> analysts have estimated that it will take $2 trillion a year in revenue to pay for the infrastructure that has already been built
I doubt any credible analyst has claimed that. What's the total AI capex that's already been spent? About $1T? It's a pretty absurd idea that those DCs need to make $2T/year for 5-7 years -> $10T-14T over their lifetime to break even.
(Yes, this is nitpicking in the sense that there are probably analysts talking about how projected and sustained capex at >1T/year will require 2T/year in revenue, so patching the article won't be a biggie. But this article is cosplaying as financial analysis and leads off with such an obviously incorrect argument. What does that say about the credibility of the rest of the article?)
> Diseconomies of Scale.
Another very basic mistake here. The author starts talking about efficiency in the context of past technologies. That's lower unit costs as scale increases.
But for AI, they seem to switch from talking about unit costs to total costs. Or at least I can't explain what they say about models getting more expensive over time in any other way, because that is not true about unit costs.
We've never seen economies of scale as large as for AI. For a given quality level, the cost has been dropping at >10x per year, not increasing.
I appreciate someone else here sees this math is way off.
With what seems to be their current revenue, the frontier companies are likely better off than Uber was pre-IPO. The "circular financing" we've seen probably lowers risk relative to external financing, since both parties get aligned incentives.
It's too bad Nvidia keeps getting hammered, they're selling the shovels in a gold rush that'll last for years, and I don't think that's priced in.
To parallel to The Big Short this is the point in the movie where folks realize it’s mathematically impossible for things to not implode and so players are quietly positioning themselves for that eventuality before things are allowed to blow.
It’s been a dramatic shift these last six months but everywhere I look now folks are quietly preparing their battle armor to survive what’s about to unfold.
The tech will stay, but the AI business landscape will have a market-cleansing forest fire.
There’s a whole generation in tech now that’s never seen what happens when a bubble like this unravels. I fully expect we’ll see the likes of offices just abandoned overnight with food still in the fridge as AI company after AI company just vaporizes.
Not financial advice, but: Honestly it’s pretty hard and I wouldn’t recommend it. Shorting stocks and getting into bonds / default swaps but the system is broadly rigged against small players doing well here. Ie in the Big Short see everything that was needed for two guys to bet with their own money.
It’s likely to be bumpy for all but stay the course with diversified strategy. The .com bust and 2008 are just blips and cheap buying opportunities for most folks with diversified portfolios and index funds. We’ll likely see similar messy markets for a while but the world will eventually recover and move on.
The ones that get truly wiped out are those with a lot of paper wealth now that implodes, but have little liquid wealth. These types go from “super rich” to can’t pay their bills almost overnight. Cash is king for times like those ahead so if you don’t have a lot of cash on hand, now is the time to secure that if you can.
Heck, it's pretty rigged against big guys too. The subprime bubble was not something that only Michael Burry knew about -- it was a front page story on the Economist several times years before the crash.
There were lots of big guys that bet against the bubble, but bet too early and weren't able to keep up with margin calls etc.
Be careful if you expect a dot com fallout - that was retail investor driven and took a long time to unwind as people sat through painful drops. Ai is much more of a private investment bubble. AI remains useful. What is likely to go away (and all at once) is investment and free rides/discounts.
So I expect more of a sobering process for AI companies rather than a blowup, simply because they all still will have cash in the bank and some have actual products with clients that make use of them.
The most straightforward thing to do that would have saved you in 2008, 2000 (and even 1929 to some extent), is to limit debt exposure and have enough cash sitting around so you don't have to sell your positions.
People who held (and invested more during the low points) did just fine. The people who got hurt the most were in a position where they had to sell (their 401k, their house etc) for a loss.
When a crash happens, it's a buyers market. In some ways it's a transfer of wealth to the top.
The problem is, timing the crash is impossible. Inflation chews into your cash portfolio, that could've been used to grow wealth.
Edit
There are of course more complex financial instruments. And they're interesting, but essentially amount to gambling. Holding short positions on sectors/companies is a zero sum game. As a retailer, you're betting against firms that have a lot of money and insider knowledge. Even if the thesis is correct, it has to be correct at the right moment.
I don't think there's a good way for individual investors to position themselves against economy-wide investment misallocation. As The Big Short also covered, any profitable short position is going to require making a risky bet about when the bubble will pop, and when things get chaotic there's no guarantee that instruments which "should" be correlated to the thesis will actually remain so. You can:
* identify a bad AI stock at $100 today
* produce an ironclad 100% guaranteed proof the company's valuation will drop below $50 by the end of 2028
* buy a bunch of long-dated puts with a break-even of $75
* see the AI investment bubble visibly falling apart in let's say March 2027
and you still lose if OpenAI acquihires the team at $80 in an attempt to prop things up.
Thats a weird take. Google/Alphabet can literaly just afford their AI investment.
Microsoft can as well.
So Anthropic only has a few thousand people which is nothing and OpenAI has also only a few thousand.
A lot of people working at these companies would find a new job easily and probably making a shit ton of money anyway.
The broader AI companies do not have that much venture capital, not that many peple and potentially not even that high of risks.
Deepseek didn't do their funding round because they could'nt even get the compute if they wanted to. So they might be very lean.
And even investors for OpenAI and Anthropic might accept that they stop R&D and just keep the services up and running as they are cost effective.
The bigger problem will be the industry itself: If i'm 5x more productive, and generate now more costs on the business, they will reduce headcounts to compensate for it.
> Most new technologies have been welcomed by the public with open arms.
I'm not going to predict how this is going to turn out in either direction but this statement gives me pause. I don't think that's ever been true. Yes, the siren song is strong but initially most new tech is met with skepticism. Are we so quick to forget "the internet/computers are just a fad"-type thinking?
Yeah backlash against railroads was very widespread, from lots of angles (farmers, property owners, canal investors, health-conscious people, religious leaders, etc). Probably the same for cars.
What's new is the inversion of who holds these opinions. With the internet/computers, company leadership was often the skeptical voice while small pockets of individuals would push these technologies from the bottom. AI is the complete opposite with skepticism coming bottom-up and push to adopt coming top-down. The last time this (almost) happened was "The Metaverse" which seemed to be getting a lot of push from the C suite.
This doesn't mean LLMs will be as useless as the metaverse, but a lot of the biggest proponents of the technology within companies are those most clueless about it which feels like a red flag.
Exactly, it's utter nonsense. The general public thought the first bicycles, automobiles, computers, mobile phones, smart watches, and AirPods were stupid. And now they are normal. Same with video games and dating apps.
I lived it. I had a teacher in high school continually scold me for reading books on programming languages and banned me from writing essays on the future of computers. Apparently, I should have been outside getting my ass kicked by the jocks like any normal kid would be. Why the guy even called up my parents to express his concerns about me. Or there was the time my 2nd grade teacher contacted my parents because I wasreading Run Silent, Run Deep during my submarine phase because that was a book for adolescents, not children. Mel Brooks wrote it best...
When ecommerce was getting big, I knew a number of people who thought it was absurd. You can't just send money over the Internet, hackers will steal it! You should use normal payment methods, like writing "fifteen dollars and 56/100------" on a special piece of paper the bank gave you.
People weren’t as ignorant back then as you think. They had very good reason to reject heliocentrism, namely that constellations don’t grow and shrink in size as the earth moves around. That, they said, would require the universe to be unimaginably big and empty, much more so than any other observation they could make at the time would suggest.
If you have the time, I recommend reading Pliny’s Natural History. I think it really shows how much more open-minded these people were than we give them credit for. They just didn’t have the technology we have.
Apparently you haven't either and you're trying to weasel out for the church because it wasn't the only reason they executed him. It amazes me that it's okay to have an invisible friend based on 2,000 to 4000-year-old writings by goat herders high on psilocybin, but if you form a parasocial attachment to an AI, that's AI psychosis. ELI5 please?
I'm not your therapist, take this question to them. And I'm not trying to weasel out of anything. You asserted that, in the past, merely claiming that the Earth revolves around the Sun was met with getting burned at the stake by a mob of priests foaming at the mouth. I corrected that, because it's a myth. Please stop sneering at centuries/millenia-old people compared to whom you live like a king, and find another group of people to feel superior to.
A modern equivalent of abandoning zeppelins because of the Hindenberg would be abandoning skyscrapers because of 9/11.
But it's humorous that Sergey Brin has a company dedicated to bringing back zeppelins and that's a lot more fun than Zuck's doomsday bunker or Bezos's doomsday yacht.
They had a ton of fundamental problems despite seeming like a good idea at first. It was only after we started using them did we realize they weren’t actually as good of an idea as they seemed.
That was ~90 years ago. We're a few branches higher on the tech tree now. Their return has been promised for over 40 years. Only the future itself will answer whether they do.
AI investment will crash but AI itself (the technology) will continue thriving, learning, improving and there is absolutely no way to stop it. The only reading on the crystal ball is if US companies fail, China will take the lead by leaps and bounds, so the only solution is to keep pushing the cart until the wheels come off or we all cross the finish line, together.
It can't 'learn' on its own. Models only get better with mountains of RnD for data, training, and lots of fine tuning.
So the moment investment dries up, models stop improving.
However, it's likely we'll get good 80/20 solutions where you get most of the performance of the then-unsustainable high end models for significantly less compute.
Unless if they take down the flat rate subs and you have to pay api prices, usage will drop a lot. Right now I freelance using gpt, if I had to pay api prices I'd probably lose 50% of the income, if not more, with the ~40% tax on top I might as well spend my time doing something else
You're unlikely to ever be priced out of tokens, at least if you'd be willing to settle for a model closer to Sonnet 4.5. That level of model certainly isn't as efficient as Fable 5, but it can crank out CRUD apps and other consulting mainstays quite well, with some supervision.
To give you an example of model in this class, the DeepSeek V4 Flash preview is a 284B A13B model, with a native quantization mixing 4 bit and 8-bit values. You can easily run it on an RTX Pro 6000 Blackwell (or 2) at a reasonable quant, especially if you offload the MoE weights to 48-64GB of system RAM. This costs US$11,800 at Microcenter right now, and it will work in any gaming box with decent cooling and a modern 1000W power supply. Over the lifetime of the card, an entire system would cost you under $4,000/year. Power is about 300W for the card (either a blower model, or a workstation model with the power cap), and another 150W or so for the rest of the server.
Or you could buy it on Open Router from dozens of different commodity vendors, starting around $0.09 per million tokens input, $0.18 per million tokens output. This is a competitive market price, so some of the providers might be losing money or reselling surplus capacity. But given the underlying hardware costs, the numbers are in the ballpark. In other words, if you're willing to settle for lower-quality tokens, you can get roughly Sonnet 4.5 for close to free, or as a modest capital expense for a successful freelancer. Halfway decent tokens are cheap, and you can generate them in-house!
Well, since you "freelance using GPT" and you expressed concern that you'd stop being able to make a profit if you had to pay API prices, I figured that swapping an off-the-shelf graphics card into a gaming rig might be a reasonable way to stay in the black. If it came to that. And the software setup on Linux is just compiling and installing llama-server, which shouldn't be enough to stop any programmer trying to make a living.
My larger point is that while frontier tokens are a near-monopoly and who knows what they really cost, many real-world workflows can be run using commodity tokens, or even served in-house by anyone who can afford to hire US or EU programmers. And if you're willing to settle for what would have been a state-of-the-art coding model in October 2025, commodity tokens are close to free.
The problem is I cannot imagine ever making a single dollar with AI. I have built about 16 SaaS apps with AI, in the time it took to build by first old fashioned human-created SaaS app. The one thing the agentic coded apps have in common with my old one is that none of them have ever won a single paying customer.
That would just create incentive to build better consumer HW for larger open weight models. And that would make me very happy. But I think the poster in another thread who stated there's a netflix subscription phenomena going on with the fixed rate pricing was onto something. My agents run 24/7 until I hit all my limits. I suspect my results are not typical. And as of yesterday I have issues with both major CEOs yet I thank them both for subsidizing my tokenmaxxing.
I moved one of our daily workflows over to Kimi K3 on Fireworks. It replaces two sales assistant positions, and was about $50/day for 14 million tokens total (in/out). They surely are not subsidizing this price as it is just inference only and other providers are even less money.
The general model for subscriptions is that power users are subsidized by subscriptions of casual users, like a gym membership or whatever.
This is a little dicier in post-agent AI, because it's easier for casual users to automate power-user consumption, but the providers have done decently in discouraging that.
it's a bit unfair to say this is the only thing that can happen. we can just as easily assume in like over a few years, US leadership identifies that AI is an arms race, and continuing down this path leads is going to lead to a zero-sum-game. and so instead of just mindlessly pushing the cart forever, we see government intervention
Just like every AI cycle that came before. At every cycle, we get a new tool. But we also get an extended "AI Winter" where investment dries up. This has been happening since the 1980s, we just keep redefining what "AI" is. At every cycle we move the line a bit; at one point voice recognition was firmly in the AI camp(and before that, science fiction). Now we have a machine that can do that in our pockets and nobody cares.
Douglas Hofstadter talked about these moving goalposts for AI in his book Gödel, Escher, Bach. I remember when a program that could win at chess was considered impossibly strong AI. Then, the game go. Etc.
Heinlein in his book The Moon Is A Harsh Mistress imagined that people in the future would find it worthwhile, even necessary, to learn an artifical language LogLan so they could talk with computers. LogLan (Logical Language) would be designed to be totally unambiguous, a necessity for speaking with a computer. He didn't imagine that we'd simply throw so much compute at the problem that we'd teach computers to understand our vague, mumbled natural languages in all their diversity.
a friend of mine did a high school project where he and a friend learnt it well enough to have a short conversation with each other. That's the only time it has come up in life until now. I hope he sees this.
That's what they'll sell you. But fundamentally current LLM implementations seem to be the wrong methodology for 'AGI' and I don't think we're remotely close to having this.
This stupid LLM is already smarter than a handful of people I know.
I don't think we have pushed LLMs to their limits already, its still progressing its still crazy good and its able to run even longer today than yesterday.
But researchers are already working on more architectures.
Investors for sure don't care about people. Replace Accenture with OpenAI and you replace 800k people immediadly.
Geoffrey Hinton said it in his Videos: a normal person costs A LOT of money to train, educate, onboard, keep etc. they forget things etc.
AGI you train once, teach once then clone and copy and just run it.
We even might already crossed the line were it is already more cost efective to train 1-5 frontier models something it doesn't know yet than teaching this to a 1000 humans.
> AI investment will crash but AI itself (the technology) will continue thriving, learning, improving and there is absolutely no way to stop it.
No, there's is a way to stop it. As obvious counterexample: go all Butlerian Jihad: death sentence for anyone caught "making a machine in the likeness of a human mind." It's doable with a few laws and treaties. I'm not saying that's the only way it could be stopped, it's just a way it could.
I'm fucking sick of TINA, especially with arguments from vigorous assertion.
However, who will pay for the very expensive training if the ROI isn't there? Absent government subsidization, then likely the frontier models will be sold off to the hyperscalers by the bankruptcy administrators to wring a few more dollars out of them as they become out of date and more and more irrelevant - assuming that inference alone is profitable at a price the market is willing to pay of course, if not then yes the lights will indeed be turned off.
When it comes to the AI crash honestly my biggest concern is what is going to happen to the job market during and the years following the crash. Not sure how things are in the rest of the world, but as someone in their early 30s working in the IT industry in Sweden I’ve never seen the market this competitive before, even for mid level and senior roles, and it worries me what the future of employment is going to look like.
Maybe those older than me have been through this kind of thing before in 2008-2009 and in the early 2000s but the state of the IT job market in the last year or two has been really concerning to me. Anecdotally I’ve also heard it’s very tough for new graduates these days.
> what the future of employment is going to look like
As an already slightly outdated senior dev in the UK rapidly approaching 50 and not wanting to touch generative AI / agentic dev… … I'm increasingly looking forward to waiting tables, being a hospital porter, or anything else away from tech!
I'm thinking that if I get out soon, I might beat most of the rush and be safely employed before many of the rest of you are forced down the same path.
[more than a little defeatist, yes, but I've be been unhappy in dev for years so for me the AI driven displacement is just the final nail rather than the main big new problem]
Over 40 myself and switched to high school teaching this year. I feel like tech in general has been degrading for a while but 2025 AI really just wiped it out imo.
It's interesting that you're looking to exit the industry as a SR dev - from my casual observations it seems like mostly it's the JR devs and younger folks that have been mostly rejecting agentic dev.
Just out of curiosity, do you have a specific objections to using the technology? I totally understand the burnout however (I'm rapidly approaching 50 myself).
get into telecom is my advice. despite it being the birthplace of monstrosities like 3gpp, there is a suprising amount of sane people. and no amount of ai crash and burn will decrease the need for communications.
I think there's a few things going on and I also believe that in many (if not most, and dare I say ALL) cases AI as an excuse for lay offs is simply a scapegoat. If AI is so great, why aren't you re-training those who are ineffective at using it, and then cutting those who still don't make the cut? If it's this amazing thing for your company, wouldn't it drive revenue up drastically?
I wish the SEC would start auditing claims about AI for job cuts, not sure if its within their ball park but there's people claiming this, when they have no receipts, and then stock goes up, if that's not market manipulation, I'm not sure what is.
This is a good example. If AI increases productivity so much, then having more employees would be beneficial, because each can produce even more output. So the firm would only be limited by the imagination of their product managers and marketing team finding new things to produce with this vast capability.
Even IF somehow you can do with less employees, in the case that you over hired after the "Great Resignation" following COVID (I think this is partial to blame for many lay offs still happening now that people aren't resigning from jobs willy nilly), where are these magical numbers that your current workforce is producing with AI? Are you just shipping more code with less testing? So many questions, only very little remarks. We need receipts.
Devil's advocate: if demand is constrained/finite in a given market, there's an upper bound on revenue, no matter how hard you try. In that case it'd make sense to work on efficiency to maximize profits given fixed revenue.
One of the strong arguments in favor of AI accelerationism is net pain reduction by making the transition period as short as possible. Ideas like “pacing” or a crash and reboot will prolong the inevitable.
That is an absurd position to take. Nobody benefits by having less time to adapt to changes in their employability
I am not calling you out personally here, but this comment bears many hallmarks of the "accelerationist disease": thinking in abstract models rather than in concrete terms; optimising for the fangs of capital rather than human beings, imaginary infinite bets that lead participants to nonsensical conclusions
If AI replaces all labor in two years, unemployment insurance and savings can comfortably carry me over the transitional period to universal income or obsolescence of money.
Should AI stagnate at the point where 70% of software engineering jobs are gone, I am going to have a problem.
Concretely, I imagine that I would need to learn other knowledge work. If AI can also perform it, I would have to go into blue collar work.
Not exactly true. If AI goes from 0 to 100 quickly (human level AGI in a few years), we will all lose our jobs together and will have to confront this fact. There is no ignoring such massive societal elephant in the room.
If instead AI improves slowly, lots of people will lose their jobs, but majority will not. Those still with jobs may not be persuaded by the sizable jobless minority to support any societal changes, since they are still OK. This means the jobless will be screwed.
Firstly, this generation of AI is not going to take away all jobs
Secondly, we already know what happens when sectors face mass unemployment from the 1980s, when steel and coal workers were laid off en-masse: nobody came to help
Accelerationists want to see their jobs go up in smoke because they think that if enough white collar workers are impoverished, everyone who was formerly poorer than them, will chip in to rescue us
Just like they didn't in 1985...
In a context where the state already has less tax revenue due to mass unemployment! It's completely bananas thinking that only survives on internet brainrot forums
Human-level AGI is by definition able to take away all human jobs.
> Secondly, we already know what happens when sectors face mass unemployment from the 1980s, when steel and coal workers were laid off en-masse: nobody came to help
There is a massive difference between 20-30% of people in a country being unemployed (historical sector-wide collapses), and 80-100% of people in a country being unemployed (human-level AGI unemployment). The second situation is completely historically unprecedented and will *not* lead to the same results.
> In a context where the state already has less tax revenue due to mass unemployment!
If human-level AGI is an actual reality, taxation (and money itself) is no longer relevant. Only energy and resources may still matter.
I am not advocating that, since I am an accelerationist, but basic reading comprehension is obviously a problem for you both here and in your other trolling efforts. (He has since edited his comment to remove the direct claim I was personally advocating terrorism).
You cannot stop the advancement of technology towards AI without the threat of violence towards those that pursue technology. The Unabomber was at least honest about that.
> edit: updated as it is not clear if Mosura is being serious or exaggerating for rhetoric
You are the HN manifestation of that woman’s face when interviewing Sydney Sweeney.
If humans are machines executing on some medium eventually it will be possible to emulate them on some other medium. Believing otherwise requires some essentially religious convictions on the existence of the soul or similar.
We already have human intelligence. Creating new humans is easy (and fun)
The only reason to emulate human intelligence is to capture it, exploit it, and reduce the negotiating power of those who have it naturally. Anyone who wants this is not a friend of human beings
Creating a new educated, competent human takes ~25 years, and significant costs and resources.
AGI can be just copy-pasted at essentially no cost. Even a robot that is mass-produced for tens of thousands of dollars is still much less expensive and takes much less time to make than a competent human.
Yes, but I think we should separate human economic value from the true purpose of our existence. If AI can replace us in economically valuable tasks, we will have no choice but to do so. The economic value part should be looked at objectively.
That does not follow. The argument about the possibility of artificial intelligence makes absolutely no claims about the purpose of human or artificial intelligence. One is a claim about "is", another is a claim about "ought".
Technological changes to economic conditions in general are not possible to opt out of forever. I'm not an AI accelerationist myself, because I think there are severe risks beyond the merely economic, but the auto industry is a good example of where many countries are engaging in what one might call "EV accelerationism". In principle you don't have to, it's possible to let them roll out only to the extent that the market naturally demands, but if you know internal combustion engines will become a niche hobbyist thing by 2050 there's a lot of value in making that transition faster and intentional.
That all persons doing intellectual work die of hunger as they cannot find a job. At least that's how I understand what the parent commenter is implying. They say that all people doing intellectual work better die quickly so they don't suffer a long 'transition period' to inevitably being dead of hunger.
We already have disruptions in Software engineering, image generation, video generation, etc.
Humans might not like AI, Companies do and with Claude we are no longer sitting on the right side, if a Product manager tells a Boss that the PM now can do also code, the boss will listen.
But we will see other effects in the next 5-15 years: population pyramid is crashing which will increase demand for people. We already have high demand for other jobs.
so i can see a transition for a lot of white collar workers to blue collar workers which will continue the disruption for everyone.
Then the robots come along and AI integration is done and we will have a real issue on our hand. When will this come? No clue i don't think its unrealistic to expeirernece this in my lifetime.
What if they're cheap, though? Perhaps the pay scale for programmers starts to resemble that for finance bros: very low pay for the first few years because so many go into the field, and those who accumulate skill (and connections, etc) can then hit the hockey stick part of the graph and get a huge pay boost when they become middle and senior. So, it would be a deskilling of programmers at the start of their careers but still worthwhile to hire them if they are cheap and can dig through the tech debt with sheer numbers/manpower. I'm just spitballing here.
I've already seen an instance of a junior developer being completely unable to fix vibe coded technical debt, even with guidance, because AI analysis is fundamental to how they understand code and they're unable to comprehend what's happening when the AI analysis is not right.
When I joined my first startup, they said if you didn't have a good plan for going public or getting acquired by series C, then the D in series D stood for death. Uber et al raised like a J round. None of the rules around markets or investments are real, it's all vibes-based, and the decision-makers love talking to their ChatGPT mistresses too much to let this all come crashing down.
Yeah. Uber's a successful company and they raised a J round. Therefore the rule (D means Death) is not real. I'm not sure what you're implying, or if you just misunderstood the original comment?
>or if you just misunderstood the original comment?
Did you? If you keep on reading, the parent comment further implies that both AI companies and uber are "vibes based". But that's hard to square with the fact that 7 years since IPO the share price is up 75%. Of course, you can argue we're still in the irrational exuberance stage, but that just creates a situation where you can never be called wrong.
I've been reading a lot of stories like this lately. I'm no business genius, but you'd assume that investors are. Are they just blind or are they burning cash on purpose. What's the steelman argument here?
> Circular Revenues: A small handful of tech firms, chip manufacturers, and AI companies are propping each other up by investing and buying from each other.
> Increasing Corporate Skepticism: The news is full of stories of corporations that are throttling the employee use of AI since the costs to use the software are a lot higher than expected.
Real AI spend is out of control, with the news is full of stories about corporations trying to keep a lid on it, but also real AI spending is low and concentrated to a few firms.
I don't know. This doesn't feel very coherent to me, but rather like a collection of assertions that are adopted because they individually say something bearish about the industry.
Certainly some investments will have been overreaches, but I find it pretty unlikely that any of the compute build-out to date is going to be left sitting idle one or two or five years from now.
I generally like this but I think the diseconomies of scale bit is narrow. Open models are showing us that it’s just the frontier companies in the west which are bloating and ignoring efficiency and that they’re leaving a lot of efficiency on the table.
Maybe it would be good if more electricity generation became available for other purposes like heat pumps and electric cars? Although, perhaps it doesn’t help as much if it’s in the wrong place.
These days we argue whether the version of SQL created by the AI Swarm at Cursor is really SQL, or whether that benchmark is invalid because SQL and its code is already in the model's training data. Does that scratch the itch for you? X-D
Most of the post makes sense to me, even if I don't agree. This part however:
> The news is full of stories of corporations that are throttling the employee use of AI since the costs to use the software are a lot higher than expected
is not something I've seen. News is not "full of stories" of this behaviour. There are a few anecdotal stories here and there.
I don't have much background in the area, but I am surprised to see that everyone here basically agrees a crash is imminent. There are disanalogies to past crashes that don't convince me that a big crash is definitively coming in the near term.
For example:
- Anthropic makes a profit right now and is seemingly on an exponential upward trajectory, so the debt being too much for it doesn't seem compelling to me.
- AI technology continues to get better exponentially and doesn't have any clear sign this trend is flattening. If anything, it's accelerating. So it's plausible the investor value is legitimate for these companies given the massive potential for continued profitability.
- I would say markets are typically very good predictors of the future. Many sophisticated investors know about the case for the future crash and are still buying at these valuations.
I am open to being wrong, but the assessment in the blog seems one-sided to me.
Polymarket currently puts the chance of such a downturn at 20% by December 2026. Seems like most people would put higher chances here, but I'm not convinced by the arguments. (https://polymarket.com/event/ai-bubble-burst-by)
I think the answer here is that we are NOT going to have the Jetson's future with humanoid robots in many homes nor self driving cars taking over the roads. I am less sure about that second one, though, since people in my family tell me their self driving Tesla's do a great job.
But those silver boxes at the side of the road? Might see many more of those, and much larger, maybe red hot on top
I think after the AI crash the biggest winner will be Apple.
They will release MacBook Pro M10 or whatever that can comfortably run Opus5 levels of performance local model for software development/general ai that is baked in the MacOS for 7k USD.
i mean my line is always "when." people have been saying ai is a bubble and will crash since the beginning. the question has always been "when."
i think the best framing of the ai crash is not in terms of software, but in terms of real estate and infrastructure. the real cost is hidden in construction contracts and people buying up land (and chips, transformers, generators, etc), not in technology. this will be the 2008 real estate crash, not the 2000 dot com bubble. (altho this point is not my most well researched point, i tbh need to look more into hard numbers of where the most risk is distributed, ai software startups or data center contracts. but i also am in grad school i have papers to publish and no time.)
im also gonna push back on the "diseconomy of scale" point. while its true that the best models rn use the most resources, we are building better smaller models as well. i recently set one up on my 12 gb vram and while its not nearly as good as claude, it works. i make it write really simple code for me. and we are still improving small models. i think theres some level of hope to economy of scale. obviously there are going to be physical limits, but im sure there will be a way to get economies of scale to work.
What implications might an AI crash have on the software developer job market? It’s likely that many AI software companies would fold and thus have to layoff their employees, but what would be the implications for the rest of the tech job market?
I often see developers giddy about the idea of AI being a bubble and waiting for the crash, but I suspect this event could actually be particularly terrible for us.
These days, I find it hard to predict the future. Things feel too complex, and the times seem different.
People say the AI bubble will burst, and I also think it will. But I want to think about how it might differ from other bubbles.
One positive factor is that during the internet startup era, there was almost no revenue. But now, big tech companies are generating profits and can absorb AI-scale losses.
The dot-com bubble burst, but the internet ended up being far more valuable than the bubble itself. I think the value was priced in early, and AI will be similar.
The core issue is always the same: there's a bubble and there are warnings, but you never know when it will collapse. And the people who act first on the warnings have to give up on the upside. The only winners are those who sell everything just before the crash. The problem is knowing when that moment is.
One difference from the dot-com bubble is that after it burst, a second wave of entrepreneurs leveraged the cheap telecom infrastructure that was left behind. So after the AI bubble bursts, will OpenAI and Anthropic fall and new companies emerge? I'm skeptical about that. Something feels different this time.
My biggest concern, though, is that if I suddenly couldn't use AI anymore, I'm worried about how long it would take to recover the coding skills I've lost after nearly a year of barely hand-coding. AI has become too deeply embedded in my life
There's a very nonzero chance that the AI expenses are not a race to recoup investments as they are a bet (with other people's money) that AI will lead to longevity or immortality, a bet made by aging tech oligarchs. Notice that Ellison is betting his entire company on it, which due to his secured debt is the same as betting his personal fortune in the most extreme case.
im sure people will continue to want to eat food, live in houses, go on vacation, look at phone, send email, drive car, even if there’s no ai money frothing around
Eh, the empire aka, protection of free trade as a service is still alive and well with no viable alternative on the horizon. Every other empire is either not free trade or - just plain mercantilism and thus will exploit way more then the quite "idealist" us- trump is just a taste of what the others in power would feel like.
I think the Hormuz saga and how China just quietly stopped importing oil are signs that we are nearing the end of the "free trade as a service" from the US. There are so many aspects of that I do not understand yet: how much oil trade will be done in yuan in the future is one that leads to other question about the current wars and what will happen when those end.
China won already with or without AI. You just did not realize it yet.
The US is in a position where they have to debase the dollar all the way to zero and the Japan carry trade will slowly unwind for the worst as soon as the Bank of Japan (BOJ) begins to increase rates in more than 30 years.
Someone has to pay for all the trillions of dollars of debt for the data center build out that is being hidden from the balance sheets of the big tech companies.
I don't think the country that singlehandedly messed up the global economy by launching a pointless war on Iran - without even coordinating with its allies! - that predictably got the Strait of Hormuz closed gets to call itself protector of free trade anymore.
The US has lost its appeal as a hegemon. Constantly threatening allies with invasion isn't helping either.
Open a history book- before the us century- this stuff was the norm. Its what is to be expected after the us- retreats inwards. China blockading japan. The world blockading china. Russia blockading eastern Europe and vice versa. The shitty old world pre the ww2 order.
It is pro-kicking out oppossition to the system that uses the system to grow in power without contributing anything back to it. So the us has abandoned the "Wandel durch Handel" ideology- which lets admit it- was a brutal failure. Russia did not change. China changed back. None of the countries in the middle east got democratic or "contributing to the worlds allmende". Turkey is sliding back.
Those analyses tend to not consider the government interest in having AI as a weapon: now AI labs are fully part of the military sector, making them much more supported by the system
This has been my take too. I think that the revenue requirements for making the industry profitable are likely never to be met, but that the US government sees AI as a strategic necessity and will ensure that the companies that are pushing the technology forward will be paid huge sums of money if they ever become insolvent. It's not ideal, but if banks are too big to fail, I'm guessing that OpenAI and Anthropic are too. This is doubly true since the defense sector is already integrating these solutions into their systems and actively using them in wartime activities.
This is a forbidden subject here. In my opinion, the foremost subject in the entire world, but for whatever reason seldom discussed and always dismissed.
All four frontier models contracted with the Pentagon. Claude already having been involved in Operation Abduct Sitting President & Use Active Denial Systems to Make People Vomit and Shit Themselves 'because'. Institutional and corporate capture . Foreign nations now having our models trained to facilitate tailored political orientation. And the fact that these systems, for all their limitations perennially bewailed, have tremendous capacity as force multipliers for all the things civil rights activists were previously terrified about apparently just before all our problems were magically solved.
Snowden docs x³
Things that once took fusion centers, thinktanks, research, etc and thousands of manual processing hours -- now real-time, instant, with real-time 'understanding'.
I think the AI crash will be primarily comprised of collectively lamenting all the obvious things we overlooked.
> analysts have estimated that it will take $2 trillion a year in revenue to pay for the infrastructure that has already been built
I doubt any credible analyst has claimed that. What's the total AI capex that's already been spent? About $1T? It's a pretty absurd idea that those DCs need to make $2T/year for 5-7 years -> $10T-14T over their lifetime to break even.
(Yes, this is nitpicking in the sense that there are probably analysts talking about how projected and sustained capex at >1T/year will require 2T/year in revenue, so patching the article won't be a biggie. But this article is cosplaying as financial analysis and leads off with such an obviously incorrect argument. What does that say about the credibility of the rest of the article?)
> Diseconomies of Scale.
Another very basic mistake here. The author starts talking about efficiency in the context of past technologies. That's lower unit costs as scale increases.
But for AI, they seem to switch from talking about unit costs to total costs. Or at least I can't explain what they say about models getting more expensive over time in any other way, because that is not true about unit costs.
We've never seen economies of scale as large as for AI. For a given quality level, the cost has been dropping at >10x per year, not increasing.
I appreciate someone else here sees this math is way off.
With what seems to be their current revenue, the frontier companies are likely better off than Uber was pre-IPO. The "circular financing" we've seen probably lowers risk relative to external financing, since both parties get aligned incentives.
It's too bad Nvidia keeps getting hammered, they're selling the shovels in a gold rush that'll last for years, and I don't think that's priced in.
It’s not a question of if but when at this point.
To parallel to The Big Short this is the point in the movie where folks realize it’s mathematically impossible for things to not implode and so players are quietly positioning themselves for that eventuality before things are allowed to blow.
It’s been a dramatic shift these last six months but everywhere I look now folks are quietly preparing their battle armor to survive what’s about to unfold.
The tech will stay, but the AI business landscape will have a market-cleansing forest fire.
There’s a whole generation in tech now that’s never seen what happens when a bubble like this unravels. I fully expect we’ll see the likes of offices just abandoned overnight with food still in the fridge as AI company after AI company just vaporizes.
How would one position oneself as an individual investor if one believed this thesis?
Not financial advice, but: Honestly it’s pretty hard and I wouldn’t recommend it. Shorting stocks and getting into bonds / default swaps but the system is broadly rigged against small players doing well here. Ie in the Big Short see everything that was needed for two guys to bet with their own money.
It’s likely to be bumpy for all but stay the course with diversified strategy. The .com bust and 2008 are just blips and cheap buying opportunities for most folks with diversified portfolios and index funds. We’ll likely see similar messy markets for a while but the world will eventually recover and move on.
The ones that get truly wiped out are those with a lot of paper wealth now that implodes, but have little liquid wealth. These types go from “super rich” to can’t pay their bills almost overnight. Cash is king for times like those ahead so if you don’t have a lot of cash on hand, now is the time to secure that if you can.
Heck, it's pretty rigged against big guys too. The subprime bubble was not something that only Michael Burry knew about -- it was a front page story on the Economist several times years before the crash.
There were lots of big guys that bet against the bubble, but bet too early and weren't able to keep up with margin calls etc.
Be careful if you expect a dot com fallout - that was retail investor driven and took a long time to unwind as people sat through painful drops. Ai is much more of a private investment bubble. AI remains useful. What is likely to go away (and all at once) is investment and free rides/discounts.
So I expect more of a sobering process for AI companies rather than a blowup, simply because they all still will have cash in the bank and some have actual products with clients that make use of them.
Not financial advice.
The most straightforward thing to do that would have saved you in 2008, 2000 (and even 1929 to some extent), is to limit debt exposure and have enough cash sitting around so you don't have to sell your positions.
People who held (and invested more during the low points) did just fine. The people who got hurt the most were in a position where they had to sell (their 401k, their house etc) for a loss.
When a crash happens, it's a buyers market. In some ways it's a transfer of wealth to the top.
The problem is, timing the crash is impossible. Inflation chews into your cash portfolio, that could've been used to grow wealth.
Edit
There are of course more complex financial instruments. And they're interesting, but essentially amount to gambling. Holding short positions on sectors/companies is a zero sum game. As a retailer, you're betting against firms that have a lot of money and insider knowledge. Even if the thesis is correct, it has to be correct at the right moment.
I don't think there's a good way for individual investors to position themselves against economy-wide investment misallocation. As The Big Short also covered, any profitable short position is going to require making a risky bet about when the bubble will pop, and when things get chaotic there's no guarantee that instruments which "should" be correlated to the thesis will actually remain so. You can:
* identify a bad AI stock at $100 today
* produce an ironclad 100% guaranteed proof the company's valuation will drop below $50 by the end of 2028
* buy a bunch of long-dated puts with a break-even of $75
* see the AI investment bubble visibly falling apart in let's say March 2027
and you still lose if OpenAI acquihires the team at $80 in an attempt to prop things up.
Thats a weird take. Google/Alphabet can literaly just afford their AI investment.
Microsoft can as well.
So Anthropic only has a few thousand people which is nothing and OpenAI has also only a few thousand.
A lot of people working at these companies would find a new job easily and probably making a shit ton of money anyway.
The broader AI companies do not have that much venture capital, not that many peple and potentially not even that high of risks.
Deepseek didn't do their funding round because they could'nt even get the compute if they wanted to. So they might be very lean.
And even investors for OpenAI and Anthropic might accept that they stop R&D and just keep the services up and running as they are cost effective.
The bigger problem will be the industry itself: If i'm 5x more productive, and generate now more costs on the business, they will reduce headcounts to compensate for it.
> Most new technologies have been welcomed by the public with open arms.
I'm not going to predict how this is going to turn out in either direction but this statement gives me pause. I don't think that's ever been true. Yes, the siren song is strong but initially most new tech is met with skepticism. Are we so quick to forget "the internet/computers are just a fad"-type thinking?
Yeah backlash against railroads was very widespread, from lots of angles (farmers, property owners, canal investors, health-conscious people, religious leaders, etc). Probably the same for cars.
What's new is the inversion of who holds these opinions. With the internet/computers, company leadership was often the skeptical voice while small pockets of individuals would push these technologies from the bottom. AI is the complete opposite with skepticism coming bottom-up and push to adopt coming top-down. The last time this (almost) happened was "The Metaverse" which seemed to be getting a lot of push from the C suite.
This doesn't mean LLMs will be as useless as the metaverse, but a lot of the biggest proponents of the technology within companies are those most clueless about it which feels like a red flag.
Exactly, it's utter nonsense. The general public thought the first bicycles, automobiles, computers, mobile phones, smart watches, and AirPods were stupid. And now they are normal. Same with video games and dating apps.
I lived it. I had a teacher in high school continually scold me for reading books on programming languages and banned me from writing essays on the future of computers. Apparently, I should have been outside getting my ass kicked by the jocks like any normal kid would be. Why the guy even called up my parents to express his concerns about me. Or there was the time my 2nd grade teacher contacted my parents because I wasreading Run Silent, Run Deep during my submarine phase because that was a book for adolescents, not children. Mel Brooks wrote it best...
https://www.youtube.com/watch?v=hYTQ7__NNDI
When ecommerce was getting big, I knew a number of people who thought it was absurd. You can't just send money over the Internet, hackers will steal it! You should use normal payment methods, like writing "fifteen dollars and 56/100------" on a special piece of paper the bank gave you.
Never thought I'd see the ycombinator peeps deny the existence of the Luddite mindset, but here we are.
It's more a post GamerGate understanding of a nerds vs the world narrative framing
The death of Harambe and Cecil the lion ruined the whole world and directly lead to GamerGate/Trump and world wise ride in right wing populaism.
Heliocentrism calling...
Yeah, we reject anything new and scary. Gotta wonder what the cavepeople made of the wheel.
Edit: still triggered by heliocentrism in 2026? OK then.
People weren’t as ignorant back then as you think. They had very good reason to reject heliocentrism, namely that constellations don’t grow and shrink in size as the earth moves around. That, they said, would require the universe to be unimaginably big and empty, much more so than any other observation they could make at the time would suggest.
If you have the time, I recommend reading Pliny’s Natural History. I think it really shows how much more open-minded these people were than we give them credit for. They just didn’t have the technology we have.
Tell that to Giordano Bruno.
You haven’t even bothered to look up what actually got him executed, have you.
Apparently you haven't either and you're trying to weasel out for the church because it wasn't the only reason they executed him. It amazes me that it's okay to have an invisible friend based on 2,000 to 4000-year-old writings by goat herders high on psilocybin, but if you form a parasocial attachment to an AI, that's AI psychosis. ELI5 please?
I'm not your therapist, take this question to them. And I'm not trying to weasel out of anything. You asserted that, in the past, merely claiming that the Earth revolves around the Sun was met with getting burned at the stake by a mob of priests foaming at the mouth. I corrected that, because it's a myth. Please stop sneering at centuries/millenia-old people compared to whom you live like a king, and find another group of people to feel superior to.
Another reason many people rejected heliocentrism was that you could get excommunicated (or worse!) if you professed to accept it.
I mean Zeppelins were supposed to be the premier form of air travel despite the glaringly obvious issues.
Invoking “heliocentrism” is the HN version of “Ive already drawn you as the virgin and me as the Chad.”
A modern equivalent of abandoning zeppelins because of the Hindenberg would be abandoning skyscrapers because of 9/11.
But it's humorous that Sergey Brin has a company dedicated to bringing back zeppelins and that's a lot more fun than Zuck's doomsday bunker or Bezos's doomsday yacht.
Who said I was referencing the Hindenburg?
They had a ton of fundamental problems despite seeming like a good idea at first. It was only after we started using them did we realize they weren’t actually as good of an idea as they seemed.
That was ~90 years ago. We're a few branches higher on the tech tree now. Their return has been promised for over 40 years. Only the future itself will answer whether they do.
AI investment will crash but AI itself (the technology) will continue thriving, learning, improving and there is absolutely no way to stop it. The only reading on the crystal ball is if US companies fail, China will take the lead by leaps and bounds, so the only solution is to keep pushing the cart until the wheels come off or we all cross the finish line, together.
It can't 'learn' on its own. Models only get better with mountains of RnD for data, training, and lots of fine tuning.
So the moment investment dries up, models stop improving.
However, it's likely we'll get good 80/20 solutions where you get most of the performance of the then-unsustainable high end models for significantly less compute.
> absolutely no way to stop it.
Unless if they take down the flat rate subs and you have to pay api prices, usage will drop a lot. Right now I freelance using gpt, if I had to pay api prices I'd probably lose 50% of the income, if not more, with the ~40% tax on top I might as well spend my time doing something else
You're unlikely to ever be priced out of tokens, at least if you'd be willing to settle for a model closer to Sonnet 4.5. That level of model certainly isn't as efficient as Fable 5, but it can crank out CRUD apps and other consulting mainstays quite well, with some supervision.
To give you an example of model in this class, the DeepSeek V4 Flash preview is a 284B A13B model, with a native quantization mixing 4 bit and 8-bit values. You can easily run it on an RTX Pro 6000 Blackwell (or 2) at a reasonable quant, especially if you offload the MoE weights to 48-64GB of system RAM. This costs US$11,800 at Microcenter right now, and it will work in any gaming box with decent cooling and a modern 1000W power supply. Over the lifetime of the card, an entire system would cost you under $4,000/year. Power is about 300W for the card (either a blower model, or a workstation model with the power cap), and another 150W or so for the rest of the server.
Or you could buy it on Open Router from dozens of different commodity vendors, starting around $0.09 per million tokens input, $0.18 per million tokens output. This is a competitive market price, so some of the providers might be losing money or reselling surplus capacity. But given the underlying hardware costs, the numbers are in the ballpark. In other words, if you're willing to settle for lower-quality tokens, you can get roughly Sonnet 4.5 for close to free, or as a modest capital expense for a successful freelancer. Halfway decent tokens are cheap, and you can generate them in-house!
> You can easily run it on an RTX Pro 6000
Sure, and people can just build their own dropbox for like $250 too.
How many people will bother though
Well, since you "freelance using GPT" and you expressed concern that you'd stop being able to make a profit if you had to pay API prices, I figured that swapping an off-the-shelf graphics card into a gaming rig might be a reasonable way to stay in the black. If it came to that. And the software setup on Linux is just compiling and installing llama-server, which shouldn't be enough to stop any programmer trying to make a living.
Or you could just take your credit card and spend $20 on credits at https://openrouter.ai/deepseek/deepseek-v4-flash. I'm not sure that I could manage to spend even a $1/day at those rates.
My larger point is that while frontier tokens are a near-monopoly and who knows what they really cost, many real-world workflows can be run using commodity tokens, or even served in-house by anyone who can afford to hire US or EU programmers. And if you're willing to settle for what would have been a state-of-the-art coding model in October 2025, commodity tokens are close to free.
The problem is I cannot imagine ever making a single dollar with AI. I have built about 16 SaaS apps with AI, in the time it took to build by first old fashioned human-created SaaS app. The one thing the agentic coded apps have in common with my old one is that none of them have ever won a single paying customer.
That would just create incentive to build better consumer HW for larger open weight models. And that would make me very happy. But I think the poster in another thread who stated there's a netflix subscription phenomena going on with the fixed rate pricing was onto something. My agents run 24/7 until I hit all my limits. I suspect my results are not typical. And as of yesterday I have issues with both major CEOs yet I thank them both for subsidizing my tokenmaxxing.
I don't necessarily buy that API prices represent "real" prices of the models.
Yes, they are ludicrously profitable - even when future training costs are taken into account!
I moved one of our daily workflows over to Kimi K3 on Fireworks. It replaces two sales assistant positions, and was about $50/day for 14 million tokens total (in/out). They surely are not subsidizing this price as it is just inference only and other providers are even less money.
It's probably a safe assumption that openrouter prices for Kimi K3 are "real".
If they're hosted in China next to a coal power plant and maintained by people getting paid 1/10th of a US engineer sure.
I doubt opanai and anthropic will go that road
Not in the slightest as there are providers selling K3 for half of what OR has them listed for. (And maintaining profitability)
Well of course they're higher than marginal cost. But these providers have to also generate a fat return on investment.
What do you think the real price is? Subscriptions are heavily subsidised, I don't know anyone who would deny that.
The general model for subscriptions is that power users are subsidized by subscriptions of casual users, like a gym membership or whatever.
This is a little dicier in post-agent AI, because it's easier for casual users to automate power-user consumption, but the providers have done decently in discouraging that.
it's a bit unfair to say this is the only thing that can happen. we can just as easily assume in like over a few years, US leadership identifies that AI is an arms race, and continuing down this path leads is going to lead to a zero-sum-game. and so instead of just mindlessly pushing the cart forever, we see government intervention
Just like every AI cycle that came before. At every cycle, we get a new tool. But we also get an extended "AI Winter" where investment dries up. This has been happening since the 1980s, we just keep redefining what "AI" is. At every cycle we move the line a bit; at one point voice recognition was firmly in the AI camp(and before that, science fiction). Now we have a machine that can do that in our pockets and nobody cares.
Does anyone remember expert systems?
Douglas Hofstadter talked about these moving goalposts for AI in his book Gödel, Escher, Bach. I remember when a program that could win at chess was considered impossibly strong AI. Then, the game go. Etc.
Heinlein in his book The Moon Is A Harsh Mistress imagined that people in the future would find it worthwhile, even necessary, to learn an artifical language LogLan so they could talk with computers. LogLan (Logical Language) would be designed to be totally unambiguous, a necessity for speaking with a computer. He didn't imagine that we'd simply throw so much compute at the problem that we'd teach computers to understand our vague, mumbled natural languages in all their diversity.
Ah, Lojban!
https://en.wikipedia.org/wiki/Lojban
Sadly this does not let you speak to computers, just other language nerds with a lot of time.
a friend of mine did a high school project where he and a friend learnt it well enough to have a short conversation with each other. That's the only time it has come up in life until now. I hope he sees this.
Yup. If there is a crash, those with cash and understanding will clean up. AI is eating the world, period.
There is "absolutely no way to stop" the thing that requires hundreds of billions of dollars of cash injected every year just to avoid falling over?
what exactly is this finish line? AGI?
That's what they'll sell you. But fundamentally current LLM implementations seem to be the wrong methodology for 'AGI' and I don't think we're remotely close to having this.
Good luck convincing HN of this point. Half the replies in this thread reek of insanity.
Don't sneer at imaginary HN commenters. It's always possible to a point more persuasively than that.
This stupid LLM is already smarter than a handful of people I know.
I don't think we have pushed LLMs to their limits already, its still progressing its still crazy good and its able to run even longer today than yesterday.
But researchers are already working on more architectures.
Yes.
Investors for sure don't care about people. Replace Accenture with OpenAI and you replace 800k people immediadly.
Geoffrey Hinton said it in his Videos: a normal person costs A LOT of money to train, educate, onboard, keep etc. they forget things etc.
AGI you train once, teach once then clone and copy and just run it.
We even might already crossed the line were it is already more cost efective to train 1-5 frontier models something it doesn't know yet than teaching this to a 1000 humans.
Economists have a term of art for this phenomenon: "automation".
What exactly was the finish line for the Internet?
> AI investment will crash but AI itself (the technology) will continue thriving, learning, improving and there is absolutely no way to stop it.
No, there's is a way to stop it. As obvious counterexample: go all Butlerian Jihad: death sentence for anyone caught "making a machine in the likeness of a human mind." It's doable with a few laws and treaties. I'm not saying that's the only way it could be stopped, it's just a way it could.
I'm fucking sick of TINA, especially with arguments from vigorous assertion.
I didn't know there is a finish line.
However, who will pay for the very expensive training if the ROI isn't there? Absent government subsidization, then likely the frontier models will be sold off to the hyperscalers by the bankruptcy administrators to wring a few more dollars out of them as they become out of date and more and more irrelevant - assuming that inference alone is profitable at a price the market is willing to pay of course, if not then yes the lights will indeed be turned off.
When it comes to the AI crash honestly my biggest concern is what is going to happen to the job market during and the years following the crash. Not sure how things are in the rest of the world, but as someone in their early 30s working in the IT industry in Sweden I’ve never seen the market this competitive before, even for mid level and senior roles, and it worries me what the future of employment is going to look like.
Maybe those older than me have been through this kind of thing before in 2008-2009 and in the early 2000s but the state of the IT job market in the last year or two has been really concerning to me. Anecdotally I’ve also heard it’s very tough for new graduates these days.
> what the future of employment is going to look like
As an already slightly outdated senior dev in the UK rapidly approaching 50 and not wanting to touch generative AI / agentic dev… … I'm increasingly looking forward to waiting tables, being a hospital porter, or anything else away from tech!
I'm thinking that if I get out soon, I might beat most of the rush and be safely employed before many of the rest of you are forced down the same path.
[more than a little defeatist, yes, but I've be been unhappy in dev for years so for me the AI driven displacement is just the final nail rather than the main big new problem]
Over 40 myself and switched to high school teaching this year. I feel like tech in general has been degrading for a while but 2025 AI really just wiped it out imo.
It's interesting that you're looking to exit the industry as a SR dev - from my casual observations it seems like mostly it's the JR devs and younger folks that have been mostly rejecting agentic dev.
Just out of curiosity, do you have a specific objections to using the technology? I totally understand the burnout however (I'm rapidly approaching 50 myself).
get into telecom is my advice. despite it being the birthplace of monstrosities like 3gpp, there is a suprising amount of sane people. and no amount of ai crash and burn will decrease the need for communications.
I think there's a few things going on and I also believe that in many (if not most, and dare I say ALL) cases AI as an excuse for lay offs is simply a scapegoat. If AI is so great, why aren't you re-training those who are ineffective at using it, and then cutting those who still don't make the cut? If it's this amazing thing for your company, wouldn't it drive revenue up drastically?
I wish the SEC would start auditing claims about AI for job cuts, not sure if its within their ball park but there's people claiming this, when they have no receipts, and then stock goes up, if that's not market manipulation, I'm not sure what is.
This is a good example. If AI increases productivity so much, then having more employees would be beneficial, because each can produce even more output. So the firm would only be limited by the imagination of their product managers and marketing team finding new things to produce with this vast capability.
And yet...
Even IF somehow you can do with less employees, in the case that you over hired after the "Great Resignation" following COVID (I think this is partial to blame for many lay offs still happening now that people aren't resigning from jobs willy nilly), where are these magical numbers that your current workforce is producing with AI? Are you just shipping more code with less testing? So many questions, only very little remarks. We need receipts.
Devil's advocate: if demand is constrained/finite in a given market, there's an upper bound on revenue, no matter how hard you try. In that case it'd make sense to work on efficiency to maximize profits given fixed revenue.
(Not saying AI helps with that yet.)
One of the strong arguments in favor of AI accelerationism is net pain reduction by making the transition period as short as possible. Ideas like “pacing” or a crash and reboot will prolong the inevitable.
That is an absurd position to take. Nobody benefits by having less time to adapt to changes in their employability
I am not calling you out personally here, but this comment bears many hallmarks of the "accelerationist disease": thinking in abstract models rather than in concrete terms; optimising for the fangs of capital rather than human beings, imaginary infinite bets that lead participants to nonsensical conclusions
Concretely:
If AI replaces all labor in two years, unemployment insurance and savings can comfortably carry me over the transitional period to universal income or obsolescence of money.
Should AI stagnate at the point where 70% of software engineering jobs are gone, I am going to have a problem.
Concretely, I imagine that I would need to learn other knowledge work. If AI can also perform it, I would have to go into blue collar work.
> Nobody benefits by having less time to adapt
Not exactly true. If AI goes from 0 to 100 quickly (human level AGI in a few years), we will all lose our jobs together and will have to confront this fact. There is no ignoring such massive societal elephant in the room.
If instead AI improves slowly, lots of people will lose their jobs, but majority will not. Those still with jobs may not be persuaded by the sizable jobless minority to support any societal changes, since they are still OK. This means the jobless will be screwed.
Firstly, this generation of AI is not going to take away all jobs
Secondly, we already know what happens when sectors face mass unemployment from the 1980s, when steel and coal workers were laid off en-masse: nobody came to help
Accelerationists want to see their jobs go up in smoke because they think that if enough white collar workers are impoverished, everyone who was formerly poorer than them, will chip in to rescue us
Just like they didn't in 1985...
In a context where the state already has less tax revenue due to mass unemployment! It's completely bananas thinking that only survives on internet brainrot forums
> Firstly, AI is not going to take away all jobs
Human-level AGI is by definition able to take away all human jobs.
> Secondly, we already know what happens when sectors face mass unemployment from the 1980s, when steel and coal workers were laid off en-masse: nobody came to help
There is a massive difference between 20-30% of people in a country being unemployed (historical sector-wide collapses), and 80-100% of people in a country being unemployed (human-level AGI unemployment). The second situation is completely historically unprecedented and will *not* lead to the same results.
> In a context where the state already has less tax revenue due to mass unemployment!
If human-level AGI is an actual reality, taxation (and money itself) is no longer relevant. Only energy and resources may still matter.
Once AI is achieved it will not be a question of “adaptation” but “acceptance”.
You are in denial. If you don’t like this the answer was what the Unabomber proposed the whole time.
> the answer was what the Unabomber proposed the whole time.
The Unabomber (Theodore Kaczynski) killed 3 people and injured 23 others across 16 bombings between 1978 and 1995
It's absolutely incredible to see this talked about as a "reasonable" position on AI taking jobs.
edit: updated as it is not clear if Mosura is being serious or exaggerating for rhetoric
I am not advocating that, since I am an accelerationist, but basic reading comprehension is obviously a problem for you both here and in your other trolling efforts. (He has since edited his comment to remove the direct claim I was personally advocating terrorism).
You cannot stop the advancement of technology towards AI without the threat of violence towards those that pursue technology. The Unabomber was at least honest about that.
> edit: updated as it is not clear if Mosura is being serious or exaggerating for rhetoric
You are the HN manifestation of that woman’s face when interviewing Sydney Sweeney.
What is inevitable?
The pacing and the prolonging, obviously!
Artificial intelligence.
If humans are machines executing on some medium eventually it will be possible to emulate them on some other medium. Believing otherwise requires some essentially religious convictions on the existence of the soul or similar.
We already have human intelligence. Creating new humans is easy (and fun)
The only reason to emulate human intelligence is to capture it, exploit it, and reduce the negotiating power of those who have it naturally. Anyone who wants this is not a friend of human beings
> Creating new humans is easy (and fun)
This could not be further from the truth.
Creating a new educated, competent human takes ~25 years, and significant costs and resources.
AGI can be just copy-pasted at essentially no cost. Even a robot that is mass-produced for tens of thousands of dollars is still much less expensive and takes much less time to make than a competent human.
But it is more than just fun , it gives purpose to our existence … pardon me but it is idiotic to look at bringing up a child so objectively.
Yes, but I think we should separate human economic value from the true purpose of our existence. If AI can replace us in economically valuable tasks, we will have no choice but to do so. The economic value part should be looked at objectively.
I do mostly agree with you, but...
> Creating new humans is easy
If you did a survey I suspect agreement with this statement would be rather bimodal.
That's fair. It starts easy, then gets difficult.
Sure, but that only stands if you believe the only purpose of human intelligence is for economic benefit.
That does not follow. The argument about the possibility of artificial intelligence makes absolutely no claims about the purpose of human or artificial intelligence. One is a claim about "is", another is a claim about "ought".
>Believing otherwise requires some essentially religious convictions on the existence of the soul or similar.
My doubt has nothing to do with religion.
I believe we'll never be able to harness the amount of technology required to replicate human intelligence.
Technological changes to economic conditions in general are not possible to opt out of forever. I'm not an AI accelerationist myself, because I think there are severe risks beyond the merely economic, but the auto industry is a good example of where many countries are engaging in what one might call "EV accelerationism". In principle you don't have to, it's possible to let them roll out only to the extent that the market naturally demands, but if you know internal combustion engines will become a niche hobbyist thing by 2050 there's a lot of value in making that transition faster and intentional.
That all persons doing intellectual work die of hunger as they cannot find a job. At least that's how I understand what the parent commenter is implying. They say that all people doing intellectual work better die quickly so they don't suffer a long 'transition period' to inevitably being dead of hunger.
Wonderful. And these people call themselves utilitarians too. God / Buddha / Allah / Batman, save me from their kindness!
Yeah I can't imagine this suddenly stops.
We already have disruptions in Software engineering, image generation, video generation, etc.
Humans might not like AI, Companies do and with Claude we are no longer sitting on the right side, if a Product manager tells a Boss that the PM now can do also code, the boss will listen.
But we will see other effects in the next 5-15 years: population pyramid is crashing which will increase demand for people. We already have high demand for other jobs.
so i can see a transition for a lot of white collar workers to blue collar workers which will continue the disruption for everyone.
Then the robots come along and AI integration is done and we will have a real issue on our hand. When will this come? No clue i don't think its unrealistic to expeirernece this in my lifetime.
> what is going to happen to the job market during and the years following the crash
Junior programmers will be in demand again. Someone will have to fix the mountains of vibe coded technical debt.
I'm not sure that you're going to fix that with juniors...
well, somebody has to learn it anew, after all the seniors are out. It's fine.
What if they're cheap, though? Perhaps the pay scale for programmers starts to resemble that for finance bros: very low pay for the first few years because so many go into the field, and those who accumulate skill (and connections, etc) can then hit the hockey stick part of the graph and get a huge pay boost when they become middle and senior. So, it would be a deskilling of programmers at the start of their careers but still worthwhile to hire them if they are cheap and can dig through the tech debt with sheer numbers/manpower. I'm just spitballing here.
I've already seen an instance of a junior developer being completely unable to fix vibe coded technical debt, even with guidance, because AI analysis is fundamental to how they understand code and they're unable to comprehend what's happening when the AI analysis is not right.
When I joined my first startup, they said if you didn't have a good plan for going public or getting acquired by series C, then the D in series D stood for death. Uber et al raised like a J round. None of the rules around markets or investments are real, it's all vibes-based, and the decision-makers love talking to their ChatGPT mistresses too much to let this all come crashing down.
>Uber et al raised like a J round. None of the rules around markets or investments are real
Uber's a poor example because it did IPO in 2019 and its stocks are up around 75% since then.
Yeah. Uber's a successful company and they raised a J round. Therefore the rule (D means Death) is not real. I'm not sure what you're implying, or if you just misunderstood the original comment?
>or if you just misunderstood the original comment?
Did you? If you keep on reading, the parent comment further implies that both AI companies and uber are "vibes based". But that's hard to square with the fact that 7 years since IPO the share price is up 75%. Of course, you can argue we're still in the irrational exuberance stage, but that just creates a situation where you can never be called wrong.
I've been reading a lot of stories like this lately. I'm no business genius, but you'd assume that investors are. Are they just blind or are they burning cash on purpose. What's the steelman argument here?
> Circular Revenues: A small handful of tech firms, chip manufacturers, and AI companies are propping each other up by investing and buying from each other.
> Increasing Corporate Skepticism: The news is full of stories of corporations that are throttling the employee use of AI since the costs to use the software are a lot higher than expected.
Real AI spend is out of control, with the news is full of stories about corporations trying to keep a lid on it, but also real AI spending is low and concentrated to a few firms.
I don't know. This doesn't feel very coherent to me, but rather like a collection of assertions that are adopted because they individually say something bearish about the industry.
Certainly some investments will have been overreaches, but I find it pretty unlikely that any of the compute build-out to date is going to be left sitting idle one or two or five years from now.
“Infrastructure spend by providers is high and increasing while customers are actively trying to spend less”
It’s not the same people doing both.
I generally like this but I think the diseconomies of scale bit is narrow. Open models are showing us that it’s just the frontier companies in the west which are bloating and ignoring efficiency and that they’re leaving a lot of efficiency on the table.
Maybe it would be good if more electricity generation became available for other purposes like heat pumps and electric cars? Although, perhaps it doesn’t help as much if it’s in the wrong place.
Also, cheaper RAM would be nice.
Oh no, no one saw it coming. Oh well, anyway.
Let’s argue about SQL vs NoSQL again. I miss that.
These days we argue whether the version of SQL created by the AI Swarm at Cursor is really SQL, or whether that benchmark is invalid because SQL and its code is already in the model's training data. Does that scratch the itch for you? X-D
https://startupfortune.com/cursors-ai-agents-rebuilt-sqlite-...
Tabs vs. Spaces!!!
Most of the post makes sense to me, even if I don't agree. This part however:
> The news is full of stories of corporations that are throttling the employee use of AI since the costs to use the software are a lot higher than expected
is not something I've seen. News is not "full of stories" of this behaviour. There are a few anecdotal stories here and there.
It won't be the same as 2000 because back then I didn't spend a dime on those ridiculous startups that went bust.
I don't have much background in the area, but I am surprised to see that everyone here basically agrees a crash is imminent. There are disanalogies to past crashes that don't convince me that a big crash is definitively coming in the near term.
For example:
- Anthropic makes a profit right now and is seemingly on an exponential upward trajectory, so the debt being too much for it doesn't seem compelling to me.
- AI technology continues to get better exponentially and doesn't have any clear sign this trend is flattening. If anything, it's accelerating. So it's plausible the investor value is legitimate for these companies given the massive potential for continued profitability.
- I would say markets are typically very good predictors of the future. Many sophisticated investors know about the case for the future crash and are still buying at these valuations.
I am open to being wrong, but the assessment in the blog seems one-sided to me.
Polymarket currently puts the chance of such a downturn at 20% by December 2026. Seems like most people would put higher chances here, but I'm not convinced by the arguments. (https://polymarket.com/event/ai-bubble-burst-by)
Wars spending will make up for the loss
a contrarian take that's been eating away at my mind lately:
- if every autonomous self-driving car is going to need some sort of edge AI data center nearby (for instantaneous exceptional incident handling)
- if every household robot is going to need some sort of edge AI data center nearby (for instantaneous exceptional incident handling)
- if there are going to be millions of self-driving cars and autonomous robots joining us in the next decade
...then we will need lots more semi-conductor chips, and data centers in lots more places, in the next decade.
I think the answer here is that we are NOT going to have the Jetson's future with humanoid robots in many homes nor self driving cars taking over the roads. I am less sure about that second one, though, since people in my family tell me their self driving Tesla's do a great job.
But those silver boxes at the side of the road? Might see many more of those, and much larger, maybe red hot on top
I think after the AI crash the biggest winner will be Apple.
They will release MacBook Pro M10 or whatever that can comfortably run Opus5 levels of performance local model for software development/general ai that is baked in the MacOS for 7k USD.
i mean my line is always "when." people have been saying ai is a bubble and will crash since the beginning. the question has always been "when." i think the best framing of the ai crash is not in terms of software, but in terms of real estate and infrastructure. the real cost is hidden in construction contracts and people buying up land (and chips, transformers, generators, etc), not in technology. this will be the 2008 real estate crash, not the 2000 dot com bubble. (altho this point is not my most well researched point, i tbh need to look more into hard numbers of where the most risk is distributed, ai software startups or data center contracts. but i also am in grad school i have papers to publish and no time.) im also gonna push back on the "diseconomy of scale" point. while its true that the best models rn use the most resources, we are building better smaller models as well. i recently set one up on my 12 gb vram and while its not nearly as good as claude, it works. i make it write really simple code for me. and we are still improving small models. i think theres some level of hope to economy of scale. obviously there are going to be physical limits, but im sure there will be a way to get economies of scale to work.
What implications might an AI crash have on the software developer job market? It’s likely that many AI software companies would fold and thus have to layoff their employees, but what would be the implications for the rest of the tech job market?
I often see developers giddy about the idea of AI being a bubble and waiting for the crash, but I suspect this event could actually be particularly terrible for us.
These days, I find it hard to predict the future. Things feel too complex, and the times seem different.
People say the AI bubble will burst, and I also think it will. But I want to think about how it might differ from other bubbles.
One positive factor is that during the internet startup era, there was almost no revenue. But now, big tech companies are generating profits and can absorb AI-scale losses.
The dot-com bubble burst, but the internet ended up being far more valuable than the bubble itself. I think the value was priced in early, and AI will be similar.
The core issue is always the same: there's a bubble and there are warnings, but you never know when it will collapse. And the people who act first on the warnings have to give up on the upside. The only winners are those who sell everything just before the crash. The problem is knowing when that moment is.
One difference from the dot-com bubble is that after it burst, a second wave of entrepreneurs leveraged the cheap telecom infrastructure that was left behind. So after the AI bubble bursts, will OpenAI and Anthropic fall and new companies emerge? I'm skeptical about that. Something feels different this time.
My biggest concern, though, is that if I suddenly couldn't use AI anymore, I'm worried about how long it would take to recover the coding skills I've lost after nearly a year of barely hand-coding. AI has become too deeply embedded in my life
There's a very nonzero chance that the AI expenses are not a race to recoup investments as they are a bet (with other people's money) that AI will lead to longevity or immortality, a bet made by aging tech oligarchs. Notice that Ellison is betting his entire company on it, which due to his secured debt is the same as betting his personal fortune in the most extreme case.
So they've been sold snake oil. This, along with colonizing Mars and mining asteroids, is complete and total science fiction.
This chart of free cash flow for the big tech companies should worry people:
https://www.advisorpedia.com/media/media/2026/02/23/trailing...
Somehow we've managed to turn software from a machine for printing money to a machine for setting it on fire.
Without AI there is no future US economy. You're all in.
im sure people will continue to want to eat food, live in houses, go on vacation, look at phone, send email, drive car, even if there’s no ai money frothing around
They can totally print more dollars to spend on things made somewhere else.
> Without AI there is no future US economy
I'm already on it, man, don't need to convince me!
You promise?
In a word, baloney.
The US produces stuff, and it produces ideas. Yeah, it imports a lot of stuff too. But it exports things like food, movies, and software.
Without AI, the financial sector may take a huge hit. But the US economy? It will still be here, still producing things that are valuable.
Eh, the empire aka, protection of free trade as a service is still alive and well with no viable alternative on the horizon. Every other empire is either not free trade or - just plain mercantilism and thus will exploit way more then the quite "idealist" us- trump is just a taste of what the others in power would feel like.
I think the Hormuz saga and how China just quietly stopped importing oil are signs that we are nearing the end of the "free trade as a service" from the US. There are so many aspects of that I do not understand yet: how much oil trade will be done in yuan in the future is one that leads to other question about the current wars and what will happen when those end.
Without AI the Chinese have won already.
If you haven't noticed it's because you've stayed too close to the homeland.
With AI the Chinese have also won.
China won already with or without AI. You just did not realize it yet.
The US is in a position where they have to debase the dollar all the way to zero and the Japan carry trade will slowly unwind for the worst as soon as the Bank of Japan (BOJ) begins to increase rates in more than 30 years.
Someone has to pay for all the trillions of dollars of debt for the data center build out that is being hidden from the balance sheets of the big tech companies.
That machinery and attached industries would keep on paddling for a while.
I don't think the country that singlehandedly messed up the global economy by launching a pointless war on Iran - without even coordinating with its allies! - that predictably got the Strait of Hormuz closed gets to call itself protector of free trade anymore.
The US has lost its appeal as a hegemon. Constantly threatening allies with invasion isn't helping either.
Open a history book- before the us century- this stuff was the norm. Its what is to be expected after the us- retreats inwards. China blockading japan. The world blockading china. Russia blockading eastern Europe and vice versa. The shitty old world pre the ww2 order.
Indeed. But the US can't do this and also assume people will still accept it as hegemon and protector of free trade.
It's lost those titles when it started behaving that way.
Isn't the US like very clearly anti free trade at this point?
It is pro-kicking out oppossition to the system that uses the system to grow in power without contributing anything back to it. So the us has abandoned the "Wandel durch Handel" ideology- which lets admit it- was a brutal failure. Russia did not change. China changed back. None of the countries in the middle east got democratic or "contributing to the worlds allmende". Turkey is sliding back.
So its cleaning the house of parasites o clock.
They still need the protection and enablement of free trade.
Those analyses tend to not consider the government interest in having AI as a weapon: now AI labs are fully part of the military sector, making them much more supported by the system
This has been my take too. I think that the revenue requirements for making the industry profitable are likely never to be met, but that the US government sees AI as a strategic necessity and will ensure that the companies that are pushing the technology forward will be paid huge sums of money if they ever become insolvent. It's not ideal, but if banks are too big to fail, I'm guessing that OpenAI and Anthropic are too. This is doubly true since the defense sector is already integrating these solutions into their systems and actively using them in wartime activities.
This is a forbidden subject here. In my opinion, the foremost subject in the entire world, but for whatever reason seldom discussed and always dismissed.
All four frontier models contracted with the Pentagon. Claude already having been involved in Operation Abduct Sitting President & Use Active Denial Systems to Make People Vomit and Shit Themselves 'because'. Institutional and corporate capture . Foreign nations now having our models trained to facilitate tailored political orientation. And the fact that these systems, for all their limitations perennially bewailed, have tremendous capacity as force multipliers for all the things civil rights activists were previously terrified about apparently just before all our problems were magically solved.
Snowden docs x³
Things that once took fusion centers, thinktanks, research, etc and thousands of manual processing hours -- now real-time, instant, with real-time 'understanding'.
I think the AI crash will be primarily comprised of collectively lamenting all the obvious things we overlooked.
And with great irony, the only competitor (or savior) to total US asymmetric top-down domination is China. Or wait, I think I got that backward... https://www.wired.com/story/super-pac-backed-by-openai-and-p...