This article is all vapor. After reading it in it's entirety, there is only a short section that's actually about the headline, and it communicates nothing about what the headline means. What is 17x what?
No link to the original research note. No real details on the methodology used. A few notes on the well-known lack of an AI business model (similar things were said about search in the late 90's).
I just don't see how the broader market is exposed to an AI crash in the way it was exposed to subprime loans. If OpenAI goes belly up is it really taking anyone else down with it?
During the dot-com era, internet or IT in general accounted for a much smaller percentage of the GDP. So, I'm not sure how the percentage of GDP can help us gauge the scale of the bubble, if any.
When you're talking the size of investment that AI-centric companies have received, on the order of hundreds of billions of dollars, there's no way it's not exposed to the wider market.
But I agree with you, the article is too light on details for how inflammatory it is.
There's good reason to believe that OpenAI's success (or failure) and the success of many other firms are correlated. If OpenAI's bubble bursts, then that is likely to spread to other close firms and – depending on severity – any other firms that are merely associated.
NVDA, MSFT, AAPL, META, and GOOG are all heavily investing in AI right now, and together make up 28% of the money tied up in S&P 500 indices. Simply investing in the S&P 500, which many people do, exposes you to meaningful downside risk of an AI bubble pop.
Don't get me wrong; I'm no fan of the billionaires. Eat the rich, etc. But I don't want the billionaires to lose everything suddenly, because I'm 100% sure my 401k will go down with them, and 50% sure my job will.
A quick and dirty google shows that global financial institutions wrote down 1-2 trillion in mortgage related securities, a ~6trn drop in the us real estate mkt and another 6trn destroyed in the equities mkt. Not sure I really trust these numbers but they are the right order of magnitude. I have a very hard time believing the AI bubble is as big as the subprime bubble. But it very well could help trigger a massive correction and recession on a similar scale because everything else that has been propped up by low rates will likely correct with it.
In the subprime crisis 10 million American families lost their homes and 8.7 million people lost their jobs.
In the case of the collapse of an AI bubble, I don't see as much of a direct relationship to effects on the average Joe. Yes all those billions spent by tech investors will get written off, the companies heavily invested in AI will shed well paid tech jobs in a sector that was craving talent anyway.
I think the biggest effect would be the fact that all that capital was spent on AI tech rather than productive assets and businesses. That's a big opportunity cost, and would hit growth, but I don't see it wiping out ordinary people in the same way. The pain will be heavily concentrated on investors and for everyone else it will just be a slow drag but not a catastrophe.
The real problem is if there are other negative economic effects that compound with it.
> In the subprime crisis 10 million American families lost their homes
When you say "lost their homes", do you mean "I owned the house and somehow I now own no house and have no money for it, it just evaporated", or do you mean "I took a loan I could not afford, on a house I could not afford, while investing a tiny amount of money or none at all into the deal, and hoping to profit from ever increasing prices, and using my equity as an infinite-money ATM, and when that stopped, the bank took the house back"? If the latter, then what was lost is not "homes" but unrealistic prospects of profits from the thin air. If the former, I'd like to know how exactly a subprime crisis could cause something like that.
That is a very broad and angry brush you're painting with.
There were plenty of people that bought houses at reasonable prices and down-payments and still lost their ass when downstream ramifications took out unrelated businessss.
I don't doubt that, I just think when throwing around million-sized numbers it is necessary to be aware that these numbers are not describing one case, they describe a lot of different cases. There were a lot of people that were hurt by the subprime fallout (myself included, though I did not lose a home because I could not afford one at the time anyway) and also a lot of people engaging in absolutely reckless speculation (which also was one of the reasons why I was not able to afford a home at the time). There's more than one side to that picture, and some of the people who gut hurt were also the people who made it possible to get that crazy in the first place.
I don't share your doubts. Even people not directly exposed to the bubble may be indirectly exposed by way of funds they've invested in or retirement accounts, or even just by having invested in something that invested in something that may evaporate with the bubble. Not to mention jobs that may disappear, which will result in the people who held those jobs being less able to spend their money, which has knock-on effects in the rest of the economy if it happens at large enough scale.
Didn't a lot of people lose their houses because they bought a house with a mortgage payment they could absolutely afford with their income, and then got laid off when the economy contracted?
And, frankly, it's literally the bank's job not to make loans that people will default on too frequently (for their own sake), so if you're not exceptionally knowledgeable about banking, it's not unreasonable to trust your bank and their advisors not to make a loan you won't be able to pay back. Like, sure, you shouldn't trust them not to screw you on the terms and with interest, but banks mostly are trying to make loans they expect to get paid back, and I would personally expect them to have a good idea of how much they can trust me with.
> And, frankly, it's literally the bank's job not to make loans that people will default on too frequently
It's not the bank's job though to decide whether it's ok for you to treat the house as a long-term asset which consumes a part of your cash flow, or as a speculative gamble. You can find a bank that will support either, but it's on you to decide which road to take. And if somebody takes the speculative road and loses, then it's not exactly the banks' fault. The adult should take responsibility for their own actions.
> it's not unreasonable to trust your bank and their advisors not to make a loan you won't be able to pay back.
No, it's not reasonable at all. Loan officers do not have a fiduciary duty towards you. They have a fiduciary duty towards the bank, so that's what they worry about - to take care of bank's interests. Assuming those interests would always align with yours is a dangerous naiveté. There are financial advisors who are fiduciaries - and you can hire one if you need - but you won't find them in your mortage bank's loan office. Yes, the bank is interested, in most cases, not to produce overtly bad loans - but that doesn't mean they care how you are going to pay it, and there's a lot of chance they'd sell your loan to another servicer in a year or two anyway. They have no duty to figure out if taking this loan won't harm you, that's your duty.
Agree with 95% of this. The only slight tweak id make is that the banks did err in assuming the value of the homes being used as collateral could not tank. Setting aside for a moment the reality of pretty rapid sales of the note to another entity, in general, it shouldn’t matter very much to the bank if you’re able to pay the loan back, assuming that their predictions are correct about the value of the home, and assuming as well that you don’t physically destroy it. Either they get their money or they get to sell the house and get their money.
Uh... neither. The former is wrong, because as you point out the housing supply isn't a matter of finance. People have homes.
But your respin is kind of a whopper too. While there were absolutely people cynically leveraging real estate to make a buck, the overwhelming majority of foreclosures in the wake of the '08 crisis were just regular homeowners. They needed a home (maybe they moved, or got married, grew up, downsized, etc... people need homes!). So they called a real estate agent and a bank to figure out what they could get, and everyone told them (correctly) that they could get a great home at a very reasonable price with very little down payment. Because everyone else was doing it. So they did.
Insert the comment everyone’s mom says about peer pressure: “if everybody else was jumping off a cliff, would you jump too?”
Everyone who took an adjustable loan, interest only loan, etc., who didn’t have an exit strategy already in place in case of inability to refinance, had themselves to blame, regardless of whether “everyone was doing it.“ I don’t mean any criticism toward people who happened to lose their jobs and would’ve otherwise been able to continue paying on the loans they’d taken. Nor am I saying it’s OK to take advantage of people who don’t bother to read or understand the assumptions inherent in the contracts that they’re signing. But people were incredibly naïve if they accepted some broker’s verbal assertion that they’ll always be able to refinance the otherwise-unaffordable house on favorable terms in 3 or 5 years or whatever.
AFAIK its theorized much of the GDP growth of the US, as well as the stock market growth of important index funds, is mostly as a result of select tech companies surfing the AI hype.
In that case, a prolonged recession may occur (that would've occurred anyway), and the effect will be felt throughout the economy.
But, again, that's just a general recession being triggered by the AI bubble bursting, i.e. AI no longer propping up the economy, so that's not a bad thing. What the results of that are in terms of severity or impact I wouldn't know, I don't think anyone knows.
Why do you think it would be a prolonged recession? Dot com bust, for example, was just two quarters of GDP decline, followed by solid growth. 2007 was much worse, because it was a crisis of financial institutions. AI bubble may be bigger than dot com, but feels more like it in that is a narrow section of the economy. Even more narrow than dot com.
Agree. The reason given and examples are totally vapid.
Artificially low interest rates have stimulated investment into AI that has hit scaling limits, says research firm
He blames "low interest rates," yet interest rates have surged since 2022 to their highest levels in decades. He cannot even get the basic facts right, which kills his credibility at the start.
This also torpedos a common narrative that high interest rates are always bad for asset prices. The difference between 1% vs 5% interest rates does not factor much into VC decisions when the expectations are for 40-100+% annual returns with the hottest AI companies, which far exceeds the additional cost of borrowing. A similar pattern was seen in the the '80s and the late '90s, in which high interest rates also coincided with high valuations of tech companies.
This means a much longer effort at reflation, a bit like what we saw in the early 1990s, after the S&L crisis, and likely special measures as well, as the Trump administration seeks to devalue the US$ in an effort to onshore jobs," he says.
In an attempt to paint a negative picture of impending crisis, he gives examples, of 2001 and 1991, of among the mildest recessions ever. The US stock market and economy would go on to boom in 1995, just a few years after the S&L crisis.
If there is a job that AI needs to automate, it's these overpaid and useless analysts.
> Make no mistake, I think that this is the biggest and most dangerous bubble the world has ever seen. The misallocation of capital in the US (which also includes housing, VC and crypto) is already 17x the Dotcom bubble and 4x the 2008 real estate bubble, and as it unwinds it will not just threaten significant economic malaise, it will threaten to overturn the entire globalist agenda, that developed with the advent of Thatcher and Reagan in from 1979 and 1982, accelerated with the fall of the Berlin Wall fell in 1988, and sped up again with China’s accession into the WTO in 2002.
The valuations vs revenue numbers definitely are alarming but the fact that it is a general purpose tech makes me reluctant to pull cash out of the market.
dotcom bubble may have been a thing, but the underlying tech was valid and created the big techs out of thin air. Admittedly after much pain & flushing out the pets.coms of the world
If we're at another moment like that do you really want to be sitting on the sidelines?
In the dot-com crash, the NASDAQ lost 80% of it's value, which is something I'd certainly like to be sitting on the sidelines for.
The trouble with a crash is that it happens fast, and you are likely not going to be able to get out fast enough. There is also the psychology involved - say you lose 20% in one day, then do you sell, or wait for a bounce back before selling? What if there is no bounce, and it goes down another 20% the next day, then what do you do?
Nowadays there are "circuit breakers" that will stop the market if it is crashing, so maybe you hear about the start of the crash an hour late, by which time the trading has been halted. What do you do now - put in a market sell order? If every one is trying to bail out at the same time, then the market will reopen at a much lower price (matching buy & sell orders), and so it goes ...
And conversely, if you try to preempt the crash, you might end up sitting out most of the run up. Lots of people called the dot.com bubble in the late 90s and got out, only to end up worse off than if they'd just held through the whole boom, crash, and recovery. (Of course, I'm sure there were also some that managed to time the market successfully! Not sure I'm that lucky though.)
All the financial engineering to prop up the bubble is going to end badly, it always does.
Edit: I think it's fair to say there is a fair bit antics by companies that are actually illegal, like in every bubble, that have been hidden by the mania. They get exposed as the tide goes out.
In short it's companies turning to issuing debt, fuelled by increase in M&A activity, potential IPO of OpenAI, followed by collapse as tricks to increase revenue will fail to meet expectations and companies that mismanage debt will go bust.
The real problem is continued socialization of risk when the inevitable bailouts come. The government is basically telling everyone to take crazy risks, because if the risk is big enough, we'll bail you all out.
> The government is basically telling everyone to take crazy risks, because if the risk is big enough, we'll bail you all out.
This is the widespread sentiment, yes, and why the bubble is so crazy. The feeling a lot of people have is; even if shit hits the fan, rather than let stocks/assets decrease in value relative to cash, governments will just print money and keep stocks/assets "stable" while making cash worthless. The US proved they're willing to do this during COVID. What you're betting on by holding assets despite the bubble is that they're willing to do it again.
Yeah, but can they afford it? The government's debt obligations and inflation made it possible, the circumstances have changed considerably. They could wreck the economy for a long time by doing something like that again (particularly if it isn't a war or something).
The people in control of the economy are not affected by the state of the economy. There is no incentive for them not to gamble because there is no downside. If they win, they get even more money and power. If they lose, they'll still have their yachts and land and hobby rocket, newspaper, and social media companies.
Putin the magnificent has said "whoever leads in AI will rule the world". So it is a war. An arms race. Weirdly being run by private capital cuz there was no incentive(unlike during WW2/Cold War) to fund a state run Manhattan Project for AI.
Watch what happens if a breakthrough comes out of China, Russia or a shed in Nigeria. What is happening on Wall Street or who is loosing cash or jobs will drop instantly down the list of priorities. Money will be printed to the moon.
I think rather than there being "no state inventive", its that there is already private incentive, so there's not enough reason for the state to get heavily involved outside of maybe providing some additional capital. Private companies are already putting in the effort and taking on much of the risk, the government can just step in later after the results show.
Of course not, but with the Fed always standing at the ready to print more money, the answer is yes. The downside is persistent inflation. Inflation and default are two sides of the same coin.
Why would the government bail out AI companies? They didn’t bail out the dot-com bubble. Bailing out the auto industry was about the UAW. Bailing out financial institutions went with programs to keep people from being foreclosed on and losing their homes. The AI bubble doesn’t have those incentives and is mostly (all?) private company valuations.
I would really hope that nobody considers AI companies to be foundational to the economy the way the housing market is. I know there is going to be people crying in Congress with "Think of the 401ks!" but I really want the hype men to feel some pain this time.
It's really a shame that it's really the regular people who are going to suffer the majority of the harm when this bubble pops, as the Wall Street insiders get off mostly scot free yet again but I really don't want the government trying to prop up these assholes again.
That's true of every bubble. It's the best thing around until it suddenly isn't.
Don't forget that bubble popping doesn't necessarily mean that the sector is dead. The Dot Com bubble popped but people are still selling things on websites. It just means the hype and unrealistic expectations come crashing down to Earth. AI is going to be with us from now on, but it won't become a God with unlimited productivity next year.
https://www.derekthompson.org/p/this-is-how-the-ai-bubble-wi... Goes into good detail on how the bubble has infiltrated everywhere. Like if you went into a REIT thinking it was safe from the bubble, nope a significant portion of your investment is in the data center buildout
It seems like an (ugly?) truth that if you don't want to be exposed to bubbles, you have to take lower-return investments, by definition. Because the bubble is raising the high bar of what kind of return people expect on investment, so if you try to get close to that, you're going to end up affected by the bubble.
The problem, of course, that no one wants to take a lower return on investment.
Yeah, but I guess the point is that it is so big you can't escape it, it's everywhere. So the fallout will not be limited to a segment of the economy, it's going to hurt everywhere.
And the reason nobody wants to take a lower return on investment is that doing so is trying to time the market.
Let's say you heard Trump's talk of tariffs, got spooked when he won the election, and moved your investments. The S&P 500 has gone up ~33% since election day. How much growth can you realistically miss before correctly hedging against a crash still net loses you money?
And that's assuming you pick the correct hedge. If the US economy well and truly crashes beyond return, to the point where the S&P 500 is actual garbage that will not recover in the next 5-10 years, that might bode poorly for the dollar, so that's it for your treasuries, money market funds, CDs, etc. Gold is always an option, until it's at an all time high because just as the economy is roaring along a ton of people are nervous. So, you can invest in gold, but it's expensive, which means it could actually dip if we get stability. Ok, so you invest in foreign markets - but the Global Financial Crisis was global, the Great Depression was global, everything was global.
What do you actually hedge in? Oil? Well, the rapid development of green energy might actually, finally render that an unsafe investment. Green energy? Think again, the president is on a crusade against it. There's literally nothing that's actually a safe investment, and even if you did find a safe investment, the big crash might not happen until everyone else riding the wave has doubled or tripled their investment, by which point they'll be on par or even above you post crash. I'm actually very happy I'm nowhere near retiring, since regardless of how else you feel of the current governing of the US, I think it would be hard to argue that they're fighting a war against conventional wisdom, and that makes investing very scary.
I can’t see the case where we won’t need more data centers in the future even if AI is a bubble as more work load is generated and even if it is a bubble. There are real world use cases where you will need ML and inference even if you don’t need to spend money on training.
> I would really hope that nobody considers AI companies to be foundational to the economy
The way people in the US government has been speaking and acting goes against that hope. Instead, it seems to be unanimous that AI will be the foundation of the economy, the US can't lose its leadership on it for even a second (because when it takes over, it will take over in a second), and it's worth breaking every rule, spending every cent to keep, and cannibalizing every other industry if necessary.
Honestly, you have to deal with the delusional con-man in power ASAP.
I think you could make an argument either way and have an equal chance of being right. Either Trump et al get bribed/flattered/are personally exposed to the fallout and bail out their fellow class members or he gets angry about it and decides to make heads roll via his increasingly captured and controlled DOJ. I still agree it's more likely the former than the latter but I can see the path that leads to the latter.
"Heads will roll" just means "severe punishments will be administered". It doesn't necessarily imply the punishments will be illicit, illegitimate, or even fatal.
It came out soon after AOL acquired Tine Warner in 2001 that they artificially inflated income to prop their stock up for the merger to go through. No one was punished for it.
Here's an useful exercise. Suppose someone makes an advance that makes LLMs training require 90% less compute, but does not make them any smarter. (Recent results from China indicate substantial compute reductions are possible.)
Demand for GPU time drops. What happens to the economics of the AI boom?
Demand most likely won't drop due to Jevons and the worst will be big labs and Nvidia probably drop significantly in valuation but still pretty good. Also recent china results only show that you could build cheaper model hitting the benchmarks not actually generally competitive irl.
Agreed, with one caveat: demand for hardware can drop if the optimization is so substantial that deployed hardware is already capable of fulfilling any possible request in a timely fashion. Say, anyone becomes able to run any state-of-the-art model on any desktop computer. 90% is not enough, though. It has to be at least three or four orders of magnitude. The ones driving demand are big datacenters, so they need to be in a position where just throwing more hardware at the problem solves nothing.
I remember something like that happening. Someone optimzed something and boom, one billion of valuation gone. It seems ridiculous on the surface, and definitely smells of a bubble to me.
That's happening to the automobile market in China. Electric car and battery manufacturing costs go down with scale. Way down. Huge electric car glut in China. Car prices crashing. Government worried.[1]
Profit margin of China car companies is down to around 4%.
I think the problem is bigger than that too. If you have a problem set where you cannot afford any kind of mistakes LLMs are never going to be the right tool, they are never going to get to that stage.
If you have a problem set where some mistakes are tolerable however, all you need to do is slightly increase your tolerance budget for mistakes and you can run something like deepseek locally at a tiny fraction of the cost.
Which means these giant data centers are useful for training but the economics of running ongoing inference for customers doesn’t make long term sense for the vast majority of people.
Which leaves you with a huge huge huge financial problem, one which may have devastating large scale impacts on the economy.
The vast majority of energy is already used at inference time. This gap is only going to widen as the number of applications increases and people ask their LLMs to do more complicated tasks.
If training costs go down, I expect models will be trained on more video, so training energy usage will not decrease either.
The mistake this article makes is assuming LLM scaling is one thing. It's not.
RL is spiky. It produces narrow improvements on specific capabilities. They're not making the model generically smarter, they're RLing in holes in the model's capabilities. In reality we don't have one scaling curve, we have thousands of them. We're in diminishing returns in "top line smarts" but we're raising the floor in a wide variety of areas that people who don't heavily eval models for a living might not notice.
Do you mean car title loans? Because that's basically what the GPU-backed loans are. Car title loans have terrible terms due to the diminishing value of the car. These GPU-backed loans are being booked at much, much more favorable terms, as if the GPU holds its value for a long time. This simply isn't the case, so if the loan is defaulted on, the loaner will have a rude awakening when all they get are worthless GPUs.
Isn't that part of why down payments are so necessary? It takes the loan amount below the immediate depreciation amount of the asset so the bank/lender is less exposed to the risk.
I doubt that's the main driver for lenders and people are generally pretty attached to and excited about their new cars. They also do have a stake they'll still be out time and money even if insurance covered the entire cost of the vehicle.
I've been playing with Llama 3.2-3B this weekend and, while I haven't spent enough time with it to figure out the limitations, it's definitely good enough to be useful. That's running at "as fast as I can read" on a £100 Intel Arc GPU. I'm tempted to buy a Arc770 (£250ish) to see if I can run the OpenAI open weights models...
I think it's reasonably likely that these massive investments turn to dust when they get squeezed by the commoditization of LLMs.
The Saudi sovereign wealth funds and similar Middle Eastern investors, Masayoshi Son, a16z (though not all of their funds), Tesla's board, and others seem to be operating in some kind of alternative reality where watching out for your limiteds and shareholders isn't a factor. Do they think they are really TBTF? Is that why JD Vance is where he is?
It's not just the size of the bubble that's scary. It's that some people obviously think they're going to get away with it.
> It's that some people obviously think they're going to get away with it.
Cynical, I know, but they absolutely will get away with it. The wrong people will be punished, as usual. They will claim that "there was no way to know" this could happen.
They will do what they always do, which is to vaguely imply that if they are left to the consequences of their gambling addiction then China will get ahead of "us". Note that the general public are only included as part of "us" when Sam Altman needs the general public's funding.
If you believe the rule of law and ~general financial stability of the US will persist just keep on investing regularly and ride it out. Markets have a long history of bouncing back and they will keep doing it until they don’t. When/if that happens your retirement dreams might seem quaint compared to the global socioeconomic reality. The people who could sustain the most near term pain are those who have already retired and are living in a fixed income.
Like everyone else, wait until the very last second, just as the market peaks, to unload your assets. Before the institutional investors and insiders beat you to the punch.
The AI hype bubble now represents something like 20-30% of the stock market depending on who you ask. For reference the great depression started with a 24% drop of the stock market. The people running this AI game know there is 0% chance the government will let this play out naturally because it would cause another great depression. Bailout Guaranteed.
The plebs can take some inflation,higher taxes and national debt increase for the hucksters to get some yachts and lambos. I mean what were you going to do with that money? Feed your family? Pay rent? Jeeze get a life....
In this round of con there is not even a way for an average pleb to get any scraps since the AI bubble is almost completely hidden behind private equity. Except maybe Nvidia stock.
The bubble is starting to froth/pop IMO. An incestuous cycle of players propping each other up with circular investments(ie I'll loan you money to buy my product so I can sell it to you) began last month with Nvidia "funding" OpenAI datacenters. Actions like that mean they are out of external ways to keep shuffling the debt. Like when WeWork CEO started loaning the company its own money to buy its own product to rent.
Edit: Oh $hit AMD just did the same thing with a circular funding deal with OpenAI. https://news.ycombinator.com/item?id=45490549 Channel Stuffing Money printer go brrr. I can't belive there is even discussion if there is a bubble at this point. Its literally wolf of wall street "Never let them cash out that makes it real" style deals all over the place.
> An incestuous cycle of players propping each other up with circular investments
Of course I'm reminded of today's announcement about the strategic partnership between AMD and OpenAI which caused AMD's stock price to jump a whopping ~35%.
AI isn't the only component of the bubble. SpaceX which has a genuinely successful business with Falcon 9, is also a huge bubble. SpaceX launches an amazing 25 times as many rockets as ULA in a year. But 2/3 of those are Starlink launches. No outside revenue. And SpaceX is valued at about 200X the valuation of ULA.
The AI hypy has inflated the values of the "magnificent seven". Meta,MS,Appl, Amzn, Nvidia ect. Retail and various professonal investment chasing indirect investment in AI,ie MS owns almost half of openai. Also 401k rebalancing. Are driving the valutation of those public stocks to very unbalanced valuations with zero increase in business.
Its complicated. Though you could argue its all good business. But first buy my swampland in florida please.
> I'll loan you money to buy my product so I can sell it to you.
I predict this will happen among Nations as well. US will provide money to Japan so that Japan can invest that money in US (at Trump's discretion) so that everything is kosher.
That is what traditional US foreign aid was all about. We give money to allies, they buy our military hardware or farm products. Trump of course has shit all over the system, so who knows what we are going to do now.
We also give out tons of subsidies and tax breaks to lure foreign investment to the US.
This article is pretty low quality IMHO. Lots of conjecture about the value of LLMs. We are far too early in this hype cycle to know what will come of this new tech. There is, however, no doubt that incredible amounts of capital are flowing into data center construction and much of that capital allocation will end in tears. There will be losers.
Bitcoin not popping is a testament to human delusion. It cant function as a stable store of value becuase theres nothing preventing if from hyperdeflating, and yet people are still buying it. What are they even buying? The deflation based gains? Deflation based on nothing but speculation? It doesnt do anything. Its madness
I dont think the GenAI bubble will be as resilient because its not being funded through discretionary expenditure. Its being funded with capital expenditure. If your cousin spent his savings on bitcoin and isnt getting any benefit from it (or even loses it in a drop) thats fine because millenials dont have savings anyway. But if a company spends on AI and doesnt get a positive value prop in a while, they will pivot and pull their money out, because shareholders will get mad
No it's not, because nobody knows how big the AI bubble is. Or even if it actually exists at all.
Because the concept behind a bubble is that it's inflated over the true value due to speculation or non-speculative overinvestment.
But there's still a true value.
So you can say AI valuations are whatever multiple of previous whatevers, but you can't say the AI bubble is. Because if there's an AI bubble (and there probably is, but nobody knows), nobody knows how big the bubble part is versus that solid part. Obviously, if anyone did, they'd make lots of money by investing accordingly.
And of course the reason we can't figure out the bubble size is because nobody knows what the true value of LLM's will be over the coming decade or two. It's hard to know even what the order of magnitude will be. And how much will be captured by existing LLM companies.
I would say if you're holding a lot of money in a S&P 500-tracking index fund, switch to holding a total stock market fund instead. It's probably a bad time to hold SP500 right now, but you don't want to get out of the market entirely because there's no way to know when the bubble pops. So hold a total stock market fund instead, so that you don't have an outsize amount of your portfolio invested in the companies leading the SP500 (which are the most likely to lose the most when the bubble pops).
I agree with using a TSM fund rather than an S&P 500 index, but be aware that you're still mostly holding S&P 500 companies. The S&P 500 covers about 80% of the total market cap of US public companies, whereas the Russell 3000 is 98%. (This is, of course, because the 500 largest companies are so much larger than the median.)
If you really want to diversify away from the biggest growth companies you could try a value and/or small/value tilt. Those might all be good advice, but they do come with their own caveats, whereas using TSM instead of an S&P 500 fund is just good advice in general.
The weird thing that makes it different to subprime is that AI is real and really could become AGI. But it might also be overvalued.
So it's not like there isn't a product (there is) and there's no growth prospect (there is). But it is scary how much now hangs in the balance of one bet.
It almost feels like it's going to end badly either way. If the Great AI Bet succeeds, a tiny proportion of the world will own all intellectual power. But if it fails, the impact of the write-offs on the broader economy will be terrifying.
Every single researcher not being paid $$$ by AI companies say there is 0 path between LLMs and AGI, but sure... and the next pfizer drug might make us immortal, who knows, everything is possible after all
I dont think we'll have AGI in 2 weeks like altman does, but its not fair to say its impossible.
LLMs also came out of nowhere, a series of discrete improvements that finally got over the hurdle and acheived an unprecedented functionality. Absolutely no one predicted its emergent capabilities back in 2016 when the transformer paper was proposed.
They didnt see scaling then, they might not see the next thing now, until its found.
So it wont be in the next 2 weeks, and it wont be from OpenAI, but it might be in 10 years from some random researcher at waterloo, or tiktok
Will AI of today definitely become AGI of tomorrow? No, for sure not, and anyone who claims this is at best crazy.
But is it imaginable? I think totally. Andrej Karpathy' blog post about RNN writing Shakespeare 1 character at a time was 10 years ago. GPT-2 was released 6 years ago. In that time we went from something that barely speaks English, never mind any other language, to something that, on a good run, is an excellent language tutor (natural and programming), can play games, can write moderately complex programs, goodness knows what else. For some people, the romance of a ChatGPT-4 was unmatched.
Even if it doesn't become "AGI", it might just get so good at being sub-AGI that the question is irrelevant. We're seriously contemplating a near future where junior devs are replaced by LLMs; and I write this as an AI sceptic who uses LLMs to write a lot of the kind of dumb code a junior dev might do instead.
I don't like AI, in that it nibbles away at my competitive advantage in life. But it's IMO crazy to pretend it is not even potentially a game changer.
The next logical step for cars were flying cars, the next logical step for planes was space travel. Both were hyped not so long ago, people believed in both, it never happened and it isn't closer to us now than 50 years ago
I'm not saying it doesn't bring any value, I'm just saying that if you think we should give $7 gazillion to Altman because he's building skynet by 2030 you're smoking crack
Ah, the same reason climate scientists not paid by oil companies agreed on global climate change, and doctors not paid by cigarette companies warned about lung cancer?
I meant it as a joke. Should've added an /s instead of "I don't know".
But in all seriousness until someone else invents AGI some other way, you can't disprove that LLMs aren't the way. My intuition says you need more than LLMs but I could very well be wrong.
Artificial Intelligence is an entire discipline of the Computer Science field. It encompasses everything from how Pinky, Blinky and Clyde chase Pac-Man, to A* search and similar pure algorithms, to machine learning, computer vision, and LLMs.
It is also a term used widely in popular culture and media to mean, essentially, AGIs—Cortana, Agent Smith, C-3PO.
The problem is not that this term is very broad, and it certainly isn't that it has come to mean something that it didn't before. It is that a bunch of people with a financial interest have been busily trying to convince the world that LLMs are Cortana.
What would count as proof of AI trajectory to AGI?
Currently we have:
- cross domain competence
- composable tool usage
- constant improvement without clear signs of stagnation...some can count gpt 5 as not much progress but there are still world models with huge gains compared to year before
There is a bunch of things missing for sure like mechanistic reasonic, better context lengths, determinism, better world modelling, continuous learning.
The bet is crazy and whole world is gambling on AI, right now, because of those signs of potential. That's precisely why we landed here, the evidence was good enough for big tech to gamble on it...
>What would count as proof of AI trajectory to AGI?
It depends on who you ask. There is no absolute for definition of "AGI". Sam Altman defines it in monetary terms, because that benefits him. I have a very different idea of what "AGI" means. It's really very subjective, so I don't have a definite answer for you. I'm sure you'd define "AGI" differently than I would, so having this discussion is kind of a waste of my time.
The peo1ple pouring money into "AI" are doing so in the hope that it will become more reliable someday. My educated guess is that it won't, due to the underlying mechanisms that it is built on. Predicting the next word in a sentence according to grammatical rules is a long, long way off from a machine knowing and understanding how truthful the resulting sentence is.
I describe it as both a bubble and the next industrial revolution. It's not so much inflated as it is far ahead of itself IMO. Kind of like NVDA stock over the past nearly 3 decades.
The talent out there is all focused on a tiny number of tasks and benchmarks in service of the AGI cult whilst the real gain is scattered amongst everything else with no staffing to build it. So I see a bust much like the dotcom bubble in the arbitrary future, but then it corrects surprisingly quickly shortly thereafter and the engineers of that bubble will have gotten out temporarily well ahead of that bust, buying back their assets at 50% or greater discounts and the broader base of retail investors are screwed as usual.
As for impacts on the economy. Meh. They're already starting to ignore the ramblings of the Orange in Chief (the guy shuts down the federal government and the entire stock market yawned last week). The weakly efficient market abides and it's been burned enough times already by the TACO trade. It'll get through this. But oh the whining that will ensue on the corporate media.
Real estate is real too. People need houses to live. Websites were also real value
The dotcom bub, subprime crisis, and AI bubble are all based on real goods being overvalued to hell and then crashing back down to their actual capabilities. The soze of the bubble is the delta between their actual value and the market's percieved value.
In that regard i think AI will be crazy crazy valuable in real terms but not like the market is using it. I think the real value of AI agents is very very low and the market will crash to that level. I think the value of genAI as much simpler interfaces for communication (RAG, translation, NLI) and for automated understanding of static sytems (rather than acting in dynamical systems) is high and will crash a bit before the market learns to use them right and then itll be a party
It might be. By feel, it is different in that a lot of the AI investment is into real things, that hold non-speculative value: data centres, GPUs, training marketable and genuinely useful products. It's not food for pets but online and sent by email.
I think that's mostly training costs. There is a serious level of diminishing returns and the primary way companies have been trying to get ahead is throwing more and more hardware at it. AI already burns something around 5x as much power as crypto mining and it's only growing.
You hear sometimes about the AI singularity and how they will continue to become smarter at an exponential pace until they're basically gods in a box, but the reality is we're already well into the top of the S curve and every advance requires more effort than the one before.
You should be more concerned that Chinese labs can train models that are just as good for 10X less because Americans treat the USD’s status as the global reserve currency as the ultimate bitter lesson. Who needs better math and engineering when you can print money to buy more GPUs???
The author literally said gpt is spending 10x more for equivalent. This really means ChatGPT had that intelligence at that cost a year or two ago. Smaller domain models are better at focused tasks in that area but can’t be generalized.
Surely you aren’t talking about Android as “open source”. To a first approximation no one wants a phone running only AOSP without Google’s proprietary parts.
You're free to use the same Android that the Chinese OEMs (and Amazon) built their phones on yourself, without traveling to China. It's true that a reasonable definition of "no one" wanted Amazon's phones, but a much more lenient definition has to be used for the Chinese OEMs.
Nvidia is way more than their data centre hardware - their software and research is just leagues ahead of the nearest competition, and their edge compute capabilities are going to be a rapidly growing market as derived models start getting put in products - think self-driving, robotics, machine vision, manufacturing and industrial automation, on and on.
So yes - while there are others who may yet compete in the data centre compute market, nothing else comes close to the monopolistic total vertical integration nvidia has built over the last decade.
I lived through dotcom boom and bust, and there had been a lot of craziness, but when the dust settled, we have now a functioning internet economy, and I am able to sit at home in my pajama pants working for an international company that has employees all over the globe, while enjoying many other conveniences the internet brings me - and all this is enabled by what started during dotcom era. It is true that it did not match the most wiled-eyes predictions, but also a lot of things did change, and a lot of things did improve. If the AI boom followed this trajectory - which is of course no guarantee of, but if we assume that - then there is going to be a lot of crazy stuff to come, but at the end of it, the dust will settle and we will be mostly better off from it. Some people would become billionaires, some undeservedly so, some people would be ruined, some also unfairly so, but for most people, at the end it won't be that huge.
We are in the era of people using geolocation to remind them about their grocery list when they're near a store. In 10 years we will be in the gps and uber era.
Its just that it will be bumpy when this bubble crashes. And the US isnt as stable and sane as it used to be. I wonder if the economy can be steered through waters this rough
The troubling part is we are in the AI bubble whether we opted in or not.
The downstream effects are everywhere - real estate (data centers), energy (power consumption), education (automated tools), employment (tech layoffs), and equity markets (Nvidia propping up the spoos). It's not one sector this time, it’s systemic adjacency.
During the dot-com crash, retail got burned. During subprime, it was the global financial system. Today, AI exposure is distributed among mega-cap public companies, a16z-style VC funds, sovereign wealth, and shadowy circular debt deals between hyperscalers and AI startups.
Even if you are not betting on AI, your retirement fund is.
What's the mood in the bay area? If this is 17 times the size I can't imaging how much optimism/fun you all are having.
I was there for the first dot com boom. So much optimism. It seemed every party someone announced they were pregnant we were all so optimism for the future and comfortable, ready to start families. The restaurants were all full with new spots popping up all the time. It was a really fun time to be there. I know there are some layoffs but at 17x even with some layoffs I can't imagine 17x more energy/money then back in the first dot com.
But what if AI succeeds and it’s the rest of the economy that implodes? Long-term, I’m sure AI will lead to massive productivity gains. Short-term, I think it’s going to be chaotic. I don't know which will implode, AI or repetitive white collar work or something else
Worse than the bubble is what's happening to the rest of the economy. If you remove AI related spending the US economy is trending in a really bad direction.
This bubble popping will definitely take down crypto with it and rip through other adjacent industries.
How can the AI bubble burst in the public markets? It is circular investment within silicon valley. Even if AI proves nothing, overall the tech market is in the same spot where it started
Not all AI investment is private. You could argue that public companies like Google, Meta, and Microsoft have had their stock appreciate at least partly because of the AI frenzy.
Has anyone articulated a theory for what the profit model might actually look like for the big LLMs? In the .com era, the valuations were crazy, but people generally understood how you could eventually make money on all of this stuff.
If models are requiring larger and larger infrastructure buildouts, does anyone have a clear sense of what users will have to pay in order to make the businesses profitable?
It seems that each model generation, and each generation of GPUs, has a limited shelf life, so you need to make more money from each generation of models/GPUs than it cost you to build/buy them. User pricing to make a profit depends on inference volume. You need to make back a fixed dollar amount before breaking even, which could come from high volume and low prices, or lower volume and high prices.
The trouble is that everything is changing so fast that any kind of forecasting is extremely error prone, especially when one forecast builds on another. First you need to guess how much more capable the models are going to get (at things where people will pay more for better performance), in what time frame, then guess what level of demand (inference volume) will exist with that level of capability...
The LLM developers like OpenAI and Anthropic like to tout things like Math Olympiad and Competitive Programming results as signs of progress, but there is no guarantee that they will be equally successful in applying RL to more general areas of commercial value where RL rewards are harder to define.
These companies also like to talk about "scaling laws" is if there was some inevitability about investing more money & compute and getting better results, but this only works until it does not, and they replace one broken "law" with another. Right now it's all about scaling of RL-training and test time compute, but how long will that last, and what type of problems will benefit?
The profit model here seems a bit like the Drake equation for calculating the probability of other intelligent life in our galaxy... it may be possible to define the equation, but the outcome depends on having the right values for all the variables, which are largely unknown.
With the direction of OpenAI, hyper personalized ads inserted directly into chat and their app experiences could be a path. Not saying it will work, but they’re definitely exploring it.
Who gets hurt when the bubble pops? And who will the government need to bail out?
Last time, a lot of the companies were public and the general public saw stock losses. This time the only companies that are publicly really exposed by the AI bubble are Nvidia with all of their circular financing and Oracle. Of course Tesla has always been a meme stock.
Defined contribution plans - for now - can’t have private equity in their funds. Defined benefit plans and endowments are exposed.
Apple famously hasn’t invested that much in AI, Google is spending a lot on infrastructure. But between search and GCP and YouTube they have a real business plan and are funding based on profits. Amazon is in the same boat. Microsoft is bowing out of spending money on training and focused on inference - and they also have Azure. Meta is making money using AI for ad targeting and probably in the future to generate ads.
I can also see consulting companies being hurt (I work in cloud consulting) as businesses are throwing money at them to “AI enable” their business.
Given how investing is heavily promoted by all these neobanks I have a feeling a lot of people will get burnt. Back in the days, not even 10 years ago, you had to research and go out of your way to invest, now you can do things like "automatically round up your transactions to buy NVIDIA" from your bank app. The only ways to get out of the middle class are: lotteries, crypto, putting everything stock market for 20 years and living like a student in the meantime.
Most people in the middle class can’t afford to put any decent chunk in their 401Ks let alone invest any meaningful amount above that.
But funny enough, most of the people I know that do invest in individual stocks, bitcoin etc are people in the servjce industry who are single and make decent money on tips. I live in very heavy tourist town.
But since this is a site of tech heavy participants, if you are a software developer or adjacent, you are on average making twice the median local wage for your area if you are in the US even as a enterprise Dev 2-5 years out of school and should be able to invest at least 15% if your income.
I wonder how the tech job market would look without the AI bubble. Is the AI the cause of the unemployment crisis(if we believe the CEOs who layoff a quarter of the company because "AI is replacing thousands of developers") or if we didn't have this crazy AI bubble, the market would be a much worse bloodbath.
Which non-bubble period of time has been crime free?
Its very unclear to me if there is actually more crime during bubbles or if people want someone to blame so the powers that be investigate harder until they find someone.
We do? Have you been reading the news where the President is firing AGs who won’t go after his enemies and firing FBI agents who didn’t agree to perp walk his enemy? Not to mention he just pardoned 1500 domestic terrorists in January.
Public companies are outright bribing the President to get mergers passed (Paramount) and the entire confiscating TikTok to give it to his buddy at Oracle on the cheap can’t be ignored.
On the federal level, there really isn’t any legal standard anymore.
"The rule of law" is the rule of the economic elites over the weak plebs. The "rule" is formed, shaped and ignored so they can have their way. Anything goes when you operate on the upper levels, we are not naive kids.
Disclaimer: I am not a doomsday apologist, just a Nerd collecting, analyzing data and doing projections out of it.
There is not only an AI bubble, but a US Dollar bubble, too. Most people don't get it what "Stock markets, gold, real estate, crypto, EVERYTHING is up" actually means: That the currency you are paying in is losing value at rapid pace.
I do not yet understand how the market manipulation actually works, but if you exchange from an Asian country to US Dollar right now, an amount that will buy you a excellent dinner for the whole family no longer even buys you a hamburger in the US.
The part I also do not understand is why the only thing that right now is "cheap" as seen from the US perspective are foreign currencies.
I also don't understand why other countries are not liquidating their USD reserves. How can you as a country not see that your reserves now would be much safer in Asian countries including even the Chinese Rimibi?
Well, maybe the answer to the USD losing value rapidly is that in reality there ARE market actors in the background moving to other currencies, but are really really good at hiding it.
But in any case: No doubt, the western part of the global financial system will be crashing within the next 3 weeks. With the US being on the brink to civil war anyway, you should be able to extrapolate what a financial crisis and bank run in the US in combination with everyone and their dog owning war-grade weapons will end up at.
Right now you have heaps of people in the US who are completely ignoring any data but invest only based on being member of a fanatic cult - every time you read news that Tesla no longer is able to sell cars, or that their robots do not work, or that their Robotaxi stuff is a scam, the Tesla share value goes further up. The same applies to AI. That AI will never make any profit is no longer niche knowledge, it's headlines at Bloomberg, WSJ, FT & co. You have to actively ignore that information - still, AI stocks go up.
So what is missing for the bubble to finally burst? Anything that makes those cult members start to doubt. Anything that triggers "Maybe putting all my retirement savings into crypto is too risky" or "Maybe Elon actually isn't the messiah", "I put all my money into a Trump meme-coin. Is this really more important than getting my teeth repaired?". Or anything that triggers those sheeps that are invested into this to simply needing cash quickly. So, for example due to a natural disaster, or some pandemic, the Epstein stuff finally blowing up on Trump etc.
The list of things that now may trigger the implosion is gigantic. Due to that it is possible to predict where it will start - it might not be the AI bubble. It very well even may be something rather classic like the "Deported migrants no longer pay towards their mortgages / loans" stuff happening right now.
But in summary: Due to the very high number of potential triggers it is stunning it hasn't imploded yet, and if you read/watch interviews from scientists in this area, the common theme these days is: "WTF is going on here?! This should have imploded LONG ago!".
Buried lede: what is the transformative potential of ML/AI?
It's not a bad bet that many are making that "changes everything" is real in this case. Hitching a ride to the technology getting civilization over the hill is a reasonable goal.
But what institution, technology, play...?
Following the money shows where bets are being placed. But none are safe, the disruptive consequences of network effects make diversification appear wise. The industry is littered with the bones of giants.
Personally I am looking at ML/AI as analogous to the Great Oxygenation Event. What things look like—at least, on land—afterward, none know, least of all the cranky old oligarchs who are hell bent on consolidating control and ownership.
If you marvel at the output of generative AI be it pictures, logo, music, 3D models, prose or code - you probably are yourself way too under-skilled to be gainfully employed or not yourself capable of generating similar or better output.
Conversely - any expert (prolific writer, coder, painter, photographer, videographer or log/web designer) isn't as much amused or going to scream out of excitement that what these models are producing (vides, pictures, logos, essays, code) - they could never ever have thought anything better than that.
This fact alone is enough to warrant that a big bust is coming. Not a matter of if but when.
> If you marvel at the output of generative AI be it pictures, logo, music, 3D models, prose or code - you probably are yourself way too under-skilled to be gainfully employed or not yourself capable of generating similar or better output.
Interesting take. I am continuously impressed by these models. I also train these models for a living, and have worked on some of the highest profile models of the last few years.
> you probably are yourself way too under-skilled to be gainfully employed or not yourself capable of generating similar or better output.
Have you considered the posibilities that "subpar" coding skills are still extremely valuable?
I say this because almost every single extremely high paid engineer that I know at rocketship unicorn startups to FAANG companies are all using AI coding as an essential part of their work flow. Its ubiquitous. And we get paid tons of money.
We aren't just copying an pasting hundreds of lines of code and pushing to prod, of course, but its an invaluable tool for significantly speeding up an engineers coding workflow.
This contradicts your claim, as these aren't need grads, these are all highly paid professionals.
This article is all vapor. After reading it in it's entirety, there is only a short section that's actually about the headline, and it communicates nothing about what the headline means. What is 17x what?
No link to the original research note. No real details on the methodology used. A few notes on the well-known lack of an AI business model (similar things were said about search in the late 90's).
I just don't see how the broader market is exposed to an AI crash in the way it was exposed to subprime loans. If OpenAI goes belly up is it really taking anyone else down with it?
A shocking statistic is that YTD, 1/6th of the 2% rise in US GDP this year is attributable to AI Datacenters alone.
Source: https://www.economist.com/finance-and-economics/2025/08/18/h...
So I think if there was an AI crash, US economy goes with it in the short term
During the dot-com era, internet or IT in general accounted for a much smaller percentage of the GDP. So, I'm not sure how the percentage of GDP can help us gauge the scale of the bubble, if any.
When you're talking the size of investment that AI-centric companies have received, on the order of hundreds of billions of dollars, there's no way it's not exposed to the wider market.
But I agree with you, the article is too light on details for how inflammatory it is.
There's good reason to believe that OpenAI's success (or failure) and the success of many other firms are correlated. If OpenAI's bubble bursts, then that is likely to spread to other close firms and – depending on severity – any other firms that are merely associated.
NVDA, MSFT, AAPL, META, and GOOG are all heavily investing in AI right now, and together make up 28% of the money tied up in S&P 500 indices. Simply investing in the S&P 500, which many people do, exposes you to meaningful downside risk of an AI bubble pop.
Yes.
Don't get me wrong; I'm no fan of the billionaires. Eat the rich, etc. But I don't want the billionaires to lose everything suddenly, because I'm 100% sure my 401k will go down with them, and 50% sure my job will.
A quick and dirty google shows that global financial institutions wrote down 1-2 trillion in mortgage related securities, a ~6trn drop in the us real estate mkt and another 6trn destroyed in the equities mkt. Not sure I really trust these numbers but they are the right order of magnitude. I have a very hard time believing the AI bubble is as big as the subprime bubble. But it very well could help trigger a massive correction and recession on a similar scale because everything else that has been propped up by low rates will likely correct with it.
In the subprime crisis 10 million American families lost their homes and 8.7 million people lost their jobs.
In the case of the collapse of an AI bubble, I don't see as much of a direct relationship to effects on the average Joe. Yes all those billions spent by tech investors will get written off, the companies heavily invested in AI will shed well paid tech jobs in a sector that was craving talent anyway.
I think the biggest effect would be the fact that all that capital was spent on AI tech rather than productive assets and businesses. That's a big opportunity cost, and would hit growth, but I don't see it wiping out ordinary people in the same way. The pain will be heavily concentrated on investors and for everyone else it will just be a slow drag but not a catastrophe.
The real problem is if there are other negative economic effects that compound with it.
> In the subprime crisis 10 million American families lost their homes
When you say "lost their homes", do you mean "I owned the house and somehow I now own no house and have no money for it, it just evaporated", or do you mean "I took a loan I could not afford, on a house I could not afford, while investing a tiny amount of money or none at all into the deal, and hoping to profit from ever increasing prices, and using my equity as an infinite-money ATM, and when that stopped, the bank took the house back"? If the latter, then what was lost is not "homes" but unrealistic prospects of profits from the thin air. If the former, I'd like to know how exactly a subprime crisis could cause something like that.
That is a very broad and angry brush you're painting with.
There were plenty of people that bought houses at reasonable prices and down-payments and still lost their ass when downstream ramifications took out unrelated businessss.
I don't doubt that, I just think when throwing around million-sized numbers it is necessary to be aware that these numbers are not describing one case, they describe a lot of different cases. There were a lot of people that were hurt by the subprime fallout (myself included, though I did not lose a home because I could not afford one at the time anyway) and also a lot of people engaging in absolutely reckless speculation (which also was one of the reasons why I was not able to afford a home at the time). There's more than one side to that picture, and some of the people who gut hurt were also the people who made it possible to get that crazy in the first place.
Actual people lived in those actual homes and ended up being actually hurt in the housing market crash.
Actual people are not relying on actual AI. And I doubt many actual people would be hurt by the AI crash.
I don't share your doubts. Even people not directly exposed to the bubble may be indirectly exposed by way of funds they've invested in or retirement accounts, or even just by having invested in something that invested in something that may evaporate with the bubble. Not to mention jobs that may disappear, which will result in the people who held those jobs being less able to spend their money, which has knock-on effects in the rest of the economy if it happens at large enough scale.
Didn't a lot of people lose their houses because they bought a house with a mortgage payment they could absolutely afford with their income, and then got laid off when the economy contracted?
And, frankly, it's literally the bank's job not to make loans that people will default on too frequently (for their own sake), so if you're not exceptionally knowledgeable about banking, it's not unreasonable to trust your bank and their advisors not to make a loan you won't be able to pay back. Like, sure, you shouldn't trust them not to screw you on the terms and with interest, but banks mostly are trying to make loans they expect to get paid back, and I would personally expect them to have a good idea of how much they can trust me with.
> And, frankly, it's literally the bank's job not to make loans that people will default on too frequently
It's not the bank's job though to decide whether it's ok for you to treat the house as a long-term asset which consumes a part of your cash flow, or as a speculative gamble. You can find a bank that will support either, but it's on you to decide which road to take. And if somebody takes the speculative road and loses, then it's not exactly the banks' fault. The adult should take responsibility for their own actions.
> it's not unreasonable to trust your bank and their advisors not to make a loan you won't be able to pay back.
No, it's not reasonable at all. Loan officers do not have a fiduciary duty towards you. They have a fiduciary duty towards the bank, so that's what they worry about - to take care of bank's interests. Assuming those interests would always align with yours is a dangerous naiveté. There are financial advisors who are fiduciaries - and you can hire one if you need - but you won't find them in your mortage bank's loan office. Yes, the bank is interested, in most cases, not to produce overtly bad loans - but that doesn't mean they care how you are going to pay it, and there's a lot of chance they'd sell your loan to another servicer in a year or two anyway. They have no duty to figure out if taking this loan won't harm you, that's your duty.
Agree with 95% of this. The only slight tweak id make is that the banks did err in assuming the value of the homes being used as collateral could not tank. Setting aside for a moment the reality of pretty rapid sales of the note to another entity, in general, it shouldn’t matter very much to the bank if you’re able to pay the loan back, assuming that their predictions are correct about the value of the home, and assuming as well that you don’t physically destroy it. Either they get their money or they get to sell the house and get their money.
Uh... neither. The former is wrong, because as you point out the housing supply isn't a matter of finance. People have homes.
But your respin is kind of a whopper too. While there were absolutely people cynically leveraging real estate to make a buck, the overwhelming majority of foreclosures in the wake of the '08 crisis were just regular homeowners. They needed a home (maybe they moved, or got married, grew up, downsized, etc... people need homes!). So they called a real estate agent and a bank to figure out what they could get, and everyone told them (correctly) that they could get a great home at a very reasonable price with very little down payment. Because everyone else was doing it. So they did.
Insert the comment everyone’s mom says about peer pressure: “if everybody else was jumping off a cliff, would you jump too?”
Everyone who took an adjustable loan, interest only loan, etc., who didn’t have an exit strategy already in place in case of inability to refinance, had themselves to blame, regardless of whether “everyone was doing it.“ I don’t mean any criticism toward people who happened to lose their jobs and would’ve otherwise been able to continue paying on the loans they’d taken. Nor am I saying it’s OK to take advantage of people who don’t bother to read or understand the assumptions inherent in the contracts that they’re signing. But people were incredibly naïve if they accepted some broker’s verbal assertion that they’ll always be able to refinance the otherwise-unaffordable house on favorable terms in 3 or 5 years or whatever.
AFAIK its theorized much of the GDP growth of the US, as well as the stock market growth of important index funds, is mostly as a result of select tech companies surfing the AI hype.
In that case, a prolonged recession may occur (that would've occurred anyway), and the effect will be felt throughout the economy.
But, again, that's just a general recession being triggered by the AI bubble bursting, i.e. AI no longer propping up the economy, so that's not a bad thing. What the results of that are in terms of severity or impact I wouldn't know, I don't think anyone knows.
Why do you think it would be a prolonged recession? Dot com bust, for example, was just two quarters of GDP decline, followed by solid growth. 2007 was much worse, because it was a crisis of financial institutions. AI bubble may be bigger than dot com, but feels more like it in that is a narrow section of the economy. Even more narrow than dot com.
[flagged]
It seems the link has been replaced with a Morningstar report instead of the blagspam trial funnel originally linked.
Agree. The reason given and examples are totally vapid.
Artificially low interest rates have stimulated investment into AI that has hit scaling limits, says research firm
He blames "low interest rates," yet interest rates have surged since 2022 to their highest levels in decades. He cannot even get the basic facts right, which kills his credibility at the start.
This also torpedos a common narrative that high interest rates are always bad for asset prices. The difference between 1% vs 5% interest rates does not factor much into VC decisions when the expectations are for 40-100+% annual returns with the hottest AI companies, which far exceeds the additional cost of borrowing. A similar pattern was seen in the the '80s and the late '90s, in which high interest rates also coincided with high valuations of tech companies.
This means a much longer effort at reflation, a bit like what we saw in the early 1990s, after the S&L crisis, and likely special measures as well, as the Trump administration seeks to devalue the US$ in an effort to onshore jobs," he says.
In an attempt to paint a negative picture of impending crisis, he gives examples, of 2001 and 1991, of among the mildest recessions ever. The US stock market and economy would go on to boom in 1995, just a few years after the S&L crisis.
If there is a job that AI needs to automate, it's these overpaid and useless analysts.
This article is referencing a subscriber-only letter.
I think this video is from the same source as the letter
MicroStrategy Partnership: “Power Plai - Julien Garran - 02 October 2025” https://youtu.be/uz2EqmqNNlE
> Make no mistake, I think that this is the biggest and most dangerous bubble the world has ever seen. The misallocation of capital in the US (which also includes housing, VC and crypto) is already 17x the Dotcom bubble and 4x the 2008 real estate bubble, and as it unwinds it will not just threaten significant economic malaise, it will threaten to overturn the entire globalist agenda, that developed with the advent of Thatcher and Reagan in from 1979 and 1982, accelerated with the fall of the Berlin Wall fell in 1988, and sped up again with China’s accession into the WTO in 2002.
The valuations vs revenue numbers definitely are alarming but the fact that it is a general purpose tech makes me reluctant to pull cash out of the market.
dotcom bubble may have been a thing, but the underlying tech was valid and created the big techs out of thin air. Admittedly after much pain & flushing out the pets.coms of the world
If we're at another moment like that do you really want to be sitting on the sidelines?
In the dot-com crash, the NASDAQ lost 80% of it's value, which is something I'd certainly like to be sitting on the sidelines for.
The trouble with a crash is that it happens fast, and you are likely not going to be able to get out fast enough. There is also the psychology involved - say you lose 20% in one day, then do you sell, or wait for a bounce back before selling? What if there is no bounce, and it goes down another 20% the next day, then what do you do?
Nowadays there are "circuit breakers" that will stop the market if it is crashing, so maybe you hear about the start of the crash an hour late, by which time the trading has been halted. What do you do now - put in a market sell order? If every one is trying to bail out at the same time, then the market will reopen at a much lower price (matching buy & sell orders), and so it goes ...
And conversely, if you try to preempt the crash, you might end up sitting out most of the run up. Lots of people called the dot.com bubble in the late 90s and got out, only to end up worse off than if they'd just held through the whole boom, crash, and recovery. (Of course, I'm sure there were also some that managed to time the market successfully! Not sure I'm that lucky though.)
Have you seen OpenAI's revenue projections?
All the financial engineering to prop up the bubble is going to end badly, it always does.
Edit: I think it's fair to say there is a fair bit antics by companies that are actually illegal, like in every bubble, that have been hidden by the mania. They get exposed as the tide goes out.
There was a nice podcast from Prof G Markets about how the bubble could collapse: https://www.youtube.com/watch?v=Oeepx2ZLrCA
In short it's companies turning to issuing debt, fuelled by increase in M&A activity, potential IPO of OpenAI, followed by collapse as tricks to increase revenue will fail to meet expectations and companies that mismanage debt will go bust.
The real problem is continued socialization of risk when the inevitable bailouts come. The government is basically telling everyone to take crazy risks, because if the risk is big enough, we'll bail you all out.
> The government is basically telling everyone to take crazy risks, because if the risk is big enough, we'll bail you all out.
This is the widespread sentiment, yes, and why the bubble is so crazy. The feeling a lot of people have is; even if shit hits the fan, rather than let stocks/assets decrease in value relative to cash, governments will just print money and keep stocks/assets "stable" while making cash worthless. The US proved they're willing to do this during COVID. What you're betting on by holding assets despite the bubble is that they're willing to do it again.
Yeah, but can they afford it? The government's debt obligations and inflation made it possible, the circumstances have changed considerably. They could wreck the economy for a long time by doing something like that again (particularly if it isn't a war or something).
The people in control of the economy are not affected by the state of the economy. There is no incentive for them not to gamble because there is no downside. If they win, they get even more money and power. If they lose, they'll still have their yachts and land and hobby rocket, newspaper, and social media companies.
I think they can afford it one more time, but this will be the last time. I think this will be the time that wrecks the economy.
Putin the magnificent has said "whoever leads in AI will rule the world". So it is a war. An arms race. Weirdly being run by private capital cuz there was no incentive(unlike during WW2/Cold War) to fund a state run Manhattan Project for AI.
Watch what happens if a breakthrough comes out of China, Russia or a shed in Nigeria. What is happening on Wall Street or who is loosing cash or jobs will drop instantly down the list of priorities. Money will be printed to the moon.
I think rather than there being "no state inventive", its that there is already private incentive, so there's not enough reason for the state to get heavily involved outside of maybe providing some additional capital. Private companies are already putting in the effort and taking on much of the risk, the government can just step in later after the results show.
> Yeah, but can they afford it?
Of course not, but with the Fed always standing at the ready to print more money, the answer is yes. The downside is persistent inflation. Inflation and default are two sides of the same coin.
Why would the government bail out AI companies? They didn’t bail out the dot-com bubble. Bailing out the auto industry was about the UAW. Bailing out financial institutions went with programs to keep people from being foreclosed on and losing their homes. The AI bubble doesn’t have those incentives and is mostly (all?) private company valuations.
I would really hope that nobody considers AI companies to be foundational to the economy the way the housing market is. I know there is going to be people crying in Congress with "Think of the 401ks!" but I really want the hype men to feel some pain this time.
It's really a shame that it's really the regular people who are going to suffer the majority of the harm when this bubble pops, as the Wall Street insiders get off mostly scot free yet again but I really don't want the government trying to prop up these assholes again.
If the economy is built on growth then AI is absolutely foundational to the economy. It is presently the fastest growing sector of the economy
That's true of every bubble. It's the best thing around until it suddenly isn't.
Don't forget that bubble popping doesn't necessarily mean that the sector is dead. The Dot Com bubble popped but people are still selling things on websites. It just means the hype and unrealistic expectations come crashing down to Earth. AI is going to be with us from now on, but it won't become a God with unlimited productivity next year.
https://www.derekthompson.org/p/this-is-how-the-ai-bubble-wi... Goes into good detail on how the bubble has infiltrated everywhere. Like if you went into a REIT thinking it was safe from the bubble, nope a significant portion of your investment is in the data center buildout
I thought data centers were always a part of commercial REITs though. Is it just that they've grown as a proportion of REITs?
It seems like an (ugly?) truth that if you don't want to be exposed to bubbles, you have to take lower-return investments, by definition. Because the bubble is raising the high bar of what kind of return people expect on investment, so if you try to get close to that, you're going to end up affected by the bubble.
The problem, of course, that no one wants to take a lower return on investment.
Yeah, but I guess the point is that it is so big you can't escape it, it's everywhere. So the fallout will not be limited to a segment of the economy, it's going to hurt everywhere.
And the reason nobody wants to take a lower return on investment is that doing so is trying to time the market.
Let's say you heard Trump's talk of tariffs, got spooked when he won the election, and moved your investments. The S&P 500 has gone up ~33% since election day. How much growth can you realistically miss before correctly hedging against a crash still net loses you money?
And that's assuming you pick the correct hedge. If the US economy well and truly crashes beyond return, to the point where the S&P 500 is actual garbage that will not recover in the next 5-10 years, that might bode poorly for the dollar, so that's it for your treasuries, money market funds, CDs, etc. Gold is always an option, until it's at an all time high because just as the economy is roaring along a ton of people are nervous. So, you can invest in gold, but it's expensive, which means it could actually dip if we get stability. Ok, so you invest in foreign markets - but the Global Financial Crisis was global, the Great Depression was global, everything was global.
What do you actually hedge in? Oil? Well, the rapid development of green energy might actually, finally render that an unsafe investment. Green energy? Think again, the president is on a crusade against it. There's literally nothing that's actually a safe investment, and even if you did find a safe investment, the big crash might not happen until everyone else riding the wave has doubled or tripled their investment, by which point they'll be on par or even above you post crash. I'm actually very happy I'm nowhere near retiring, since regardless of how else you feel of the current governing of the US, I think it would be hard to argue that they're fighting a war against conventional wisdom, and that makes investing very scary.
I can’t see the case where we won’t need more data centers in the future even if AI is a bubble as more work load is generated and even if it is a bubble. There are real world use cases where you will need ML and inference even if you don’t need to spend money on training.
> I would really hope that nobody considers AI companies to be foundational to the economy
The way people in the US government has been speaking and acting goes against that hope. Instead, it seems to be unanimous that AI will be the foundation of the economy, the US can't lose its leadership on it for even a second (because when it takes over, it will take over in a second), and it's worth breaking every rule, spending every cent to keep, and cannibalizing every other industry if necessary.
Honestly, you have to deal with the delusional con-man in power ASAP.
Maybe though, if that happens, some heads will roll and some of the white collar criminals get decades in prison for their behavior unlike in 2008.
Why do you think that will happen, when it pretty much never has historically happened?
That's probably less likely than ever.
I think you could make an argument either way and have an equal chance of being right. Either Trump et al get bribed/flattered/are personally exposed to the fallout and bail out their fellow class members or he gets angry about it and decides to make heads roll via his increasingly captured and controlled DOJ. I still agree it's more likely the former than the latter but I can see the path that leads to the latter.
Not if they bribe the big man
It’s not a felony to bankrupt your company or lose your investors’ money through good faith bad bets.
Presuppose everything else that contributed to the GFC was straight up fraud: Prevent it all and the crisis is basically just as bad.
Maybe I'm reading too far but "heads will roll" has connotations outside the strictly legal sense
"Heads will roll" just means "severe punishments will be administered". It doesn't necessarily imply the punishments will be illicit, illegitimate, or even fatal.
It came out soon after AOL acquired Tine Warner in 2001 that they artificially inflated income to prop their stock up for the merger to go through. No one was punished for it.
I'm shocked, shocked to find that gambling was going on there.
One of the best ways to be in a bubble is to live in the middle of transactions.
Here's an useful exercise. Suppose someone makes an advance that makes LLMs training require 90% less compute, but does not make them any smarter. (Recent results from China indicate substantial compute reductions are possible.) Demand for GPU time drops. What happens to the economics of the AI boom?
Demand most likely won't drop due to Jevons and the worst will be big labs and Nvidia probably drop significantly in valuation but still pretty good. Also recent china results only show that you could build cheaper model hitting the benchmarks not actually generally competitive irl.
Agreed, with one caveat: demand for hardware can drop if the optimization is so substantial that deployed hardware is already capable of fulfilling any possible request in a timely fashion. Say, anyone becomes able to run any state-of-the-art model on any desktop computer. 90% is not enough, though. It has to be at least three or four orders of magnitude. The ones driving demand are big datacenters, so they need to be in a position where just throwing more hardware at the problem solves nothing.
I remember something like that happening. Someone optimzed something and boom, one billion of valuation gone. It seems ridiculous on the surface, and definitely smells of a bubble to me.
That's happening to the automobile market in China. Electric car and battery manufacturing costs go down with scale. Way down. Huge electric car glut in China. Car prices crashing. Government worried.[1]
Profit margin of China car companies is down to around 4%.
[1] https://www.cnn.com/2025/09/26/cars/chinese-electric-cars-pr...
I think the problem is bigger than that too. If you have a problem set where you cannot afford any kind of mistakes LLMs are never going to be the right tool, they are never going to get to that stage.
If you have a problem set where some mistakes are tolerable however, all you need to do is slightly increase your tolerance budget for mistakes and you can run something like deepseek locally at a tiny fraction of the cost.
Which means these giant data centers are useful for training but the economics of running ongoing inference for customers doesn’t make long term sense for the vast majority of people.
Which leaves you with a huge huge huge financial problem, one which may have devastating large scale impacts on the economy.
The vast majority of energy is already used at inference time. This gap is only going to widen as the number of applications increases and people ask their LLMs to do more complicated tasks.
If training costs go down, I expect models will be trained on more video, so training energy usage will not decrease either.
The mistake this article makes is assuming LLM scaling is one thing. It's not.
RL is spiky. It produces narrow improvements on specific capabilities. They're not making the model generically smarter, they're RLing in holes in the model's capabilities. In reality we don't have one scaling curve, we have thousands of them. We're in diminishing returns in "top line smarts" but we're raising the floor in a wide variety of areas that people who don't heavily eval models for a living might not notice.
I did not know GPU backed loans was a thing. It seems risky.
It's not just, risky, it's insane. The GPUs lose value almost as soon as they're installed.
On the other hand, auto loans are a thing. It’s all a question of how the assets are valued.
Do you mean car title loans? Because that's basically what the GPU-backed loans are. Car title loans have terrible terms due to the diminishing value of the car. These GPU-backed loans are being booked at much, much more favorable terms, as if the GPU holds its value for a long time. This simply isn't the case, so if the loan is defaulted on, the loaner will have a rude awakening when all they get are worthless GPUs.
Isn't that part of why down payments are so necessary? It takes the loan amount below the immediate depreciation amount of the asset so the bank/lender is less exposed to the risk.
I’d imagine the main benefit of downpayments to be behavioral, since otherwise the borrower has no stake in the asset initially.
I doubt that's the main driver for lenders and people are generally pretty attached to and excited about their new cars. They also do have a stake they'll still be out time and money even if insurance covered the entire cost of the vehicle.
I don't care as long as the bank losing on it wouldn't be bailed out with my money. Which unfortunately is a common thing now.
I've been playing with Llama 3.2-3B this weekend and, while I haven't spent enough time with it to figure out the limitations, it's definitely good enough to be useful. That's running at "as fast as I can read" on a £100 Intel Arc GPU. I'm tempted to buy a Arc770 (£250ish) to see if I can run the OpenAI open weights models...
I think it's reasonably likely that these massive investments turn to dust when they get squeezed by the commoditization of LLMs.
The Saudi sovereign wealth funds and similar Middle Eastern investors, Masayoshi Son, a16z (though not all of their funds), Tesla's board, and others seem to be operating in some kind of alternative reality where watching out for your limiteds and shareholders isn't a factor. Do they think they are really TBTF? Is that why JD Vance is where he is?
It's not just the size of the bubble that's scary. It's that some people obviously think they're going to get away with it.
> It's that some people obviously think they're going to get away with it.
Cynical, I know, but they absolutely will get away with it. The wrong people will be punished, as usual. They will claim that "there was no way to know" this could happen.
They will do what they always do, which is to vaguely imply that if they are left to the consequences of their gambling addiction then China will get ahead of "us". Note that the general public are only included as part of "us" when Sam Altman needs the general public's funding.
Hopefully their grift fails when china also doesnt have AGI or anything like it. But idk, people can be delusional
So as someone investing for hopefully an eventual retirement, what am I supposed to do?
If you believe the rule of law and ~general financial stability of the US will persist just keep on investing regularly and ride it out. Markets have a long history of bouncing back and they will keep doing it until they don’t. When/if that happens your retirement dreams might seem quaint compared to the global socioeconomic reality. The people who could sustain the most near term pain are those who have already retired and are living in a fixed income.
If the US collapses, conversely, stock/bond market investing choices won't matter.
And frankly I wouldn't trust my foreign market investments managed by a US company to still be there either.
Also I'm not sure that foreign markets would be that thrilled by a US collapse either, at least not immediately.
And, unfortunately, a sandwich-rich portfolio doesn't really help in the long run either.
Don't do stupid things, don't bet everything on a single prospect, don't listen to random people on the internet :)
Don't be employed by anyone who does; don't work with companies or industries that do; live in a shack and poo in a latrine you dug yourself.
In the end, DCA will prevail.
Depends: when are you planning to retire?
20 years hopefully
Spend less, save, put it on index funds. The bubble popping will mess you up a little but time in the market > timing the market
live long enough to retire on the upswing
if the upswing doesn't come our lifestyles are all screwed anyway
Like everyone else, wait until the very last second, just as the market peaks, to unload your assets. Before the institutional investors and insiders beat you to the punch.
The AI hype bubble now represents something like 20-30% of the stock market depending on who you ask. For reference the great depression started with a 24% drop of the stock market. The people running this AI game know there is 0% chance the government will let this play out naturally because it would cause another great depression. Bailout Guaranteed.
The plebs can take some inflation,higher taxes and national debt increase for the hucksters to get some yachts and lambos. I mean what were you going to do with that money? Feed your family? Pay rent? Jeeze get a life....
In this round of con there is not even a way for an average pleb to get any scraps since the AI bubble is almost completely hidden behind private equity. Except maybe Nvidia stock.
The bubble is starting to froth/pop IMO. An incestuous cycle of players propping each other up with circular investments(ie I'll loan you money to buy my product so I can sell it to you) began last month with Nvidia "funding" OpenAI datacenters. Actions like that mean they are out of external ways to keep shuffling the debt. Like when WeWork CEO started loaning the company its own money to buy its own product to rent.
Edit: Oh $hit AMD just did the same thing with a circular funding deal with OpenAI. https://news.ycombinator.com/item?id=45490549 Channel Stuffing Money printer go brrr. I can't belive there is even discussion if there is a bubble at this point. Its literally wolf of wall street "Never let them cash out that makes it real" style deals all over the place.
> An incestuous cycle of players propping each other up with circular investments
Of course I'm reminded of today's announcement about the strategic partnership between AMD and OpenAI which caused AMD's stock price to jump a whopping ~35%.
AI isn't the only component of the bubble. SpaceX which has a genuinely successful business with Falcon 9, is also a huge bubble. SpaceX launches an amazing 25 times as many rockets as ULA in a year. But 2/3 of those are Starlink launches. No outside revenue. And SpaceX is valued at about 200X the valuation of ULA.
> now represents something like 20-30% of the stock market
> the AI bubble is almost completely hidden behind private equity
How are both of these statements true?
The AI hypy has inflated the values of the "magnificent seven". Meta,MS,Appl, Amzn, Nvidia ect. Retail and various professonal investment chasing indirect investment in AI,ie MS owns almost half of openai. Also 401k rebalancing. Are driving the valutation of those public stocks to very unbalanced valuations with zero increase in business.
Its complicated. Though you could argue its all good business. But first buy my swampland in florida please.
> I'll loan you money to buy my product so I can sell it to you.
I predict this will happen among Nations as well. US will provide money to Japan so that Japan can invest that money in US (at Trump's discretion) so that everything is kosher.
That is what traditional US foreign aid was all about. We give money to allies, they buy our military hardware or farm products. Trump of course has shit all over the system, so who knows what we are going to do now.
We also give out tons of subsidies and tax breaks to lure foreign investment to the US.
This article is pretty low quality IMHO. Lots of conjecture about the value of LLMs. We are far too early in this hype cycle to know what will come of this new tech. There is, however, no doubt that incredible amounts of capital are flowing into data center construction and much of that capital allocation will end in tears. There will be losers.
If bitcoin has not popped so far, AI will never pop.
Bitcoin is propped up by individuals spending discretionary funds no one will really hold them accountable for
AI is pripped up by corporations spending capital expenditure that shareholders will want to see a positive ROI on, otherwise they bail.
So the AI bubble will definitely not be as resilient as the crypto bubble
Bitcoin needs less resources than AI, and Bitcoin totally failed its purpose as a currency.
Whether it failed is (unfortunately) beside the point. Number still go up.
Bitcoin not popping is a testament to human delusion. It cant function as a stable store of value becuase theres nothing preventing if from hyperdeflating, and yet people are still buying it. What are they even buying? The deflation based gains? Deflation based on nothing but speculation? It doesnt do anything. Its madness
I dont think the GenAI bubble will be as resilient because its not being funded through discretionary expenditure. Its being funded with capital expenditure. If your cousin spent his savings on bitcoin and isnt getting any benefit from it (or even loses it in a drop) thats fine because millenials dont have savings anyway. But if a company spends on AI and doesnt get a positive value prop in a while, they will pivot and pull their money out, because shareholders will get mad
In the high 90%s of ownership of Bitcoin is owned by ~6 entities.
No it's not, because nobody knows how big the AI bubble is. Or even if it actually exists at all.
Because the concept behind a bubble is that it's inflated over the true value due to speculation or non-speculative overinvestment.
But there's still a true value.
So you can say AI valuations are whatever multiple of previous whatevers, but you can't say the AI bubble is. Because if there's an AI bubble (and there probably is, but nobody knows), nobody knows how big the bubble part is versus that solid part. Obviously, if anyone did, they'd make lots of money by investing accordingly.
And of course the reason we can't figure out the bubble size is because nobody knows what the true value of LLM's will be over the coming decade or two. It's hard to know even what the order of magnitude will be. And how much will be captured by existing LLM companies.
Is anyone acting on this? E.g. selling index funds, moving to HYSAs/bonds, etc.?
I would say if you're holding a lot of money in a S&P 500-tracking index fund, switch to holding a total stock market fund instead. It's probably a bad time to hold SP500 right now, but you don't want to get out of the market entirely because there's no way to know when the bubble pops. So hold a total stock market fund instead, so that you don't have an outsize amount of your portfolio invested in the companies leading the SP500 (which are the most likely to lose the most when the bubble pops).
I agree with using a TSM fund rather than an S&P 500 index, but be aware that you're still mostly holding S&P 500 companies. The S&P 500 covers about 80% of the total market cap of US public companies, whereas the Russell 3000 is 98%. (This is, of course, because the 500 largest companies are so much larger than the median.)
If you really want to diversify away from the biggest growth companies you could try a value and/or small/value tilt. Those might all be good advice, but they do come with their own caveats, whereas using TSM instead of an S&P 500 fund is just good advice in general.
The weird thing that makes it different to subprime is that AI is real and really could become AGI. But it might also be overvalued.
So it's not like there isn't a product (there is) and there's no growth prospect (there is). But it is scary how much now hangs in the balance of one bet.
It almost feels like it's going to end badly either way. If the Great AI Bet succeeds, a tiny proportion of the world will own all intellectual power. But if it fails, the impact of the write-offs on the broader economy will be terrifying.
> AI is real and really could become AGI
Every single researcher not being paid $$$ by AI companies say there is 0 path between LLMs and AGI, but sure... and the next pfizer drug might make us immortal, who knows, everything is possible after all
I dont think we'll have AGI in 2 weeks like altman does, but its not fair to say its impossible.
LLMs also came out of nowhere, a series of discrete improvements that finally got over the hurdle and acheived an unprecedented functionality. Absolutely no one predicted its emergent capabilities back in 2016 when the transformer paper was proposed.
They didnt see scaling then, they might not see the next thing now, until its found.
So it wont be in the next 2 weeks, and it wont be from OpenAI, but it might be in 10 years from some random researcher at waterloo, or tiktok
> its not fair to say its impossible
Nothing is impossible!
> LLMs also came out of nowhere
No they didn't.
I think this is a disingenuous take.
Will AI of today definitely become AGI of tomorrow? No, for sure not, and anyone who claims this is at best crazy.
But is it imaginable? I think totally. Andrej Karpathy' blog post about RNN writing Shakespeare 1 character at a time was 10 years ago. GPT-2 was released 6 years ago. In that time we went from something that barely speaks English, never mind any other language, to something that, on a good run, is an excellent language tutor (natural and programming), can play games, can write moderately complex programs, goodness knows what else. For some people, the romance of a ChatGPT-4 was unmatched.
Even if it doesn't become "AGI", it might just get so good at being sub-AGI that the question is irrelevant. We're seriously contemplating a near future where junior devs are replaced by LLMs; and I write this as an AI sceptic who uses LLMs to write a lot of the kind of dumb code a junior dev might do instead.
I don't like AI, in that it nibbles away at my competitive advantage in life. But it's IMO crazy to pretend it is not even potentially a game changer.
The next logical step for cars were flying cars, the next logical step for planes was space travel. Both were hyped not so long ago, people believed in both, it never happened and it isn't closer to us now than 50 years ago
I'm not saying it doesn't bring any value, I'm just saying that if you think we should give $7 gazillion to Altman because he's building skynet by 2030 you're smoking crack
> Every single researcher not being paid $$$ by AI companies say there is 0 path between LLMs and AGI
So maybe it's jealousy?
I mean I don't know.
Ah, the same reason climate scientists not paid by oil companies agreed on global climate change, and doctors not paid by cigarette companies warned about lung cancer?
I meant it as a joke. Should've added an /s instead of "I don't know".
But in all seriousness until someone else invents AGI some other way, you can't disprove that LLMs aren't the way. My intuition says you need more than LLMs but I could very well be wrong.
The goal posts will constantly shift just like Self driving. AI used to really mean AGI but that doesn’t get investments.
AI always meant AGI.
AI also never meant AGI.
Artificial Intelligence is an entire discipline of the Computer Science field. It encompasses everything from how Pinky, Blinky and Clyde chase Pac-Man, to A* search and similar pure algorithms, to machine learning, computer vision, and LLMs.
It is also a term used widely in popular culture and media to mean, essentially, AGIs—Cortana, Agent Smith, C-3PO.
The problem is not that this term is very broad, and it certainly isn't that it has come to mean something that it didn't before. It is that a bunch of people with a financial interest have been busily trying to convince the world that LLMs are Cortana.
As long as people keep spending all their money on LLMs, we are safe from an AGI.
> AI is real and really could become AGI
There is zero evidence that current "AI" (LLMs) are ever going to become "AGI".
If you want a pathological liar for an "AGI", then sure, LLMs are already there.
What would count as proof of AI trajectory to AGI?
Currently we have: - cross domain competence - composable tool usage - constant improvement without clear signs of stagnation...some can count gpt 5 as not much progress but there are still world models with huge gains compared to year before
There is a bunch of things missing for sure like mechanistic reasonic, better context lengths, determinism, better world modelling, continuous learning.
The bet is crazy and whole world is gambling on AI, right now, because of those signs of potential. That's precisely why we landed here, the evidence was good enough for big tech to gamble on it...
>What would count as proof of AI trajectory to AGI?
It depends on who you ask. There is no absolute for definition of "AGI". Sam Altman defines it in monetary terms, because that benefits him. I have a very different idea of what "AGI" means. It's really very subjective, so I don't have a definite answer for you. I'm sure you'd define "AGI" differently than I would, so having this discussion is kind of a waste of my time.
The peo1ple pouring money into "AI" are doing so in the hope that it will become more reliable someday. My educated guess is that it won't, due to the underlying mechanisms that it is built on. Predicting the next word in a sentence according to grammatical rules is a long, long way off from a machine knowing and understanding how truthful the resulting sentence is.
I describe it as both a bubble and the next industrial revolution. It's not so much inflated as it is far ahead of itself IMO. Kind of like NVDA stock over the past nearly 3 decades.
The talent out there is all focused on a tiny number of tasks and benchmarks in service of the AGI cult whilst the real gain is scattered amongst everything else with no staffing to build it. So I see a bust much like the dotcom bubble in the arbitrary future, but then it corrects surprisingly quickly shortly thereafter and the engineers of that bubble will have gotten out temporarily well ahead of that bust, buying back their assets at 50% or greater discounts and the broader base of retail investors are screwed as usual.
As for impacts on the economy. Meh. They're already starting to ignore the ramblings of the Orange in Chief (the guy shuts down the federal government and the entire stock market yawned last week). The weakly efficient market abides and it's been burned enough times already by the TACO trade. It'll get through this. But oh the whining that will ensue on the corporate media.
Real estate is real too. People need houses to live. Websites were also real value
The dotcom bub, subprime crisis, and AI bubble are all based on real goods being overvalued to hell and then crashing back down to their actual capabilities. The soze of the bubble is the delta between their actual value and the market's percieved value.
In that regard i think AI will be crazy crazy valuable in real terms but not like the market is using it. I think the real value of AI agents is very very low and the market will crash to that level. I think the value of genAI as much simpler interfaces for communication (RAG, translation, NLI) and for automated understanding of static sytems (rather than acting in dynamical systems) is high and will crash a bit before the market learns to use them right and then itll be a party
You can’t buy stock in AI. You can buy stocks in companies.
The internet was destined to be big sure during the dot.com but most companies crashed.
The bubble popping issue would be that there isn’t a good way to recover the capital used to build the AI models.
You can't(couldn't) buy stock in the internet, either. It's always "stocks in companies".
I mean, unless you go back to "tulips are a sure thing; their prices always go up!".
But buying stocks in an S&P massively invested in AI is essentially the same as "buying AI", when the bubble pops.
It might just be very similar to the dotcom bubble.
The trillion dollar question isn’t “is AI a bubble”, it’s “which of these companies are pets.com and which are Google (if any)”.
Even investing into a basket might not be a winning strategy even in a world in which AGI is imminent.
It might be. By feel, it is different in that a lot of the AI investment is into real things, that hold non-speculative value: data centres, GPUs, training marketable and genuinely useful products. It's not food for pets but online and sent by email.
But otherwise agreed.
But that’s my point: The dot com bubble burst, yet the Internet still changed everybody’s life.
Buy precious metals.
> ChatGPT-3 cost $50 million, ChatGPT-4 cost $500 million and ChatGPT-5, costing $5 billion
Is that actually true? And is most of it because of the compute requirements of the models or scaling cost due to exponential growth in usage?
I hope it didn't actually cost ten times more to create ChatGPT-5 than it did ChatGPT-4.
I think that's mostly training costs. There is a serious level of diminishing returns and the primary way companies have been trying to get ahead is throwing more and more hardware at it. AI already burns something around 5x as much power as crypto mining and it's only growing.
You hear sometimes about the AI singularity and how they will continue to become smarter at an exponential pace until they're basically gods in a box, but the reality is we're already well into the top of the S curve and every advance requires more effort than the one before.
Those numbers do in fact match what a quick search says.
But I can't find anything directly released by OpenAI, so maybe these are all just estimates. And presumably include price of hardware.
You should be more concerned that Chinese labs can train models that are just as good for 10X less because Americans treat the USD’s status as the global reserve currency as the ultimate bitter lesson. Who needs better math and engineering when you can print money to buy more GPUs???
The author literally said gpt is spending 10x more for equivalent. This really means ChatGPT had that intelligence at that cost a year or two ago. Smaller domain models are better at focused tasks in that area but can’t be generalized.
Nvidia certainly has a headstart, and great products.
But many (many) labs around the world are working on alternative chip designs for math processing.
Once a couple open-source chip designs come online, that can compete with Nvidia, it will all come crashing down.
Think Android vs iPhone.
Stoked.
Has Apple crashed?
At least it has an open-source capable competitor.
Surely you aren’t talking about Android as “open source”. To a first approximation no one wants a phone running only AOSP without Google’s proprietary parts.
For definitions of "no one" that include 1 billion people.
So yes, if I’m willing to go to China and use Chinese OEM customized phones with Chinese apps, I will be golden…
You're free to use the same Android that the Chinese OEMs (and Amazon) built their phones on yourself, without traveling to China. It's true that a reasonable definition of "no one" wanted Amazon's phones, but a much more lenient definition has to be used for the Chinese OEMs.
And in any other country besides China, how many of the apps that most people use don’t have a dependency on Google Play Services APIs?
Google has been moving much of what a western user considers “Android” out of the AOSP code since 2013.
you might want to recheck your measurements
I prefer to think 3DO, Atari Jaguar, Sega Saturn and Apple Pippin versus the Playstation myself.
Nvidia is way more than their data centre hardware - their software and research is just leagues ahead of the nearest competition, and their edge compute capabilities are going to be a rapidly growing market as derived models start getting put in products - think self-driving, robotics, machine vision, manufacturing and industrial automation, on and on.
So yes - while there are others who may yet compete in the data centre compute market, nothing else comes close to the monopolistic total vertical integration nvidia has built over the last decade.
On the one hand, I'm like: so what, I got no skin in this game... But then again, they will most assuredly get community-money bail outs.
And _also_ raise prices on GPU's.
I lived through dotcom boom and bust, and there had been a lot of craziness, but when the dust settled, we have now a functioning internet economy, and I am able to sit at home in my pajama pants working for an international company that has employees all over the globe, while enjoying many other conveniences the internet brings me - and all this is enabled by what started during dotcom era. It is true that it did not match the most wiled-eyes predictions, but also a lot of things did change, and a lot of things did improve. If the AI boom followed this trajectory - which is of course no guarantee of, but if we assume that - then there is going to be a lot of crazy stuff to come, but at the end of it, the dust will settle and we will be mostly better off from it. Some people would become billionaires, some undeservedly so, some people would be ruined, some also unfairly so, but for most people, at the end it won't be that huge.
100%
We are in the era of people using geolocation to remind them about their grocery list when they're near a store. In 10 years we will be in the gps and uber era.
Its just that it will be bumpy when this bubble crashes. And the US isnt as stable and sane as it used to be. I wonder if the economy can be steered through waters this rough
Are those numbers inflation-adjusted? I feel like inflation in market caps/valuations was a lot higher than in consumer prices
> Artificially low interest rates have stimulated investment into AI
So chatgpt was released in nov 2022. Interest rates started going up shortly after.
Am i missing something, it seems like the AI bubble and very low interest rates don't overlap except maybe at the very beginning.
The troubling part is we are in the AI bubble whether we opted in or not.
The downstream effects are everywhere - real estate (data centers), energy (power consumption), education (automated tools), employment (tech layoffs), and equity markets (Nvidia propping up the spoos). It's not one sector this time, it’s systemic adjacency.
During the dot-com crash, retail got burned. During subprime, it was the global financial system. Today, AI exposure is distributed among mega-cap public companies, a16z-style VC funds, sovereign wealth, and shadowy circular debt deals between hyperscalers and AI startups.
Even if you are not betting on AI, your retirement fund is.
[dupe] https://news.ycombinator.com/item?id=45465969
Thanks! Macro-expanded:
The AI bubble is 17 times the size of the dot-com frenzy, analyst says https://news.ycombinator.com/item?id=45465969 - 2 days ago, 94 comments
There is some higher quality discussion there, and still ongoing.
You only know it was a bubble after it burst.
Right now, I don't know if it's a bubble, and neither do you.
A famous economist who has predicted 17 of the last 3 recessions...
dupe: https://news.ycombinator.com/item?id=45465969
Actually no-one can see a bubble, that's what makes it a bubble.
What's the mood in the bay area? If this is 17 times the size I can't imaging how much optimism/fun you all are having.
I was there for the first dot com boom. So much optimism. It seemed every party someone announced they were pregnant we were all so optimism for the future and comfortable, ready to start families. The restaurants were all full with new spots popping up all the time. It was a really fun time to be there. I know there are some layoffs but at 17x even with some layoffs I can't imagine 17x more energy/money then back in the first dot com.
But what if AI succeeds and it’s the rest of the economy that implodes? Long-term, I’m sure AI will lead to massive productivity gains. Short-term, I think it’s going to be chaotic. I don't know which will implode, AI or repetitive white collar work or something else
Worse than the bubble is what's happening to the rest of the economy. If you remove AI related spending the US economy is trending in a really bad direction.
This bubble popping will definitely take down crypto with it and rip through other adjacent industries.
How can the AI bubble burst in the public markets? It is circular investment within silicon valley. Even if AI proves nothing, overall the tech market is in the same spot where it started
Not all AI investment is private. You could argue that public companies like Google, Meta, and Microsoft have had their stock appreciate at least partly because of the AI frenzy.
TINA. There is no alternative. People are forced to play the game.
Has anyone articulated a theory for what the profit model might actually look like for the big LLMs? In the .com era, the valuations were crazy, but people generally understood how you could eventually make money on all of this stuff.
If models are requiring larger and larger infrastructure buildouts, does anyone have a clear sense of what users will have to pay in order to make the businesses profitable?
It seems that each model generation, and each generation of GPUs, has a limited shelf life, so you need to make more money from each generation of models/GPUs than it cost you to build/buy them. User pricing to make a profit depends on inference volume. You need to make back a fixed dollar amount before breaking even, which could come from high volume and low prices, or lower volume and high prices.
The trouble is that everything is changing so fast that any kind of forecasting is extremely error prone, especially when one forecast builds on another. First you need to guess how much more capable the models are going to get (at things where people will pay more for better performance), in what time frame, then guess what level of demand (inference volume) will exist with that level of capability...
The LLM developers like OpenAI and Anthropic like to tout things like Math Olympiad and Competitive Programming results as signs of progress, but there is no guarantee that they will be equally successful in applying RL to more general areas of commercial value where RL rewards are harder to define.
These companies also like to talk about "scaling laws" is if there was some inevitability about investing more money & compute and getting better results, but this only works until it does not, and they replace one broken "law" with another. Right now it's all about scaling of RL-training and test time compute, but how long will that last, and what type of problems will benefit?
The profit model here seems a bit like the Drake equation for calculating the probability of other intelligent life in our galaxy... it may be possible to define the equation, but the outcome depends on having the right values for all the variables, which are largely unknown.
With the direction of OpenAI, hyper personalized ads inserted directly into chat and their app experiences could be a path. Not saying it will work, but they’re definitely exploring it.
The dot.com had no idea. They talked about eyeballs.
I think they had an idea that it would be fees per transaction or a subscription model. They just had no customers.
no one can predict the future, they were guessing then and they're guessing now
guesses in the past look better because we tend to pay more attention to the correct ones
The boldest claim isn't he boldest claim first - it's not just that AI is in a huge bubble, it is that interest rates are artificially low.
and the US has a president that's seemingly doing everything he can do make them lower
Who gets hurt when the bubble pops? And who will the government need to bail out?
Last time, a lot of the companies were public and the general public saw stock losses. This time the only companies that are publicly really exposed by the AI bubble are Nvidia with all of their circular financing and Oracle. Of course Tesla has always been a meme stock.
Defined contribution plans - for now - can’t have private equity in their funds. Defined benefit plans and endowments are exposed.
Apple famously hasn’t invested that much in AI, Google is spending a lot on infrastructure. But between search and GCP and YouTube they have a real business plan and are funding based on profits. Amazon is in the same boat. Microsoft is bowing out of spending money on training and focused on inference - and they also have Azure. Meta is making money using AI for ad targeting and probably in the future to generate ads.
I can also see consulting companies being hurt (I work in cloud consulting) as businesses are throwing money at them to “AI enable” their business.
> Who gets hurt when the bubble pops?
Given how investing is heavily promoted by all these neobanks I have a feeling a lot of people will get burnt. Back in the days, not even 10 years ago, you had to research and go out of your way to invest, now you can do things like "automatically round up your transactions to buy NVIDIA" from your bank app. The only ways to get out of the middle class are: lotteries, crypto, putting everything stock market for 20 years and living like a student in the meantime.
Most people in the middle class can’t afford to put any decent chunk in their 401Ks let alone invest any meaningful amount above that.
But funny enough, most of the people I know that do invest in individual stocks, bitcoin etc are people in the servjce industry who are single and make decent money on tips. I live in very heavy tourist town.
But since this is a site of tech heavy participants, if you are a software developer or adjacent, you are on average making twice the median local wage for your area if you are in the US even as a enterprise Dev 2-5 years out of school and should be able to invest at least 15% if your income.
I would rather be in AI bubble than Crypto currency bubble.
I wonder how the tech job market would look without the AI bubble. Is the AI the cause of the unemployment crisis(if we believe the CEOs who layoff a quarter of the company because "AI is replacing thousands of developers") or if we didn't have this crazy AI bubble, the market would be a much worse bloodbath.
now do crypto
Some day, some wizened legal types will prepare their cases as these bubbles take shape.
Instead of waiting for a warmed over post-mortem, long after the bubble pops, before even considering their options.
It's a bubble. Of course there's malfeasance, lies, corruption, etc. Assume criminality.
These endless boom-bust cycles will continue unabated until there's credible threats of doing some hard time.
> It's a bubble. Of course there's malfeasance, lies, corruption, etc. Assume criminality.
Fun fact, we live in a society with rule of law. You can't just assume criminality because you don't like something, you have to actually prove it.
Which bubbles have been free of criminality?
A cynic, though certainly not me, would argue that "bubble" is an euphemism for fraud.
Which non-bubble period of time has been crime free?
Its very unclear to me if there is actually more crime during bubbles or if people want someone to blame so the powers that be investigate harder until they find someone.
We do? Have you been reading the news where the President is firing AGs who won’t go after his enemies and firing FBI agents who didn’t agree to perp walk his enemy? Not to mention he just pardoned 1500 domestic terrorists in January.
Public companies are outright bribing the President to get mergers passed (Paramount) and the entire confiscating TikTok to give it to his buddy at Oracle on the cheap can’t be ignored.
On the federal level, there really isn’t any legal standard anymore.
"The rule of law" is the rule of the economic elites over the weak plebs. The "rule" is formed, shaped and ignored so they can have their way. Anything goes when you operate on the upper levels, we are not naive kids.
Disclaimer: I am not a doomsday apologist, just a Nerd collecting, analyzing data and doing projections out of it.
There is not only an AI bubble, but a US Dollar bubble, too. Most people don't get it what "Stock markets, gold, real estate, crypto, EVERYTHING is up" actually means: That the currency you are paying in is losing value at rapid pace.
I do not yet understand how the market manipulation actually works, but if you exchange from an Asian country to US Dollar right now, an amount that will buy you a excellent dinner for the whole family no longer even buys you a hamburger in the US.
The part I also do not understand is why the only thing that right now is "cheap" as seen from the US perspective are foreign currencies.
I also don't understand why other countries are not liquidating their USD reserves. How can you as a country not see that your reserves now would be much safer in Asian countries including even the Chinese Rimibi?
Well, maybe the answer to the USD losing value rapidly is that in reality there ARE market actors in the background moving to other currencies, but are really really good at hiding it.
But in any case: No doubt, the western part of the global financial system will be crashing within the next 3 weeks. With the US being on the brink to civil war anyway, you should be able to extrapolate what a financial crisis and bank run in the US in combination with everyone and their dog owning war-grade weapons will end up at.
Right now you have heaps of people in the US who are completely ignoring any data but invest only based on being member of a fanatic cult - every time you read news that Tesla no longer is able to sell cars, or that their robots do not work, or that their Robotaxi stuff is a scam, the Tesla share value goes further up. The same applies to AI. That AI will never make any profit is no longer niche knowledge, it's headlines at Bloomberg, WSJ, FT & co. You have to actively ignore that information - still, AI stocks go up.
So what is missing for the bubble to finally burst? Anything that makes those cult members start to doubt. Anything that triggers "Maybe putting all my retirement savings into crypto is too risky" or "Maybe Elon actually isn't the messiah", "I put all my money into a Trump meme-coin. Is this really more important than getting my teeth repaired?". Or anything that triggers those sheeps that are invested into this to simply needing cash quickly. So, for example due to a natural disaster, or some pandemic, the Epstein stuff finally blowing up on Trump etc.
The list of things that now may trigger the implosion is gigantic. Due to that it is possible to predict where it will start - it might not be the AI bubble. It very well even may be something rather classic like the "Deported migrants no longer pay towards their mortgages / loans" stuff happening right now.
But in summary: Due to the very high number of potential triggers it is stunning it hasn't imploded yet, and if you read/watch interviews from scientists in this area, the common theme these days is: "WTF is going on here?! This should have imploded LONG ago!".
[dead]
Buried lede: what is the transformative potential of ML/AI?
It's not a bad bet that many are making that "changes everything" is real in this case. Hitching a ride to the technology getting civilization over the hill is a reasonable goal.
But what institution, technology, play...?
Following the money shows where bets are being placed. But none are safe, the disruptive consequences of network effects make diversification appear wise. The industry is littered with the bones of giants.
Personally I am looking at ML/AI as analogous to the Great Oxygenation Event. What things look like—at least, on land—afterward, none know, least of all the cranky old oligarchs who are hell bent on consolidating control and ownership.
If you marvel at the output of generative AI be it pictures, logo, music, 3D models, prose or code - you probably are yourself way too under-skilled to be gainfully employed or not yourself capable of generating similar or better output.
Conversely - any expert (prolific writer, coder, painter, photographer, videographer or log/web designer) isn't as much amused or going to scream out of excitement that what these models are producing (vides, pictures, logos, essays, code) - they could never ever have thought anything better than that.
This fact alone is enough to warrant that a big bust is coming. Not a matter of if but when.
Possibly - or we might discover that for a large amount of use cases, AI output is "good enough," even if the experts don't love it.
A C programmer will snub an Excel sheet, but it doesn't change the fact that it's genuinely useful for a wide number of use cases.
As an aside, the AI Art Turing Test was a bit eye-opening for me: https://www.astralcodexten.com/p/how-did-you-do-on-the-ai-ar...
I say all of this as someone who hopes that art remains human.
No. We'll find out that good enough is going to get beaten by the best. That's how capitalism works.
> If you marvel at the output of generative AI be it pictures, logo, music, 3D models, prose or code - you probably are yourself way too under-skilled to be gainfully employed or not yourself capable of generating similar or better output.
Interesting take. I am continuously impressed by these models. I also train these models for a living, and have worked on some of the highest profile models of the last few years.
> you probably are yourself way too under-skilled to be gainfully employed or not yourself capable of generating similar or better output.
Have you considered the posibilities that "subpar" coding skills are still extremely valuable?
I say this because almost every single extremely high paid engineer that I know at rocketship unicorn startups to FAANG companies are all using AI coding as an essential part of their work flow. Its ubiquitous. And we get paid tons of money.
We aren't just copying an pasting hundreds of lines of code and pushing to prod, of course, but its an invaluable tool for significantly speeding up an engineers coding workflow.
This contradicts your claim, as these aren't need grads, these are all highly paid professionals.
I use these models myself. Day in and out. But I'm not impressed by their capabilities or quality.
I'm only impressed by the speed.
So would be every engineer out there that you're referring to. None of them probably is incapable of doing better than an LLM.
It's... Just the speed.