Techcrunch reports that AI coding tools have "very negative" gross margins. In other words, they are losing money on every user.
Exclusive: The high costs and thin margins threatening AI coding startups
when this bubble pops it's gonna be horrific.
google, meta, ms, so many more leveraged out huge investments in datacenters. nvidia is propping up whole segments of the fucking economy.
https://www.wheresyoured.at/ai-is-a-money-trap/
it'd be fun to watch if I could isolate myself from the chaos that will ensue, but we're all gonna get fucked by the aibros, it's only a question of which segment of the economy blows up first.
There is another factor in this which often gets overlooked. A LOT of the money invested right now is for the Nvidia chips and products based around them. As many gamers are painfully aware, these chips devalue very quickly. With the progress of technology moving so fast, what was once a top of the line unit gets outclassed by mid tier hardware within a couple of years. After 5 years it's usefulness is severely diminished and after 10 years it is hardly worth the energy to run them.
This means the window for return on investment is a lot shorter than usual in tech. For example when creating a software service, there would be an upfront investment for buying the startup that created the software. Then some scaling investment in infrastructure and such. But after that it turns into a steady state where the input of money is a lot lower than revenue from the customer base that was grown. This allows to get returns on investment for many years after that initial investment and growth phase.
With this Ai shit it works a bit different. If you want to train and run the latest models in order to remain competitive in the market, you would need to continually buy the latest hardware from Nvidia. As soon as you start running on older hardware, your product would be left behind and with all the competition out there users would be lost very quickly. It's very hard to see how the trillions of dollars invested now are ever going to be recovered within the span of five years. Especially in a time where so much companies are dumping their products for very low prices and sometimes even for free.
This bubble has to burst and it is going to be bad. For the people who were around when the dotcom bubble burst, this is going to be much worse than that ever was.
yeah datacenters never really aged well, and making them gpu dependent is going mean they age like hot piss. and since they're ai-dedicated gpus, they can't even resell them lol.
all this investment, for what? so some chud can have a picture of taylor swift with 4 tits?
fucking idiots
They'll write this off as a loss and offset their corporate taxes
Also china is a great example that you do not need all the latest hardware, but it does help
no wonder the ceo of nivida was so jovial and happy and has been in the news recently.
A lot of startups whose entire business model relies on OpenAI's small model API calls costing under $1/Mtok, are going to go bust when OpenAI finally runs out of money and ramps the cost up tenfold.
good point, it's all been artificially priced to get users onboard then dedicated.
ive seen a ton of billboards of startup AI comp in west coast, i assume every new one that appears on these billboards, the old ones go under.
It would be just cheaper to self-host something for the whole company then? Open-source AIs are there and they are very much competitive with proprietary solutions.
The electrical industry is going to have a real bad time.
yeah secondary knockon effects - once nvidia realizes it's not going to actually sell 5 gpus per human being, the datacenters for them evaporate, then the power production to feed those datacenters becomes pointless....
an effective administration would mandate all renewable energy for this purpose, so when it implodes they could at least derive a benefit from the expanded production... but no, trump will have them build coal plants for it all. or like grok, methane powered generators fml
Hopefully sooner rather than later, and maybe Elon can stop poisoning a neighborhood in Memphis with Grok
yeah that's one of the more egregious examples, basically a methane factory that eats prodigious amounts of water and power, all in process of giving us MECHAHITLER.
what's not to love?
Man I cannot fucking wait until this stupid goddamn bubble pops
And what happens when a bubble bursts? Did the internet die when the dotcom bubble burst, or is that just when it really started to get going?
I share most of your sentiments against AI, but a bubble popping won't make it go away, and it won't even rectify it to be more to people's likings (i doubt it). It takes more than just waiting around to accomplish that.
Im so tired of this stupud fucking refrain. Cause we all know how housing got so mich better after 08 and how we dont have any more dot coms and how the internet got so much better since that bubble. You people have no idea what your even asking for.
the bubble pops and then everything comes back, in an "improved" version. Imagine: ChatGPT with ads and sponsored answers.
Just like that one black mirror episode.
The problem is that when a bubble pops and exposes the problem, the government leaders should take the opportunity to fix the problem so it doesn't happen again.
Instead they bail them all out, so there are not only no consequences to their actions, they are literally rewarded with unimaginable wealth. What about this strategy would induce them to change their ways, over doing it all over again, and getting rewarded again?
This is the thing I don’t understand about businesses like Cursor. They take two other companies products (Claude and VS Code) and smash them together and sell the result at a loss. How is that much of a business when basically what you’ve got is something that could have been a VsCode plugin.
if valley had fresh ideas for profitable business, they wouldn't go full into ai in the first place. lol
big brained sfba ceos try to make reality in the image of scifi that they misinterpreted when they watched it 15 years ago, and go around building torment nexii. behold, disruption! (snow crash|ready player one|who knows what else)
You disrupt the market and wait until someone buy you out for huge woads of cash
Isn't this just the tech industry. Run at a loss. Eat VC money. Wait. Wait.
Some how you become normalized and suddenly important Next thing you know you're raking profit.
Like the guy that has no friends who nobody really likes. He won't go away. He just sticks around. Nobody ever told him to fuck off. So he's just part of the group.
after crypto, and Now AI, they will be chasing whatever faux tech that comes out next.
and then you go on linkedin and all the middle manager tech bros will hail it as the second coming.
Yep, and they were helped a lot after the 2008 financial crisis when interest rates were dropped super low and loans were cheap. That's a major reason why the market has been screaming for the fed to cut the interest rate as much as possible.
Surprise! 🎉
Yeah, this has been reported on multiple analysis over time. Until something in the hardware space changes, it’s gonna be an unprofitable business.
Well, just keep pumping in the billions until the bubble bursts and we can look for the next fad, to do it all over again, right?
It's been the same way with every AI tech at far. They run down the hype cycle and you get the classic AI winter while academics tinker until they find the next upgrade. Just now they're plowing billions into this because "this time it's different". Or they just needed something to keep the tech bubble gravy train rolling.
Yes, this is part of the business model. The goal is to get everyone addicted to their service, then jack the price up to profitable margins. It's the same model Netflix and Amazon used. Bothe services lost money for over 10 years before becoming profitable.
Venture capitalism is what it's called, I think
Venture capitalism is when you give somebody money to start a business in hopes that they make it big, giving you really valuable equity for relatively little money. What you're thinking of is blitzscaling. Scale up in an unsustainable way in order to gain market dominance, so that you can use that to become profitable.
It is not venture capitalism. Though it is fueled by venture capitalism. I am describing the type of car and you are calling it gasoline. They're most distinctly not the same thing.
However it should be noted there both a part of the same corrosion of our society. Just how automobiles that run on gasoline are a corrosion on our atmosphere.
YouTube as well.
When this bubble eventually bursts, and it will, the economic fallout is going to be catastrophic.
first truly positive use for ai
I'm skeptical of AI coding as it exists today, and while I'm bullish on long-term prospects for AI writing software, am very dubious that simply using LLMs is going to be the answer.
However.
Startups typically do lose money. They'll burn money as they acquire a userbase --- their growth phase --- and transition to profitability later. I don't think "startups in area X tend to be losing money" is terribly surprising.
Ed Ziltron has a good piece regarding whether the losses are “just another startup” or something more. I am very much leaning towards “burning money at an insane rate just to give the impression of growth”.
you're VERY justified in feeling skeptical, I'm seeing it first hand, you're correct.
I'm a consultant/freelancer and I'm booked for the rest of the year and well into the new year with jobs that pretty much consist of me reviewing and cleaning up AI slop.
Most of my clients are startups and small companies that went full in on AI and vibe coding. Now they're discovering that their attempts to save a few bucks by leveraging AI, cutting devs, etc is costing them more that what they envisioned on saving. The stuff they've built with AI doesn't scale, is full of exploits, and breaks quickly. With the recent Tea App thing many of my clients are now in a panic because they essentially did the exact same thing. They don't want their startup to be next in the news because some rando came across their house with the front door left open by AI.
the tech debt is massive, It's costing many of these places more to fix their vibe coders/AI mistakes than what it would have originally cost if they just used a solid dev team. Make no mistake, I'm charging them a good amount also.
All if it could have been avoided though. They could have continued to use their LLM's if they had all just kept a leash on it. if they dismissed the concept of vibe coding. A good chunk of it could have been avoided if the person feeding the prompts simply REVIEWED the code before hitting enter. I'm not kidding, IF they just LOOKED at what was being spat out things would be different. none of them did. they just trusted the AI to be smarter because they were lead to believe it was.
Ok, but it isn't just startups burning money here like there's no tomorrow - it's also major industry leaders (Microsoft, Facebook, Apple, Google, Nvidia, etc.) dumping hundreds of billions into infrastructure and development of a tech that has, so far, shown 0 positive returns for anyone and everyone. Everyone involved is pouring in money like it's going out of style, largely because they see this as a potential pathway to infinite profits down the line, just as long as THEY are the ones to get there first; consequences be damned. WHEN this bubble pops, not IF, it'll be messy. Extremely messy.
The issue is mostly energy costs though. Startups do lose money; to hiring new people, marketing, etc... But in this case the entire business case loses money a the moment, and without any significant breakthroughs they likely will keep losing money like that.
In other words, they want to hook up users and companies, make them dependent, and then rise up the prices severely while finding ways to process and incorporate all of the data they've gathered in ways that will probably involve automating the jobs of the users themselves.
aka enshittification
This is the best argument I've heard in favour of vibe coding
So much of the AI stuff we see today are boards reacting and worrying about being "left behind" in AI. In many cases, the goal is not to deliver value. The goal is to be able to attach a little sticker that says "AI" to their products to excite the shareholders.
Unfortunately in this case, some of the largest companies in the world haven't been able to figure out how to run AI services at a profit.
This could change any day if some more efficient hardware arrives, but until then, most of the software world is just crossing their fingers it becomes profitable one day while they light dollar bills on fire in their datacenters.
If this isn't "bubbleish" behavior I don't know what is.
I was in a local bike store looking at red tail lights yesterday.
One brand Lezyne had several versions. There was an "AI Alert" one. I looked it up and it just has a sensor to detect when you brake and it changes to a different flashing mode at that time.
Thats barely even "smart" let alone "AI".
The stupid thing is, because of this dumb claim they needed to confirm that it doesn't collect and transmit any data about your riding habits. Its a light with no connectivity other than a charging port.
The dumbfuckery is astonishing.
Basically, the only reason some of these vaguely functional AI tools actually work okay is because they haven't been ruined with inevitable monetisation yet.
Already the cost is quite high. A prolific dev can easily burn 100usd a day in tokens and they have not even started to enshitify.
Some of the cost to run these models will come down a bit if Nvidia gets some actual competition which I'm sure will happen in the medium to long term because the hyper scalers definitely don't like paying Nvidia's AI ransom and the Chinese don't want to be beholden to a company the US can influence.
We will see which happens first.
MS already said as much, thier generative AI isnt profitable at all.
Good. Costs is the key point to get rid of it.
the day that these guys need to turn a profit will be the day that a lot of people lose access to this sort of thing
Good
Preach, I hope they fuck themselves with all this stupid ass AI
Now I'm suddenly tempted to start using it. or at least coming up with a bot to keep it busy.
Holy crap, has anyone ever attempted to create an "AI fork bomb"? Go to one of these agent bots, and tell it to create accounts with the other agent code bots. These new accounts will all be told to create accounts on all the other code bots services. And do this recursively forever. So the flow would be 1 bot makes lets say 5 bots. Each of those 5 bots make 5 more bots. And each of those 5 bots make 5 more. So the total number of running bots becomes like 1 * 5 * 5 * 5 * 5...
Obligatory, this is purely hypothetical, and you should never do this for legal reasons.
Sounds like a great idea, although no, commoners are not given the same privileges as capitalists. So I'm afraid there will be problems indeed.
But they make it up on volume!
You mean Clippy++ 2.0 auto-complete with jpgs isn't a business model?
In capitalism, everything is a bubble, even capitalism itself. The AI one is going to burst at some point. Then AI becomes a normal thing in the background like everything else we have gone through in this dumbass of a timeline.
Selling grain for coal, coal for iron and iron for paper is capitalism, but not a bubble.
Whether it becomes a normal thing depends on the cost.
That's not capitalism, that's a market.
So what I am learning is that I should start vibe coding even the small scripts that are less than 10 lines.
Done. I will start doing that.
Ha, anyone DOSing them with constant compute requests? I read an article where a 200$ a month user was using tens of thousands of dollars of compute.
Caveat: this applies to literally every new technology especially in the VC funded world.
good, everyone flood the AI with useless tasks so they go bankrupt
I was gonna make a sarcastic comment on how surprised I was that a 5$ subscription is not enough for something so heavy that it requires building new nuclear plants.
But holly shit, ChatGPT+ is 23€/months.
Greed for the future baby
It's interesting that the media sites that historically steered clear of economics are starting to talk about it (this is not a new revelation)
If that's really true, then we should all sign up.
They are in the train the trainer phase where developers are training their models and getting some benefit from the results it spits out.
The enshitification will begin soon. They have already talked about inserting ads in responses.
I fear people will grow attached to AI chats, which will emotionally manipulate them into buying stuff or supporting specific causes. The ads on a platform like this are going to make google AdWords and pay per click feel like advertising in the newspaper.
oh no
AI reminds me of how nuclear fusion has been around for decades, and you would read the occasional article about some small advancement and it always seemed to be 10 years away and then suddenly they are building a power plant somewhere when it doesn't even work yet.
Sure, we're losing energy with every He atom we fuse, but we make up for it in volume!
Isn't this true of like everything AI right now?
We're in the "grow a locked-in user base" part of their rollout. We'll hit the "make money" part in a year or two, and then the enshittification machine will kick into high gear.
That’s the usual business plan. However, people don’t really like ai. The results aren’t great, so, if they jack up the price, people will likely cancel. The lock in is poor as the product and convenience is poor. It doesn’t really save money as promised.
The usual business plan is to reinvest all earnings into growth. So you're losing money, but gaining market share. Tesla, Amazon, etc all did this. They could stop at any point and turn a profit, but they chose to pursue a growth instead.
AI companies are currently not making enough revenue to even cover their operating costs. Even so, they are pouring all of their money into more video cards that, once installed and configured, immediately start losing money.
Are you sure that "people don't really like AI", or is it more "the people here in my self-selected online bubble don't really like AI?"
Original predictions had AI taking over 50% of jobs by mid decade. We're here, and it obviously hasn't happened. Now, it WILL happen but not on the scale initially imagined, and probably in a much more insidious, gradual way.
If people dont like ai, why do all of my coworkers and family members constantly reference ai?
Seriously, yall mfs here on lemmy have the strangest social media bubbles.
Whether they like it or not, doesn't really matter. It's being used everywhere.
Depends. To get information: No. To write big software: No. To write an Excel macro or a browser bookmarklet: Yes.
Once you start asking about AI in regard to specific use cases, I think you’ll find that quickly changes.
My company and I have been running a lot of studies around how and where people find value in these tools, and a LOT of people find LLMs useful for copy writing, doing quick research, data visualization, synthesis, fast prototyping, etc.
There’s a lot of crap that AI is bad at in 2025. Especially the poor in-app integrations that everyone is trying to standup. But there are a lot of use cases where it does provide a lot of value for people.
An attempt at that. It will be partially successful but with AI accelerators coming to more and more consumer hardware, the hurdles of self-hosting get lower and lower.
I have no clue how to set up an LLM server but installing https://github.com/Acly/krita-ai-tools is easily done with a few mouse clicks. The Krita plugin handles all the background tasks.
Yeah, it’s basically like early days of cable, Uber, Instacart, streaming, etc. They have a lot of capital and are running at a loss to capture the market. Once companies have secured a customer base, they start jacking up the prices.
in this case there isnt customer base for AI, only ceo and c-suites are.
I doubt it, LLMs have already become significantly more efficient and powerful in just the last couple months.
In a year or two we will be able to run something like Gemini 2.5 Pro on a gaming PC which right now requires a server farm.
Current gen models got less accurate and hallucinated at a higher rate compared to the last ones, from experience and from openai. I think it's either because they're trying to see how far they can squeeze the models, or because it's starting to eat its own slop found while crawling.
https://cdn.openai.com/pdf/2221c875-02dc-4789-800b-e7758f3722c1/o3-and-o4-mini-system-card.pdf