up, rare earth metals are found concentrated in coal ash.
Burn baby burn could be the next gteat investment opportunity!
It’s a lot of work. The issue is partly that while GPUs and computer components are tarrif free (hence why it’s the only part of the economy growing) that the parts for the electrical grid are not. So you have a frictionless product dragging an anchor through everybody else’s energy needs.
Like in oil/gas, the price drops are bad. But what they can’t say out load for Cultural reasons is that the price of drill pipe has gone way the hell up to tariffs so the break even price of domestic production has gone up.
The grid could handle it in the middle of the night when power demand is very low. But AI activity cannot be restricted to the hours of 10pm – 6am.
Why not? A schedule of very different fees for different time periods? Also, time zones. I know this is probably a primitive way of looking at it, and I’m not at all pro “AI will conquer the world”, but couldn’t this be done?
The training of the models is hard to stop or suspend and restart. It essentially needs to run full throttle for weeks.
Won’t we all be working from home, just turning on our AI work machines in the morning and going out to the pool to enjoy our coffee? Then get dressed and ready to head out to a nice restaurant for lunch. Coming home and putting on our virtual reality glasses for a little fun before ordering out for dinner and enjoying the evening with a few cocktails on the deck. Don’t forget to turn off your work machines.
only without the billionaires
Wolf, it’s pretty obvious this is a bubble, and you’re well aware of it.
When people start talking about “this time is different” or “this thing will change the world” or otherwise trying to make up reasons why things aren’t a bubble or can’t lose money, it’s a bubble, it’s going to crash, and it’s going to crash hard.
I really wish people (and more so investors) were smart enough to dump any investments idiots start saying that sort of stuff in en-masse, but we have idiots running the show.
There is an awful lot of trash paper out there. You are just another debit on somebody else’s balance sheet. AI is going to take all the wiggle room out of every transaction. Money is like a big wave, it flows in and then out.
What’s next?
Great piece.
I suspect many AI investments will yield little, and those who made them will get hurt.
On the other hand, those who pick the winners will be greatly rewarded.
However, if you doubt this, think back to the internet’s coming of age and the number of changes it brought about – just think of what Amazon did to retail shopping, how about the coupling of phones with cameras and the internet, Nvidia and video gaming…
Folks who invested in the winners were rewarded quite handsomely.
As AI becomes more intuitive and incorporates more real-time data and decision-making into its models, it will begin to replace humans in many areas. This is where the winners will be chosen.
Just do an internet search of companies with announced staff reductions due to AI, and you will see AI is not going back into the bottle anytime soon. The investors who pick these companies that can monetize AI to offer cost reductions will be quite valuable.
The equilibrium for me between risk and reward is about bet sizing and patience. I have picked my horses (future Amazons as I see it), and made bets I can afford to lose (and sit on through a decline and consolidation). But I want to have some capital in the game at this point, for possible eventual rewards.
The difference is that this time the companies investing huge amounts of money have operating businesses that throw off huge amounts of cash, and very strong balance sheets. If AI doesn’t work out, they can survive.
However, I agree that the stock market would not be very happy about this, and the valuations of companies like Meta, Apple, and Amazon would plummet.
The stock indices are being held up by a MERE 7 STOCKS as many of the other 7,000 companies are struggling to survive.
Just like Alan Greenspan solely caused Black Monday, he also caused the dot.com bubble and burst. Greenspan didn’t ease monetary policy until November 2002 coincident with the bottom in stocks.
Powell is just facing a downswing in R-gDp in the 4th qtr. and beyond. The economy has been kept afloat with the draining of the o/n rrp facility and the diversion of TDs into DDs.
Inflation should fall in the 1st qtr. of 2026 supporting the U.S. dollar’s exchange rate.
There are no surprises.
Alan Greenspan certainly did not cause Black Monday back in 1987. What caused it was rising interest rates in Europe and a sudden shift away from the US stock markets as is very clear from history.
Sure did. Legal reserves dropped at the fastest rate since the GD.
The Greenspan Fed began cutting rates in January 2001. By January 2002, the FFR had been cut from 6.5% to 1.75%. The Fed cut another half point in November 2002, before hitting the cycle low of 1% in July 2003. The Fed was easing monetary policy long before November 2002, so if that was the bottom in stocks it was not coincident with Greenspan starting to ease.
At the height of the Doc.com stock market bubble, Federal Reserve Chairman Alan Greenspan initiated a “tight” monetary policy (for 31 out of 34 months).
No, those rate cuts were behind the rate-of-change in monetary flows.
Interest is the price of credit, not the price of money.
So basically you’re making up your own definition of monetary “easing” as having something to do with the rate-of-change in monetary flows. Under your definition, the 11 Fed cuts totaling 475 basis points in 2001 were not “easing” even though they were reported as such at the time. However, the twelfth cut in November 2002 was the start of “easing” under your personal definition (“Greenspan didn’t ease monetary policy until November 2002 coincident with the bottom in stocks”).
If you have a different definition of monetary “easing” from what is commonly understood, then you need to explain what that definition is. Unless, of course, your comments are only meant to make sense to yourself.
I think many believe ChatGPT and other LLMs are where it is at, which does make sense given that is the consumer exposed aspect of it. That is neither the largest market for AI, and perhaps never profitable, and clearly not the most dangerous. LLMs are not without exposure as they can either accidentally or purposefully given misleading or bad information, but that can be said for almost all media.
This is chat gpt answer to the economy
In the first half of 2025, AI-related investment accounted for 92% of U.S. GDP growth, even though it made up just 4% of total GDP.
This staggering imbalance is driven by a massive surge in capital spending on AI data centers and infrastructure, especially by tech giants like OpenAI, Meta, and Microsoft. Without this AI-driven investment, GDP growth would have been nearly flat—just 0.1% annually.
So while AI’s direct share of GDP is modest, its impact on growth is disproportionately large. It’s essentially propping up the economy right now.
Historically, a well diversified portfolio (e.g S&P500) could ride out bubbles with a manageable 30% down turn. This time will be different, I fear. With Nvidia alone being 8% and technology being 40% of S&P500, the AI bubble collapse will cut much deeper. I could see S&P500 down 60% to roughly where it was in 2017. That will be a generational buying opportunity if you have any cash left.
I am really curious. What stocks/ETFs would you buy if the S&P goes down by say >40%. Index funds with good dividends comes to my mind. Stocks like Exxon, coke, or walmart dont go down most of the times. What individual stocks would you buy?
wonderfully well written (the three “w”s)….. thank you… I cut and pasted it and will send it along to others.
Bots do that all the time! They just don’t say it
Trump’s new 100% tariffs on China triggered an $18 billion crypto sell-off
There’s some really weird stuff going on, in case you don’t know
No, nothing weird. Stocks and crypto sold off, and bonds rose, when Trump threatened to impose additional 100% tariffs on China. This happened before. Don’t you remember? Happens like clockwork.
The tariffs were just an excuse for Binance to cause market wide liquidiations and flush the market so they can pump their capital in the next three months.
This wasn’t a surprise at all. What was surprising to me were just how many idiots we’re trading on margin without have spot positions.
A big part of this growing stock market bubble might be tax avoidance. Lots of folks are sitting on huge unrealized gains in taxable accounts, me included. You might think the market will go down, but do you have sufficient confidence to pay a 30% tax on the gain? If it’s not clear, tax avoidance incentivizes people to let it ride, particularly when governments have established a pattern of protecting markets and inflation looks to run 3% or more for years.
This bubble was created by monetary and fiscal policy.
The U.S. has a small fraction of the global population. And the AI-driven potential for GDP per capita growth may not be in the U.S., but elsewhere. Take for example an annual check-up with a primary care doctor, blood work, follow-ups, maintenance of individual health records, etc. The whole process (including training of doctors and nurses and medical techs) is ridiculously inefficient in the U.S.; replicating it across the entire world population would seem like a non-starter. This is one of N examples. Sure there is a bubble, but on a generational basis, ~25 years, what is the AI market opportunity globally?
I don’t know if anyone mentioned but there is a big difference between the Dot-Com era and this bubble and that is the Government budget deficit. Back then we were technically in a surplus (the famous “Clinton Surplus”) which, according to some commentators especially on the MMT side, triggered the bursting of that bubble where currently we run at 6% or more (Wolf can be more precise about that) that is a lot of money sloshing around looking for a place to go…
Wolf,
Many thanks. Had no idea how nutty this bubble was until you mentioned GPU’s being pledged as collateral. Wow, lenders mainlining stupidity steroids.
Unlike the bubble of a quarter century ago this AI version will leave us very little residual value. Power generating infrastructure of course will be a benefit. Nothing else. LLM’s are this era’s poster child for GIGO. OpenAI eventually devolves into ShutAI. The mania implodes sooner rather than later. We shall see.
What are all the ruses in the last couple of years? The trigger words that set investors going crazy.
Spacs
Disrupter (WeWork, Pelaton, etc)
NFTs
Bitcoin
Anymore?
I feel like AI is the newest and biggest
WeWork, SoftBank, NFTs, FTX, SPACs, the economy just jumps from bubble to bubble. If you just keep jumping from sinking ship go sinking ship you’ll never drown.