alt Hacker NewsI'm not convinced the AI bubble exists
Comments
>There's a weird contradiction you see from AI skeptics where they're afraid of how AI will change the world, but also seem convinced that AI will only ever produce slop that nobody would ever pay money for. [would -> should]
There is no logical contradiction even, just another race to the bottom, drying out of markets. The good ol' disruption playbook, like Amazon shrank or destroyed the markets to capture them, only now they are targeting all white collar jobs.
The conclusion starts: >I didn't write this article as some big takedown of the AI bubble theory. Even with this evidence, there are serious risks to the financial sustainability of the AI industry.
But the presented "evidence" is largely only: I don't know, nobody can tell the future. I mean there are also good reasons to be skeptic of skeptics like Ed Zitron, but Idk "modest valuation" seems like a big stretch. Also, I'd challenge the notion, that you'd only ever have to build that data centers _once_. Especially with the AI pace the chips are said to be replaced every three years, while the building itself could last 25 yrs. So even, if the run at the frontier cools down, I don't think running your data centers and models would come for free.
>Shameless Self-Plug I’ve applied to seek the Liberal Party of Canada nomination in Sudbury East—Manitoulin—Nickel Belt
Why do I always jump to the HN comments before opening the link
> Most (or maybe even all) programmers and mathematicians are worse at their respective field than an AI swarm equipped with a good harness, and a good deal more expensive as well.
We can START discussing this when someone defines "worse" and we all agree on it. For SW engineering, we've been trying to define "good" for decades and we still don't know shit.
> There's a weird contradiction you see from AI skeptics where they're afraid of how AI will change the world, but also seem convinced that AI will only ever produce slop that nobody would ever pay money for.
No, that's a total straw man. For example, I am afraid of how AI will change the world, but for that it doesn't have to stop producing slop. I'm afraid of how the slop is CURRENTLY changing the world, people and societies. Frankly, I don't even care whether someone ever pays for the slop. This isn't a "I'm going to lose my job" kind of fear, it's a "we're running education into the ground and rapidly transforming into a low-trust society" kind of fear.
It’s orthogonal to the point of the article but:
“AI can't replace lawyers and doctors, but they can automate their paperwork and free up time for them to do other things”
So my question is - in software engineering (where this has bitten hardest): does anyone know _anyone_ who is less busy and has time to “do other things”? Nope, nor me.
The bubble is not about whether AI is profitable or whether profits can grow.
The question is how much? If they grow 5x, then these companies are very profitable, changed entire industries,... etc. Yet, the bubble still pops because investors assumed 10x at least.
I liked his point about bubble theories being more about antipathy towards the tech firms rather than rational thought. I definitely experienced this myself, where acquaintances are like "the bubble will burst any time now" and when talking to them, it's very clear that it's an anti-AI stance that motivates them to say it rather than any sober economic analysis.
My thoughts has been: what would happen if GPU prices fell 50% over the next weeks?
My gut feeling is that there would be such a massive demand to buy them to run our own models.
I am also not sure there is a (structural) bubble.
I do speculate whether some companies will survive, also some of the big ones.
>Anthropic's S-1 filing says that they make $2.08 for every $1 of compute. In other words, once you ignore the cost of building infrastructure, AI is very profitable.
If you ignore the cost of building cars, they are also very profitable.
>Profits Can Grow, By A Lot
They can also decline by a lot, and Anthropic's S-1 suggest that is what is happening, as their losses are widening, not narrowing.
>Google Has No Moat, But Neither Does McDonald's
That's why no one is sinking several trillion dollars into a new fast food chain.
>Alphabet, Meta, and NVIDIA are all AI companies that have seen their market capitalization explode over the past few years, but all of them have stable P/E ratios in their 20s.
The "E" in their PE ratios is circular revenue that is ultimately derived from debt and equity, not real end-user demand from profitable companies. For some of these bubble companies, if you subtract the debt-fueled private stimulus money, their actual PE ratios skyrocket.
What a weird thing to say: a business is profitable if we ignore a key part of its cost base. That’s true of any business, you know. Even if they make a 100% coverage of every request served, it’s still a far cry from the usual internet business model where responding to a request is almost free and most of the expense is the initial investment in software and its upkeep. This is why Alphabet used to be flush with cash and we get free services.
The trouble with Anthropic and OpenAI is, that not only do they need to get operationally profitable, they also need to be sufficiently profitable to provide a return on capital on all the money they have burned to date. Much of which has gone (and keeps going) not to assets but to expenses like the electricity bill. Thus, once the models are efficient enough to be served profitably, a relatively small amount of assets needs to be generating income for a relatively large amount of capital. It is my understanding, that currently the efficiency of the models does not grow even fast enough to outpace the need for new capital.
One more thing, that gets conveniently ignored when talking about AI cost is the cost of training the models. From one hand, there’s but a few institutions capable of building them and, from the other, it’s like the models grow on trees. Yes, you could run your open-weight model on your own hardware and pretend it’s profitable because you don’t have a comprehensive costing model and the infrastructure costs from security to electricity to risk are mixed in with your general cost base. But someone has spent astronomical amounts of money training that model you run and would probably like to get it back with some interest. Let’s hope that return is just money and not, say, the ability of a certain country to control the global software development process.
> Google Has No Moat, But Neither Does McDonald's
Sounds like a valid argument until you realise McDonald's is a real estate business.
There are many problems with this article. I think the bubble is real and at some point the investor money will fall off a cliff.
The value is there but the cash burn rate is extreme to say the least; and the moat is only in what you haven't released yet.
"If the AI industry has to pay off 5 trillion dollars over the course of 20 years, that's 250 billion dollars a year", 20? more like 5 years befroe they degrade and be useless
reminds me of Dow 36,000 [1]
I think one of the things that people analysing the finances of the AI sector often miss is that the capital expenses associated with data centre buildout involve spending a lot of money on compute hardware which depreciates and becomes obsolete very fast.
Large companies have often changed the depreciation schedule for server hardware to 6 years rather than 3. I think it remains to be seen if this is justified as it’s not as though the pace of change in hardware has slowed down by 50% arbitrarily.
I think it’s pretty obvious to me that a data centre full of ampere GPUs is worth less than a data centre full of Blackwell GPUs. I suspect this is also why AI companies prefer to talk in terms of gigawatts of compute, as if efficiency per watt does not exist.
Data centre costs are not one-off. Some are, for buying land and building buildings and infrastructure. The actual compute however is a constant investment. I’m not familiar enough with the market to know what proportion of expenses are ongoing but I bet it’s higher than people think.
The 16% yield on Coreweave bonds probably indicates otherwise.
His analysis fails to consider trend lines in inference costs. A couple of years ago inference required a supercomputing cluster. Today, you can run quantized frontier tier models on video cards with a $1600 MSRP. At current trends we'll be running frontier models on a typical gaming computer setup in a couple of years. That is going to destroy centralization possibilities.
It also fails to consider the impact of 'good enough.' A lot of the build-out right now is in anticipation of trying to push the frontier level. But there comes a time, a time we're already arguably past, where models being used for a specific task will already be able to complete that task efficiently. Going way beyond the present should enable LLMs to solve ever more esoteric problems, but in economic terms if a company's goal is to harvest corn, and your corn harvester works well enough, they're not going to care whether or not that corn harvester can also solve Fugazi's 7th Great Conjecture in exomathematics.
I don't think these companies would be calling for regulation if there wasn't a bubble. I think they know that these models cannot keep the same progress forever, and potentially not even a lot longer. Instead of admitting this to the shareholders, they instead want to go the government and say "REGULATE ME DADDY", and blame them for not being able to keep all the promises that they're implying.
I'm not convinced by this article. It's as weak as it is long.
> The reason why AI companies are spending so much money is to rollout data centres to run the AI in the first place. These expenses are one-off purchases, not regular operating expenses
Companies are investing in scaling models. Doesn't that mean that they'll need to keep investing to stay on the edge? After all, when they lose the edge they lose customers.
> For starters, the average person isn't going to use a locally hosted model, ever
If open weights weren't a problem for these companies, ram altman wouldn't have accused the people of stealing his compute.
When are we getting an AI_OS? I just tell the prompt what to do and the entire OS is run by an agent (likely Linux with sudo permissions). Everything including browsing and copying files and what not is an agent driven task.
This sounds like a nightmare but so is using windows and office everyday and everyone is already using it.
This is a remarkably weak article to be getting as much attention as it is. It would be easy and boring to shred the whole thing systematically, so here's one example:
> When a bubble happens, it's usually preceded by the companies involved reaching obviously inflated evaluations [...] Alphabet, Meta, and NVIDIA are all AI companies that [aren't that]
Alphabet, Meta, and NVIDIA are not AI companies, and [I don't really need to explain this, right?] ignoring the inflated elephant here is a hell of a thing.
Honestly, criticising this is beneath me (sorry) and I don't really want to be here, doing that. Oh, but the irony. This article ("I'm not convinced the AI bubble exists") -- this copium -- being a busy front page submission today is such a neat way to demonstrate the bubble's existence. Truly, a thing of beauty.
Just replace AI by block chain or Internet and you'll see that none of these arguments hold water to claim there's no bubble.
> I don't think the rhetoric about the AI bubble is really about finances
But that's purely what bubbles are about though. People see something become popular, want to be in, over invest, and the thing becomes overvalued.
Internet was a bubble. I don't think anyone would claim the value people saw in it doesn't exist, and there's way more investment in internet today than there was when the dotcom bubble burst. It was a bubble nonetheless.
> Nomination in Sudbury East
Go Bulldogs!
I am convinced for a long time. Just the way the companies CEOs talk and move suggest that they are looking for infinite investment without delivering AGI or singularity and we are always "almost there", though it does not arrive.
AGI and singularity are the Nuclear Fusion of energy.
To write such an article with zero mention of interest rates tells me everything I need to know
Like others have said, it is not that AI/LLMs//ML is not perfectly valid technology, it is a auestion of whether the current investment into it is justifiable relative to the demand and profit. We still buy and sell tulips y’know.
> Outside of generative AI, self-driving cars are becoming a reality, with some evidence to suggest that they're even better than human drivers.
Haven’t they have been just around the corner for 10 years now? And the problem is not technology as such, but the fact you are inventing trains that use roadnetwork hardly large enough for cars?
> AI adoption takes time, but there's no reason to think that AI usage isn't going to massively skyrocket, even if AI doesn't get any better than it is today (and it certainly will get better). In other words, there's massive unmet demand for AI companies to profit from.
[Emphasis mine]. This might a good example of a person writing and thinking with an LLM and letting the model enchant you with its sycophantic bullshit. We might speculate endlessly about AI or any technology, but none of that is a basis on which you should throw money at, especially not as a retail investor. We can see value production of AI in software development, but the productivity increase promised is either not in the data, or it is too early to say yet. So the real valuation of AI is really difficult say still, a lot of things needs to happen, before it can be seen to produce concrete value increase. People need to stop really really fucking hating it for one.
It kinda pointless to talk whether there is a bubble or when it will pop if there is. It might happen today at market open or never. Timing it is a fools errand. Also it is kinda pointless to shout into the void how much you love/hate AI. Better use of time would be to develop the different systems and applications that will produce value for somebody. It all sounds too similar to crypto to me still…
Going in reverse:
"The AI Industry Doesn't Look Like Any Bubble We've Seen Before" - It looks exactly like the two railroad bubbles. Potential profit is there, in the future. But you have to invest in the infrastructure first. If you can't make a profit and pay back the loans fast enough, the infrastructure goes bye-bye, along with all the funds used to finance it.
"There's Still Plenty Of Capital To Raise" - IPO is not how you raise a trillion dollars. A million, sure. A billion, maybe? But a trillion puts retail investors - mom and pop - at risk. IPO is not how you raise funds for a railroad.
"Once AI companies become profitable" - there's the big red flag. This isn't Uber or Lyft that can float along on tens of millions in corporate hand-outs every quarter. They need 100x that much money, and nobody has it to hand out. So you can't wait around for profits. There's no magic cushion to float on.
"For starters, the average person isn't going to use a locally hosted model, ever" - They will in 4 years. "normal users don't want to have to deal with the hassle or expense of setting up something" - What hassle? You download a program with a fancy GUI and run it. There are a half dozen of these projects already, that provide ChatGPT on your computer from one binary. All you need is a GPU.
"Profits Can Grow, By A Lot" - Only if they remain cheaper than local hardware or open-weight inference providers. Which won't happen because they owe more money than they can make.
The sign of the bubble is all the investment made not based on stable returns, but on hype. The weird geniuses from San Francisco said the future will depend on this one technology, all work will be changed forever, and only their company can deliver it. If you don't get in now, you'll miss out on a fortune! That hype just doesn't work out. And the proof is here: the superhuman machine sucks at writing, the documents it creates are lies, and we seem to be spending more on it than we get out of it. Because it's not magic. It's just a machine.
Meta, I have constructive feedback on your choice of titles with this post https://news.ycombinator.com/item?id=50415422
There are very few people in the world where when they declare that they have an opinion on it, that the declaration itself is title worthy.
There's a weird contradiction you see from AI skeptics where they're afraid of how AI will change the world, but also seem convinced that AI will only ever produce slop that nobody would ever pay money for. You can't really have it both ways here.
It's more accurate to say that there's an OpenAI-Anthropic-X bubble. Financially sustainable investment in LLMs is useful and not going anywhere, as China shows. Dumping trillions of dollars into companies that make models that are 1.2x better than Chinese models at 100~1000x the cost, ehhh... good luck with that.
This analysis marginally better than Rob Ford smoking crack, I'm not convinced you should run for office, but if you succeed, please get us more Shorsey.
Putting data and rational analysis aside, the style of this post is so similar to what I was seeing during the crypto bubble.
Sure, the largest IPO in history was slashed by fifty percent before it came onto the market because no investor wants to underwrite its original offering, but it’s not a bubble
"If you check any online discussion about the financial situation of the AI industry, it's essentially taken for granted that the bubble exists and is going to pop any day now."
-------------
If everyone, or even the majority, believed that there is a bubble that's going to pop with near certainty, that would tank stock prices immediately. Why would anyone hold a stock they know is going to go down precipitously? If the majority of people know it's a bubble, the bubble immediately pops. AI stocks still have worth, so most people believe there is no bubble or that prices will go up before going down.
People talk about bubbles. They worry about bubbles. For the most part, though, they go on investing as though there isn't going to be a bubble. Every bubble looked like a solid investment to most people right up until it burst, or it wouldn't have been a bubble.
So, no, most people do not think AI is a bubble. That doesn't necessarily it's not though.
if i plug my ears and close my eyes the 16% neocloud bond yields suddenly don't look so bad
am i getting this right`?
dump n pump. day trading. long term strategies. build up for the next big thing or delusion or hysteria or lie.
if you read mainstream narratives like "AI bubble", translate it to several millions of people making lots of bucks via a lot of fraud mixed into some good, genuine, valuable work of productive people.
investors and wall street people doing anything financial is not tied to bottom up signals but to insider knowledge and top down supply chain LEGAL manipulation by people not directly involved in the actual work done.
or it's a wrap up ( made money, time to shift or prep something that will hypergrow again, maybe already re-pre-hyped ).
or it's reactions to having maxed out some part of the supply chain which had negative consequences to some other market which reduced overall gains enough to realize "oops", if we don't re-act now, future gains in other markets will be reduced almost catastrophically.
bubble talks always feels like a mix of hyped people really getting fucked over by fraud runners
the details are not too relevant for the moment as I can't have play money to invest but the narratives are interesting because they directly translate into what I read in the suggestions and advice, subliminal or direct, from people to other people.
Profits will drop from competition. McDonalds has a moat.
Those are errors just from the subheadings.
It’s the unemployment that will disrupt not the AI.
This person lives in a bubble. IPO of Antropic is literally a sign there is no more capital in private markets and it’s up to the public to take the shit bucket.
Let’s not even talk about the second headline that costs are constrained, while one company spends few dozens off billions per quarter.
The thing is, the bubble exists because even though there is no guarantee of any big profit coming from the big techs, there are billions of dollars funneling into these companies. If you look at their current numbers, they are drowning in losses. Yet they are forcing to keep this bubble alive
The discussion of the “AI bubble” often conflates two different things: (1) is AI going to justify the valuations we are seeing? and (2) are the specific AI companies we have going to justify their own valuations.
The rapid progress of multiple closed frontier models, along with the rapid progress of open models, suggests that LLMs will be like electric motors, not like social media. Are electric motors worth a huge amount of money economically? Sure. But none of the companies who invented electric motors are especially valuable today, because the technology became a commodity.
> The reason why AI companies are spending so much money is to rollout data centres to run the AI in the first place. These expenses are one-off purchases, not regular operating expenses. Once all the data centres are built, AI spending is going to plummet.
Because hardware doesn’t need updates. Newer hardware is more capable which gives competitors with newer hardware an advantage.
> Outside of these expenses, AI doesn't cost that much to run
Outside the expensive stuff it’s cheap.
So if a shoeshine boy tells you which stocks to buy, it's time to cash out.
But if the shoeshine boy tells you it's a bubble, it's time to buy?
If only shoeshine boys still existed, I would know what to do.
> If you check any online discussion about the financial situation of the AI industry, it's essentially taken for granted that the bubble exists and is going to pop any day now.
You complain that people take the bubble as a given. Yet you continue by doing what a lot of people on the other side do:
> Most (or maybe even all) programmers and mathematicians are worse at their respective field than an AI swarm equipped with a good harness, and a good deal more expensive as well.
This isn't established. Like, at all. There's serious doubts as to whether AI-driven codebases remain maintainable, understandable and trustworthy over time. You look at a snapshot and conclude, while riding the hype wave.
The technology itself is here to stay i believe, but until they figure out a way to make it more efficient and not just burning money, a part of it does feel like a bubble
I think AI is fine. I suspect there is a data center bubble.
ok
“Its a bubble”
“Its a bubble”
“Its a bubble”
“Its a bubble”
“Maybe its not a bubble”
*bubble pops*
every time.
Well, first of all, these are not one off purchases. GPUs do break and newer generations come out rendering the old GPUs outdated. I don't know if the 2.08 figure is very valid. This is based on costs of leasing GPUs from companies that may not be making money on the lease deal. What I do know is that Anthropic weren't profitable in Q3 and I'm hearing rumblings they are losing even more money in Q4.
Also Popeyes never had over $800 billion in uncancellable compute agreements. And I'm glad you're having fun with GPT, but I've moved all my projects to use DeepSeek 4.1 Flash. There are plenty of different providers. And as I use OpenRouter the change is one line.
The reason I think the PE ratios are lower than previous bubbles are because everyone already knows its a bubble. Also all these tech companies already have massive businesses and AI is not going to change that.