Is the A.I. Boom Turning Into an A.I. Bubble? | Hacker News Reader

Is the A.I. Boom Turning Into an A.I. Bubble?

newyorker.com

113 pointsFinnLobsien112 comments

Open articleView on HN

> The global pandemic will tank the stock market

> War in Ukraine will tank the stock market

> High interest rates will tank the stock market

> Tariffs will tank the stock market

> IA will tank the stock market <- We are here

All those statements made sense to me at the time. And I have no doubt that one of these days, someone will make a correct prediction. But who the hell know what and when.

Diversify, be reasonable and be prepared for it to happen someday. But freaking out with any new prediction of doom is not the winning strategy.

Ai is the stock market so there is that problem

Trump election will tank the stock market - Many nobel prize economists and experts .

He did though. The S&P 500 fell 18% around Mar/Apr. That's unprecedented for a US market experiencing no external shocks.

And the US dollar is still down 10%.

> The S&P 500 fell 18% around Mar/Apr.

where ? i don't see it.

oh, wait and see, my child, this is merely the beginning of the story.

Yeah, the way I describe it is: rich people are rich and the money has to be put somewhere.

And rich people aren't getting richer through smarter investments, but by wealth accumulation on existing assets that people need.

Being able to hold on to wealth is an underestimated and undervalued ability.

It's actually pretty easy to lose money. Everybody is happy to help you part with yours.

I really want to believe this, but I have a hard time seeing this applying to the ultra-wealthy nowadays. There are far too many strategies to hold on to your wealth or to get it back.

Aside from lottery winners whose assets start liquid, are there examples of people in the last 5 years with over $100 million in assets that permanently lost most of their wealth without doing something really stupid?

> It's actually pretty easy to lose money.

By what mechanism / who exactly? Or referring explicitly to the ultra wealthy? When I think of the normal wealthy, the 1-20 million camp, can't they just stuff all of their money into index funds for the last 20 years and everyone grows 10%+ a year (most years)? That's what the few multi millionaires I personally know have done -> step 1 put money in e.g. Vanguard, step 2 maintain job and spend based on income, not accumulated wealth, step 3 do nothing as wealth builds itself. Once they get past 2 million, I think they are making 80-200k per year at capital gains rates and maybe don't even need to put more in, just ignore it. I don't personally have that situation though so maybe I'm glossing over details or have it wrong.

Earning $1mil and stuffing 100% of it into the stock market is a risky move.

You lay out basically the situation I found myself in. But I set aside some for a future house down payment (less risk for that chunk), future wedding (less risk for that chunk), and even ignoring that, most advice wouldn't say to stuff the remaining 100% into markets.

So maybe 60% went into markets. With hindsight it oftentimes sounds like a good idea to just drop it all in the market. But life is uncertain and crystal balls are in short supply... you would feel stupid if you made $1mil, put it all in stocks, and life came out of nowhere and took a dump all over you when the markets tanked.

Not saying I'm in a bad position, but its nowhere near "1mil at 10% per year (which is in itself kinda a wild base assumption)".

TLDR: people have a rosy unrealistic view of how this works out.

Most people earn money over time. Putting these savings 100% into the stock market as you earn them is pretty reasonable. It’s only a sudden cash windfall that presents problems. The fix for that is to put money in 1, 2, and 3 year treasuries, and invest that money into equities as the treasuries mature.

That's why the first million is the hardest since it's working capital for most working folks. After that, it can stay invested and compound.

From 2004 to 2009, the inflation rate was less than 4% each year. Gold was $443 an ounce in March 2005 and $975 an ounce in March 2008. Yes rich people had to put their money somewhere as banks and then corporate America started to collapse during the surprise crisis, they took them out of equities for banks offering subprime mortgages and put them into assets like precious metals and the like.

Also in 1999 MSFT was $57 a share. In 2009 it was $16 a share. It cracked $57 again in 2016. 17 years to go sideways.

Cisco never reached its 2000 peak again. Of course companies like Pets.com just went out of business.

To be fair, there's a subtle but important difference to:

- we're going into the next Great Depression (a once in a lifetime occurrence)

- a small subset of stocks (that happens to make up a huge portion of the entire equity market in the US) has extreme PE and PEG ratios and will pop (which happens every few years)

I think your point largely stands for both cases, but it's important to delineate them.

If you're preparing for a Great Depression - you're likely only going to be right by coincidence. If you're preparing for a stock bubble to pop, at the very least, you've got better odds.

All of these did tank the stock market, but it also recovered quickly.

Therefore the dips were an aberration - they should not have happened. Recovery is a sign that the previous dip was an incorrect prediction by the market.

No, the dips resulted in government interventions--like the Fed raising liquidity or Trump removing some tariffs--which then caused stocks to rally again.

The term "Fed put" is decades old.

You present these as straw man exaggerations, but they were real things that happened.

> The global pandemic will tank the stock market

It actually did screw up the economies of a lot of countries and companies and it messed with stock markets as well. It increased debt worldwide, it increased unemployment that took years to recover.

> High interest rates will tank the stock market

There is a well established relationship between lower stock market returns when there is high interest rates.

> War in Ukraine will tank the stock market

It significantly hurt the Russian stock market, at least the sanctions did. But war tends to be a stimulus for the economy as long as it isn't too large of a war.

Yea, after the pandemic I've stopped making any predictions about the stock market and bubbles.

It's clear we're living in an illusion, but I'm pretty sure there are enough people invested in that illusion that it won't stop until it is no longer physically possible to maintain. I'm increasing convinced that when whatever dream we're living in ends, it will end catastrophically, but I'm not even certain I'll live to see that happen.

All it takes is a couple hours of consumer sell-off to trigger a complete collapse.

Very true, and the hedge funds and institutional investors will have game plans for this, some will work, some won’t.

Retail investors, as usual, will be worst hit (I say this as a European with no stake in US stocks).

I believe their plan is to offload the risk to the chains. I've been asked by several firms if I would be interested in joining them - so I assume that want to lock up their assets on the chain or want to offload risk to it. I'm more keen on the latter.

Still - nobody make any sudden moves and we'll ride out what remaining value we can squeeze.

I think this expresses how many of us feel very eloquently.

See the documentary Hypernormalization for a more in depth look at this feeling

A great work of art. One that identifies this spectrum of feeling. But one that should be taken with a grain of salt, a subjective arrangement of visual material as captured by the BBC.

It's less of an illusion but more that the reality is separating into two, one which we live and breathe as day-to-day working class/consumers and another that caters to the market makers of government & private sector.

https://archive.is/bxEtW

All those events don't tank the stock market because they were based primarily on fear.

What will tank it are lies based on greed.

> Dot-com: overvalued companies with no revenue

> Mortgage: banks lying/lending about credit scores

> Covid: continuing online & EV trend, meme stocks, SPACs

> AI: continuing scaling laws, high ROI?

Inessential Affluence?

As others have said, we did see effects, but a quick recovery of many of these events or the full scale of damage still isn't known and quantified. We definitely are in a bubble, but like you suggest, diversify. Don't stop investing, just be smart. Market can remain irrational longer than you can remain solvent. Time in the market also beats timing the market. And lastly, it sorta isn't surprising that things quickly recover. Tons of people's retirement accounts are injecting money into the stock market regardless of what is going on. We will likely never see long term significant drops just due to the fact that we've devised a system of a non-stop garuntee stream of money into it.

The New Yorker also has a vested interest in making people freak out about AI. If anyone with a prompt can churn out something that passes as a New Yorker essay, the monopoly on tone, cadence, and authority starts to crumble.

Kind of a weird post given that the pandemic did tank the stock market (-35%), interest rates and the war in Ukraine did tank the stock market (-28%), and the tariff announcement did tank the stock market (-21% and we have yet to see the full fallout).

> Diversify, be reasonable and be prepared for it to happen someday.

Indeed, but even with something like SPY, there’s quite the concentration in tech:

    Top 10 Holdings (37.99% of Total Assets)
    NVDA  8.07%
    MSFT  7.37%
    AAPL  5.77%
    AMZN  4.11%
    META  3.12%
    AVGO  2.57%
    GOOGL 2.08%
    GOOG  1.68%
    BRK-B 1.61%
    TSLA  1.61%

Now that’s intentional as it’s market cap weighted. But the investing world is in for a rude awakening if things start to pop.

Just buy $RSP or a similar equal-weightage ETF.

Those techniques rarely worked in the past as the broad stock market retreat usually affects stocks across the board. The best diversification is into uncorrelated asset classes. But there is some art in determining what's likely to remain uncorrelated. Bonds and gold are usually decent guesses.

They certainly solve the stated problem of concentration in a few tech companies. But I agree, non-tech companies are overvalued. Bonds are terrifying, they’re a bet against inflation and in favor of the US dollar. MLPs and oil companies are a nice place to stash money because they aren’t correlated with tech. MLPs flow about 7% tax free yield which isn’t bad. Bitcoin and Gold are good hedges against inflation.

We need a etf that weights with the log of the market cap

I don't think putting all your net worth in S&P would be considered diversified.

With an MSCI world, those companies would drop to ~22% exposure. Throw a bit of real estate, more exposure to your home country if you are not in the US some real estate, some bonds and you can make it drop to <10%.

I'd argue only four of those are purely tech companies (NVDA, MSFT, AAPL, AVGO).

AMZN does tech stuff, but also retailing, grocery, logistics, media and more.

META and GOOG are advertising businesses.

BRK is a basically a holding company, its businesses are in a wide variety of markets, mostly non-tech (and it owns public equity too).

TSLA is a car manufacturer.

The primary harm of a bubble is *not* a crash in equity values, it is the misallocation of capital. The worst outcome would be for the misallocation to continue due to the intervention of asset owners with the most to lose who are also in control of the state.

All of the events you listed have had significant economic effects and required massive intervention from the state to buoy asset prices. The longer this continues the more our economy becomes geared to producing "value" for this small, and shrinking, group of owners at the expense of everyone else.

The way I see it, prices are too high. War is one thing, but it's abroad and the restrictions have been minimal. But interest rates went from <1% to 4.5% without valuations going down or profits going up. This means that investors are pricing companies as if though interests rates can be lowered. I don't think they can, because inflation in the US already quite high despite the 4.5% interest rates and with these tariffs upon that, it probably can't be lowered. For this reason I think companies should be priced using the 4.5% interest rate as a best guess, maybe 4% works, maybe 3.5%, but they're current priced as if though the interest rate is is <1%.

Tesla has a P/E ratio something like 26 times what other well-run automotive companies have. Boeing has a price like Airbus, despite no profits, etc.

Commercial real estate prices have gone down, and that's reasonable, but you've had both the interest rate increases and this WFH+hybrid remote thing becoming common, and since the interest rates have gone up from such a low level I think this is still overvalued, because 1% -> 2% should in theory mean halving the value if the rent is constant, and it went <1% to 4.5%.

I think it's a miracle that there hasn't been a crash. I wonder what weird things have been going on that have ensured that there hasn't been one yet. So I think your perspective is strange. The situation is absolutely crazy, and has been for years, but that doesn't make it not crazy.

> This means that investors are pricing companies as if though interests rates can be lowered. I don't think they can, because inflation in the US already quite high despite the 4.5% interest rates and with these tariffs upon that, it probably can't be lowered.

This is why Trump is replacing people doing the statistics with people who very publicly say that they will print what Trump wants. He really wants to cut rates, and I think he will eventually get his way.

If employment is down though, then that would be justification for lowering the interest rates. I think it's more a matter of maintaining appearances than something strategic to make the, by Trump desired, low interest rate economy possible.

Even though he wants lower rates, rather than saying 'the Fed is prioritizing price stability over employment and that's wrong-- we can't waste American lives, they need to have jobs' he's denying the numbers despite that they could in principle allow him to do something he wants.

Maybe other things are signs of knowing these problems, maybe this demand for investments from Japan and the EU are motivated by concerns that stock market and to some degree commercial housing valuations are unreasonable, but it's not certain that Trump or anybody else is fully aware. This isn't complicated stuff, and I've even seen newspapers detail ideas like this, maybe toned down a bit, but it's hard to know who understands what.

- The pandemic did in fact tank the market - Wars never tank the market and they jack up the military industry so no one would have said that - High interest rates did tank the market once last year already and in the past because it provides for risk free gains - Tariffs did tank the market so Trump played the reversal games - Will AI tank the market? It will if it creates unemployment.

Diversifying does not stop the tanking. It will reduce the risk related to poor choices all at once

One interpretation is that asset owners as a class have complete control in our system and they will always be made whole no matter what happens.

This may require extracting additional rents from consumers, from workers, from renters, from debtors (including the government) but whatever changes have to be made to protect asset holders will be made regardless of the cost. An example of this in action was the collapse of SV Bank where the rules of our federal deposit insurance program were rewritten on the fly to protect the depositors. Imagine having an insurance policy, incurring an uncovered loss, and then compelling the insurance company to retroactively rewrite your policy to cover the loss!

This interpretation seems accurate to me. I think the big concern is what happens once the asset and passive wealth class have squeezed the working class dry. And the governments don't have enough wealth for a welfare state solution. What is the point of working hard and smart and trying to innovate as a young person without family wealth when you can't even achieve basic economic security.

There was an article I saw this morning saying only the very top of the S&P 500 are the stocks showing substantive growth in recent years, companies below that have been relatively slow to show growth. Additionally, since the pandemic and Ukraine war started global cost of goods have been rapidly increasing, much faster than they should be. Now with AI, the market in the US is losing a lot of jobs - both entry level and above. The latest US job numbers were so terrible Trump fired someone to try and cover it up.

I'm not sure what else needs to happen to show the economy has been doing poorly for all but the richest segments. The return of a blatant and severe caste system and mass starvation?

Some of those things did happen though.

I think there's an incorrect valuation by looking at where things are today. I mean Black Monday, the 2009 housing crash, DotCom Bubble, and others were times the market did tank yet we've since recovered.

So how are we measuring the accuracy of those predictions? From Jan to April the Trump admin was announcing tariffs. VOO's (S&P500) lowest price this year was on April 8th at $456.74 and on Feb 19th it was $563.67. We see similar patterns with covid and invasion of Ukraine. Do we consider a 25% reduction "tanking"?

I agree that with enough time that everything will work itself out. But I do not think that this means we should ignore or downplay damage done in the short term.

> The global pandemic will tank the stock market - the market did crash and then Fed stepped in. Rates were cut and Fed

> War in Ukraine will tank the stock market - Market did go down by -30% between Jan-Oct 2022 and went nowhere for sometime after that.

> High interest rates will tank the stock market - Impact for this remains to be seen. Even during 2008 the high interest rates risk persisted for couple of years before the crash. So, I'd give it more time.

> Tariffs will tank the stock market - Did we not see a 20% drop before tariffs were put on hold for 90 days?

> IA will tank the stock market - I don't have much conviction on this one.

It is true that no one knows when a prolonged market crash like 2008 will happen. Maybe never. Government has figured out that Fed intervention can help the market stay afloat. So, maybe unless Fed doesn't step in for a long time these predictions will come true.

The flip side of this is that investment returns of a diversified portfolio (net of inflation) is slowly going down. The choices are to either concentrate or find alternative investing vehicles. That is one of reasons private equity is out looking for alternate income sources like buying up houses and hiking up the rents etc.

In 2000 and 2008 how many people falsely predicted the top before it actually came?

I’m too young to remember the naysayers, just the hype.

You can watch the Big Short. Michael Burry started paying out premiums on CDS in 2005. The premiums were huge. He got into a fight with his biggest investors and then had to freeze withdrawals. So, that he could still pay the premiums on the CDS. Did he falsely predict the top?

Right, but was he an outlier or part of a chorus?

And how many people before him incorrectly predicted the top? For how long?

> But who the hell know what and when.

It’s not hard to predict that events will happen. It’s hard to do so with more precision than the next guy.

Something I learned in H2 of 2021, when I nearly went broke betting on a correction. Wasn’t wrong, one did occur. Just failed to realize I couldn’t figure out exactly when.

All of your examples did tank the stock market. At least in the meaning a 20+% drop over a time period in the order of months of less.

They are however not examples of long depressions. Which I think is reasonable to expect we'll see less of, given that more and more regular people feel the need to put their money in the stock market. There is simply no alternative. And every time we reach new heights. This is also expected.

Will this go on forever? Probably not. But it won't look like the 1930s. The stock market is a crowd, not a science.

Turning into? It already is. I wish I had a hedge fund to direct towards shorting AI-driven companies.

It's helping me find things, understand things and do things. I'm sure some of those startups will figure something out that makes sense. And eat the rest.

This was the story of Amazon in the .com bubble, but buying Amazon in 1999 was not a good deal. Even if you know who the winner is going to be, you can still lose.

Please share examples. Real life game changing examples that were not possible before.

I get this sinking feeling that this "bubble" may never burst on its own, since the political environment now is different from the 2000s. Inequality is much higher post-COVID, with the ultra-wealthy sucking up the assets of the middle class. Those assets are going into the stock market for AI companies with the objective of eventually replacing people so they never need to pay the working or middle class ever again. Even if the AI doesn't perform quite at the level of normal employees, there's very little competition in the market anyway.

I'd be happy for someone to tell me I'm wrong. Otherwise it will take something dire to break this cycle.

I saw a figure yesterday showing that current AI data center investment exceeds all consumer spending in terms of GDP. This is important, because consumer spending is normally 2/3rds of GDP. This means consumer spending isn't great, and data center investments probably can't sustain this pace forever. It's also not clear that there will be a return on these investments (one that flows to the specific companies seeing the investment.)

I think the figure wasn't that data center capex exceeded consumer spending, but that data center capex contributed more to GDP growth than consumer spending did. Still really really bad and indicative of a bubble, but an important difference.

TINFOIL HAT SAYS: Since you used the P-word, consider the 2008 election and the possibility that a similar bubble-bursting occurring at the appropriate time to maximize chances to flip Congress in 2026.

You think the kind of people propping up the AI bubble want congress to flip?

I've been trying to figure this out, like I think the 2008 bubble burst for a lot of reason but it was basically a ponzi scheme that ran out of new people to buy houses at constantly inflated rates.

Bernie Madoff got caught when 2008 happened and he didn't have enough money to pay back investors, if 2008 never happened he maybe keeps going.

For Elon Musk, and maybe why he bought Twitter, he has to stay popular enough that enough people believe his companies are the best investment available. If someone surpassed him, and Tesla didn't seem like the place where robots or self-driving might happen, then people would probably move away.

For AI, maybe something similar? People would have to start deciding that investing somewhere else is better than investing in AI and let the AI companies start to fail.

Presumably the cause of a bubble is just too many people thinking AI is the best place to invest their money? And the bubble will pop when they decide to invest somewhere else?

9 of the past 5 market crashes have been successfully predicted

they wouldn’t have happened if they were predicted largely by the market.

With MSFT being worth 4 trillions right now, it looks like a legitimate AI bubble. Those valuations have entered the realm of fantasy.

No matter how right the AI crowd is, elements of it are/will be a bubble.

No matter how right the bubble crowd is, the market becoming irrationally exuberant for a brief period of time does not invalidate the technology or the rapid change we'll see as a result of it.

For sure, the Dotcom crash didn’t mean the web was bad technology.

Not technically bad, anyway. Just… socially destructive to the fragile web that held society together.

Yeah, this is what all the bubbleists miss. We got FAANG++ out of the last bubble, and they literally rule the world with the tech that was promised during said bubble. Catsdotcom and dogsdotcom failing had 0 impact on the tech itself.

It's the same today. A lot of the hyperVCfunded startups will fail, without a doubt. But the tech giants will get their money from this tech, and it will be ubiquitous in ways we can't even imagine now.

I'm not sure that the dotcom bubble leaving behind the web means that AI (and by AI I mean LLMs) are going to be transformative in the same way the internet was. Just because it happened once doesn't necessarily mean it's going to happen again.

To be clear, I don't think LLMs are going to vanish, there's clearly some things they're good for, but there's also some really big differences between the AI bubble and the dotcom bubble. People are very skeptical and worried about AI, they (the general public) aren't really using it for anything more than search, there hasn't been any clear economic data indicating that it improves productivity in a meaningful way, and a killer app hasn't really emerged that isn't a free chatbot. Plus, the majority of the money is concentrated in the same 5 or 10 companies just passing it back and forth between each other before eventually handing it over to NVIDIA. Maybe I was just too young to pay attention to the dotcom bubble, but the vibe seems completely different.

I think the comparison between AI and the web is interesting. I still feel like AI has this assumption that it's going to surmount a gap that it may not surmount.

The internet basically just got gradually better at all the things around it. Better graphics, faster internet, better programming languages, bigger hard drives.

But we still don't know what LLM's are. They're maybe just a feature, not a whole suite of technologies creating an ecosystem, the idea that they'll make the leap from Fancy Lisa AI Companion to colleague is fanciful. I don't think there's a direct line that shows that up and to the right automatically means superintelligence.

Of course it's a bubble. It's only about 20% as useful as the claims driving the current irrational exuberance. All it can do is generate pictures and text, and we had those _coming out our eyeballs_ for at least a decade. Prior to generative AI, each of us already had more access to images and text with which to stimulate ourselves than we could consume in a lifetime.

When did we forget that discovering "truth" via symbol manipulation is a fraught proposition at best? It was in the 17th century that Leibniz proposed that encoding logical propositions into a propositional calculus would allow all intellectual disputes to be resolved mechanically.

"For it would suffice for them to take their pencils in their hands and to sit down at the abacus, and say to each other (and if they so wish also to a friend called to help): Let us calculate."

The original AI bro! Any day now...

I've been thinking lately that the real value of a piece of code is that there is at least one human alive somewhere supporting it. You remove that, and the value proposition gets extremely shaky. Folks are going to have to learn this first hand as their brain becomes full of echoes of LLM output, rather than the output being an echo of some brain process (you know, _actual_ intelligence).

But sure, if you can convince enough people that you've invented a real magic 8-ball, you might be able to convince enough of them to shake it for the rest of their lives. Me, I'm not convinced that the marginal value of "new" text and images is there.

https://archive.is/bxEtW

People have been saying this at least for the last 10 years. I think if advancements stop being made at a rapid clip, the bubble might implode. But every year it seems like its a whole new ballgame.

When LLMs start playing the stock market autonomously, then we should start worrying.

Oh, man, there have been 10 years since ChatGPT made the news in 2022?

How time flies!

People have been predicting an AI winter ever since DNNs hit it big, and that was 2012/2013 or so. And then every year since 2015 they've been saying "this is the year the AI bubble will finally burst!" And then repeat that the next year.

ChatGPT was just another big leap in a bunch of big leaps that occurred in the last 15 years. You are being disingenuous in pretending that people didn't see AI as being hyped before ChatGPT came out.

The finance bubble is pretty much a post-ChatGTP phenomenon.

There was some hype about AI before it, for completely unrelated products, with varying levels of real value behind them. What bubble there was about those has already popped long ago, missing the headlines because of the crypto hype.

If you've been with HN for awhile, or you search the archives, similar topics to "Is the AI booming turning into an AI bubble" kept popping up long before CHatGPT came out. And this was DNNs, and it was same people (e.g. Hinton and his students) who were at the forefront of that also. And no, their previous projects didn't go bust.

Well, ok, good point.

There has been people claiming that "their work will stop scaling any time now" for some 12 years. (To be fair, it can't scale forever and seems to have stopped scaling by now.)

That is a hype bubble. I maintain that it's not the bubble the article and everybody is talking about. People are talking about the billions (but the plural doesn't do it justice) running around looking for a return. A lot of their earlier projects weren't even supposed to make a profit.

> There has been people claiming that "their work will stop scaling any time now" for some 12 years. (To be fair, it can't scale forever and seems to have stopped scaling by now.)

Their work kept scaling, they just branched out into generative AI to keep going. Now, eventually they won't be able to scale up or branch out anymore, and I guess your bet is that this will happen before AGI is reached?

AI was over-hyped. I remember how "AI" felt like marketing bullshit at the lab when I was there between 2015 and 2020.

Generative AI took it at the next level though. There is no comparison.

The meaning of AI has brutally changed. It used to refer to ML, deep learning and all these things. Now, it is a synonym of generative AI.

The original paper where they took wikipedia as a corpus and generated a character by character generator (the precursor to the token-based transformer stuff) came out in 2012 (I think?). It isn't unrelated to the AI that was hyped up then, it is and was a direct evolution of the same ML we were talking about today (Hinton was an author on that paper).

Yeah yeah, there was also SCIgen in 2005 (when 3 MIT students submitted a fake paper and got it accepted).

I believe the difference is scale. It's not just Wikipedia, it's the whole web. Billions and billions are burned (and it seems highly unsustainable, where previously, stuff was made to work in public labs - possibly with things like Grid5000, but that's nothing compared to what's training and running LLMs today). One fifth to one third of HN's front page is AI. The thing is mainstream and omnipresent (before 2022, that stuff was mostly people working in research and tech companies, the informed general public might have heard of ML).

Around 2015, we were there: https://xkcd.com/1425/ (explain xkcd says September 24, 2014)

I remember reading this two years ago. And one year ago, too.

Yeah, I mean there's a chance they're wrong and some chance they'll eventually be right. At this point it's clearly impossible to say for sure what the outcome might be. There were plenty of people in 2005 saying there was a housing bubble in the making and plenty of evidence that they were wrong. Were they right but too early? Were they wrong on fundamentals but happened to have guessed right? We could be seeing something similar here where the nay sayers are right for the wrong reasons, or they're simply haters whose desires may align with the eventual outcomes. Or not.

There were also people who said in 1999 that there was an unsustainable bubble of "internet" businesses. And they were right. The bubble burst, and the internet still went on to change the world

No doubt there's a lot of stuff right now that's completely overvalued, and companies that try to do things that the technology just isn't ready for yet. Those will eventually be culled, either one by one or all at once in a big burst. But for the way AI changes industries and our lifes that will only be a speed bump

This is essentially what I’m saying. More succinctly, you can be right or wrong about a macro trend but your correctness is probably more time-boxed than people expect.

The market can remain irrational longer than you can remain solvent

Bubbles can last much longer than a couple years. The housing bubble in the US that finally popped in 2008 was building for at least a decade.

I remember reading this for years before the dot com bubble imploded, years before the housing bubble impoded, and months before the market woke up to the fact that COVID existed. (At least governments decided to print that last one away.)

On a more serious note, the fundamental issue in the first two crashes was that the markets stopped being able to rationally price things. My view is that this doesn't always mean there will be an immediate crash. It just means that when the inevitable downturn does happen, nobody will be able to figure out the correct pricing for anything -- and so a correction turns into a catastrophe.

Turning into?

Vibes have told me it long ago became one.

Of course it is.

Investors can't set realistic expectations for AI, so an ocean of capital is chasing a dream. When the rent comes due, the quick will cut and run. The rest will turn to extraction. If you want a front-row seat to the clusterfuck that is aligning LLMs with product placement or selling users' conversations to advertisers, you're in luck. Grab some popcorn.

"In the A.I. economy, it seems possible that many of the rewards will go to top firms that can afford to build and maintain large A.I. models..."

This is a flawed (and common, still, somehow!) idea of where value is within AI. The gigantic models are commoditizing rapidly. They practically incinerate cash.

There is real value being created at the layer where users can use a product they can actually trust. LLMs are certainly not that.

Our system, Aloe, is model agnostic - so when next month's model comes out we just get better. And we already beat the pants off the frontier models in capability, spending a tiny fraction of a percent of the capital to build the system.

The existing concentration of capital into big LLM companies is indeed a bubble. Eventually some of those VCs will wake up and invest where the value is being created, I presume.

A Tobin tax would do wonders to help decrease the delta between boom and bubble. But with the international economic order being shredded as we speak, that possibility seems further away than ever.

Over-investment in a new technology seems to be a common thing. It happened with the Internet and railroad. A bubble forms which pops. But the process leaves behind valuable infrastructure.

It's too bad the infrastructure in this context largely becomes obsolete in under a decade. Optimistically, we'll have a lot of data centers with a lot of generalized resources, but I have a feeling the margins get pretty thin on operating/maintaining those facilities absent high demand.

To analyze whether AI is a bubble you need to analyze its value (like value investing).

I find this article hollow because nowhere does it analyze the value or potential of AI.

Considering that the average Joe is talking about profiting from AI, then yes, AI boom becomes AI bubble.

was cloud computing ever a bubble. yes. did it stabilize. sure. but its a powerful utility that shifted and created new value. the only speculation is where and by how much.

the rush of new tech is always confusing, new tools requires new skills and time to find its place in the world.

The AI bubble is, IMO, more like the DotCom bubble, where every startup believed having a modern website would return investments many times over. The only companies becoming rich are the ones selling shovels.

Some articles arguing that's already the case (and why):

- https://www.wheresyoured.at/ai-is-a-money-trap/

- https://www.wheresyoured.at/anthropic-and-openai-have-begun-...

For me, the only consolation is things seem to be moving fast nowadays so the inevitable collapse must be close. See what happened to Github's CEO, it took 1 week from cocky to fired, hopefully the same happens to sama, Zuck, Musk and all the snake oil salesmen in this space:

- 2-nd of Aug 2025 Github CEO delivers stark message to developers: "Embrace AI or get out of the industry" https://www.businessinsider.com/github-ceo-developers-embrac...

- 11-th of Aug 2025 Github CEO resigns https://www.theverge.com/news/757461/microsoft-github-thomas...

Hopefully the amount of instability in the space and the many signals of stagnation (e.g OpenAI's valuation divorced from reality, the GPT-5 release, Claude changing their limits, Meta spending millions on single hires) will indicate to people paying attention that it's a very volatile and bubbl-y space to invest in.

Sources:

https://www.theregister.com/2025/03/26/microsoft_ai_apocalyp...

https://nypost.com/2025/08/01/business/meta-pays-250m-to-lur...

https://techcrunch.com/2025/07/28/anthropic-unveils-new-rate...

https://arstechnica.com/information-technology/2025/08/the-g...

https://fortune.com/2025/08/08/openais-reported-500-billion-...

https://finance.yahoo.com/news/beneath-ai-bubble-economy-loo...

I personally can't wait for the bubble to burst so I can "told you so" to all the enthusiasts/fanboys.

Now I’m torn between Betteridge's law and my belief that we’re in a bubble.

No. It was always a bubble. You can thank me later.

I think a Wharton Professor Ethan Mollick said it nicely: Whats the capex on creating GOD?

Yet, it's not GOD it's a parrot.

squaaauk there are THREE b's in blueberry!

What a facile thing to say. Embarrassing, too

Reply on news.ycombinator.com