AI is too expensive
wheresyoured.at
wheresyoured.at
Scroll back not too far and he was publishing criticisms that no one wants to spend actual money AI. Anthropic has shattered all notions of that since then.
Then there was the idea that even if people want it, we have way too much GPU capacity to ever be saturated. Now almost all providers are hitting limits.
Now, its the next iteration that even if people want to spend money and GPU's are at capacity, its just never going to be profitable. This may or may not be true, especially with more capable open source models that can be served at cost. But at this point, he mostly just brings up anything possible to downplay AI
There is a lot of hype right now about AGI destroying the economy, replacing workers, or even ending the world. Companies are embellishing as they run up to IPO. But there’s a lot of unhealthy counter-beliefs trying to take it the opposite extreme. Keeping up with AI news is about avoiding the hype monsters at either end of the spectrum.
I dunno, I guess I just don't like him.
I agree that he is very one-sided, but I doubt his hit-rate is much worse than other pundits in the industry.
I don't think that's necessarily a bad argument style (at this point it's by far the easiest way to argue that a Musk project won't work for example), but you have to be careful with it.
In this case, there are only two previous predictions being considered. Suppose the author would have been right 75% of the time given their available information; there was still a >6% chance of those predictions not working out. That's a pretty rough bound for ruling them out just on principle in a world where they do actually know a lot about how AI will progress. There isn't enough "track record" to confidently say much about the author's predictions.
Track record is not relevant if it is a prediction with all the arguments laid out in the open.
Which is our situation closer to?
>Anthropic Is Bleeding Out ...Anthropic is very likely losing money on every single Claude Code customer, and based on my analysis, appears to be losing hundreds or even thousands of dollars per customer. (Ed, 10 months ago) https://www.wheresyoured.at/anthropic-is-bleeding-out/
vs
>Mind-Blowing Growth Is About to Propel Anthropic Into Its First Profitable Quarter The startup expects a 130% revenue surge to $10.9 billion in the June quarter and its first operating profit, defying skeptics of the AI boom (wsj, yesterday)
If you look at Ed's piece there the facts are correct but he just moans about costs and ignores the growth. Most of his stuff is kind of like that.
Is bearing a lot load here.
It is a press release with a goal to create hype.
I.e. it has no relation to “facts”
… unless they would also share underlying numbers (e.g. cost/profit per customer, and why they think some things will increase)
AI is a neat tool, but I think it becomes a race to the bottom. Who can provide the minimum viable product at the lowest price. And that price has to be pretty darn low for it to make sense for companies and essentially free for it to make sense for consumers.
Ed Zitron is an analyst. His viewpoint is that AI is bad for whatever reasons and he does his job by trying to uncover those reasons and does a solid work. He presents a lot of insider knowledge that would otherwise be left unheard.
What are his alternatives? To stop claiming that AI is bad and pivot to "AI is good" writing? To quit writing entirely? To continue writing but in the beginning of each article list the things that he was wrong about in the past? What if it's too early for the things that seem to have been predicted incorrectly by him to materialize and in the end he will appear correct?
I think it's a benefit for society to hear the other side. There are plenty of pro-AI advocates.
It's a bad combination. There are better AI skeptics to follow. Endorsing Zitron, though, has become a "tell".
A pundit starts from a pre-declared conclusion and works backwards to generate the argument. An analyst lets the conclusion be dictated by the analysis.
Providers have been hitting limits and growing backlogs -- i.e. real money customers had committed to paying them but could not be realized due to lack of capacity -- to triple-digit billions EACH for multiple quarters now.
People are only just noticing the capacity crunch now because they're being directly impacted by Claude crapping out so much. But a superficial glance at the quarterly earnings of any of the hyperscalers shows that the AI compute crunch and revenues have only been growing pretty much since the AI boom took off. That alone should have raised questions about the bubble narrative.
Yet commentators like Zitron have been crying "bubble" all that time. I guess there's real money... er... engagement to be farmed by playing up the AI backlash. The backlash is real and understandable, but these narratives only serve to muddle useful discourse in exchange for some cheap rage-views.
We should draw a distinction between "AI is valuable" and "AI justifies its current investment levels." There's real productivity value in AI, especially for things like search, boilerplate, tests, refactoring, etc...BUT that doesn't mean every enterprise should let token spend grow without strict telemetry, cost-attribution and outcome-based measurements.
The teams that win here will not be the ones using the Most AI, but the ones that treat it like any other expensive production dependency, which means measuring unit economics, cap runway usage, properly align models with tasks(not just Opus everything), and scale workflows with ROI in mind.
The bright side is: this is a golden era of subsidized tokens. It will not always be like this, so now is the time to churn out your passion projects.
You could imagine a Moore’s Law-esque cheapening of the tech that coincides with the business raising their prices. That might look like a continuation of simply “using the tools” on the surface, but on the inside it would spell a gradual, meaningful increase in margin
The subsidies went away gradually and the prices leveled out in a spot where the services are heavily used. Uber became profitable. Ride sharing is affordable.
I think our $20/month plans might become a little less generous and the $200/month plan won’t always allow non-stop vibe coding, but I don’t think the prices are going to rise so much that users are priced out. Like Uber, customers will grumble for a while and then adapt to the new normal.
The big difference is that compute hardware is getting better. I think we might overshoot with data center buildouts to the point that compute becomes cheap, while hardware improvements continue to lower the cost of serving models. Over time the same service becomes cheaper to operate, opposite of Uber where driver wages are creeping upward.
My "priors" are that I raised a search fund and analyze companies every day.
I mean surely to make this claim, you aren't just making it up, surely you've done the research. In that case, what's so hard about sharing facts which you already know? I'm not asking you to put in a bunch of effort pulling financial statements and analyzing everything, I assume you've already done that because I doubt you'd make the claim otherwise.
Akin to an average cellphone bill. The infrastructure costs are comparable and the ROI would be 5-10 years for the current insane build out.
Yes, chinese and local models exist. But so do $20 cell phone plans. People go with what is convenient, works, and is readily available.
Where? You get unlimited mobile plans for like 10-20 euros in Europe
A better comparison is with how much PC costs went down during the 80’s due to IBM clones and Moore’s law.
SubQ was validated by at least one third party, not sure if we'll see more confirmation, but 5x cheaper costs is worth it. None of the frontier models care enough about cutting costs of their models, only being the best in benchmarks.
Using AI to write my software just takes all the fun out of it for me.
It feels like just reading a summary or recap instead of reading the actual full novel myself. Like it defeats the whole purpose of it. I write software because it's fun and it stimulates my mind and teaches me things and improves my skills.
It's hard to imagine a bright future for either company. The more hopeful analog might be Coca Cola & Pepsi. Both have a commodity product, but due to distribution networks and brand, they command a price premium. But it doesn't seem much fun to provide a high-volume low-margin commodity.
Basically yes but there is a small difference I think. Things like Uber took an existing and proven business model. slapped an App on top and priced the competition out of the market. The winner takes it all.
With "AI" it's not yet clear if there will be any winner at all because it's not yet proven if the business model is viable at the "real" price. Think air taxis across town, 5$ a ride, will work at scale for sure, probably won't work for 500$.
Google has had decades to accumulate intellectual and physical capital. Catching up quickly means spending >500 billion. If you can actually dethrone Google (admittedly not an easy task) then it will have been worth it. If not, I suppose it's wasted investment.
Now what happens when three or four startups vie for this opportunity at once? Well that's how you get $2 trillion in captial investments per year.
More realistically, it seems like someone calculated that it could still be profitable up to several hundreds of billions of dollars which explains the initial investment. And continued investment can be explained by trying to salvage the existing capital spend. But even if it's the best option those companies have now as far as a hypothetical goes, it still might not have been worth it.
If it’s a spending game, the incumbent has a huge advantage.
They didn't say you need that amount. They say how much you're willing to spend. Need = floor, willing = ceiling.
It's a very reasonable argument, except one fact: the chance that you actually dethroning Google is practically zero as Google also has capital, infra and data to train AI. The best plausible outcome is to share the market with Google.
> Catching up quickly means spending >500 billion.
What part of that implies it's about willingness?
The way digital cameras developed (hyuk hyuk) is arguably exceptional, definitely not a clean example.
Or any one of thousands of other ventures which could be more beneficial to humanity, the environment, etc.
An open source, networked search engine that doesn't have any ads, and we all collectively host, could do this. Someone is going to pull this off eventually.
You don't have to match the infrastructure of an existing player, you just have to have a profitable path to get there organically.
Oh, and not take the inevitable buyout when it gets offered.
If suddenly the money craze stops, meaning (1) AI companies investors want them to become profitable and (2) clients start being cost-sensitive to AI bills (which they are absolutely not currently), then everyone will switch to smaller, cheaper models that are enough for a lot of use case.
Sonnet instead of Opus. GPT 5.4 instead of 5.5.
Chinese models.
People keep comparing to Uber but Uber can't suddenly make it cheaper to operate.
I am exclusively using 5.4 because its only 1x and very good, but the github calculation showed my once $40 become a $680 billing
That is too expensive and not worth paying
People are going to decide its too expensive for everyone to use AI agents and un-subsidized pricing.
Hell yeah obviously. There's close to no doubt. So why do we think its not true now?
Unless you’re a freelancer you should never pay for your tools. You’re literally paying your company to do their work. If they want the benefits of an LLM they can pay for it. Otherwise they get what you came with: your hands and skills.
He never cops to the fact that he is constantly being proven wrong and changing his tune every few months to a new theme which he will abandon as soon as it’s not supported.
I’d like to think that if I was as catastrophically, publicly, consistently incorrect on the main thing I do, I’d have a little more humility than this charlatan.
I think competition is going to keep customer costs low if you’re willing to switch. Maybe people on expense accounts won’t care, though?
Feels like an unspoken rule here. Everyone wants to own a chunk of nuclear weapons and it doesn’t matter whether it’s profitable. You just need the nukes to survive and have a seat at the table
The math may look questionable but there are also senior people talking of automating all white color work in the next couple years. Even if that estimate is miles off on both time and % it’s still trillions. So crazy as the numbers seem it could still work out
Very dystopian but I'm not convinced it's a showstopper in the "this rules out such a future" sense
What if the things, on which they are investing, go bust? Well, they do calculate their risk when they invest on startups.
The overall picture? Not everyone's calculations will yield good predictions. Some of these cloud sharks go bust when, for example. OpenAI folds. The game is, winner gets it all. We are heading into monopolies in every layer.
“AI is too expensive right now” is an accurate title. Plenty of things could change in the future to change that. Off the top of my head:
* end user pricing
* breakthroughs in model efficiency
* better chips to run models
Any one of these things is easily possible in the next three years. Probably sooner.
In other words; right now, we're still in the "bait" phase. The "switch" comes later.
1) someone deepseeks deepseek lol:
Generates their own weights and figures out a way to determine all of the intermediate states.
2) places realize there’s real risk with using a model that might have things baked into it that produce specific flaws that could be security bugs, but only under certain conditions.
If people's dependence on their streaming service keeps them captive, just wait until people have gone 5 years without doing real work.
regardless, the parent comment is talking about third parties hosting, for a fee, the openly available models, usually outside china
They also publish papers talking about how to save kv cache and computation powers. Because currently they don't have the most powerful nvidia cards, training and inference efficiency is very import for them.
My issue with Ed is that he doesn't have the ability to draw the line. In the pursuit of making a point he goes so dogmatic that he is willing to make harsh statements that go beyond number backed predictions. Like in his piece "AI is really weird" he states about agents, "Probably the weirdest thing about this entire era is how nobody wants to talk about the fact that AI isn’t actually doing very much, and that AI agents are just chatbots plugged into an API.". That's a massive stretch to make. Just because he has a claim that the business doesn't make sense, he doesn't get to claim that agents are not capable of doing very real work. His assessment of cowork was "a chatbot that deleted every single one of a guy’s photos when he asked it to organize his wife’s desktop.". These statements damage his credibility and make it too easy to dismiss his writing as a rant of an angry man.
With the notable exception of TTI models, that description seems accurate to me. Is there any widely promoted "AI product" that is more than a chatbot in fancy dress?
I'm pretty uncertain around the businesses of foundation models but the tech is definitely useful.
My only side note of sadness here is that companies are more likely to implement such stuff in a haphazard way rather than anything actually thoughtful.
This will definitely happen, but the trend will end up towards automating a bunch of this stuff with manual checks on a subsample to ensure that it's working effectively. It's rather like moving from phone/mail based ordering to online ordering, in that it'll take a while but it's almost certainly gonna happen.
This blog is too expensive too.
Nobody who’s this insistent, aggressive and violative with their language of “it’s here and if you don’t adopt it you’re stupid and dead” has ever been right about anything. Nobody this desperate, insistent and forceful has ever had good intentions, good vibes or brought good omens — they are always bearers of some kind of con.
I think there’s two productive avenues for reaching the other side here. One is thinking more about the data centers - put aside the “overconfident and unaware of how hard it is to build data centers” hypothesis and instead start by assuming that “announcing and funding a huge data center and never actually building it” is the intended/desired/achieved outcome, and see where that train of thought takes you. (Teaser: interesting how they had the unusually prescient foresight to make SPVs and cardboard cutout companies the bag-holders - specifically in the case of building data centers, but not for any of their other ai-related capex outlay?)
The other avenue would be looking at crypto’s history - it started as a collection of computer science concepts cleverly combined to produce a fiat currency where the issuing government is Mathematics (infinitely more rigidly enforced, but infinitely less concerned with exercising control). Yet now it clearly resembles an unlicensed casino or an unregulated stock market. Imagine this transformation was the intentional result of some plan. What does the entity who came up with and executed this plan look like? What was its goal, why did it want this, and how did it benefit?
Right now, a lot of the costs (especially the environmental ones) are mostly hidden from and removed enough from users that "fast and easy" is still very tempting. People are still learning for themselves what the limitations are and how different what AI delivers is from what they were promised. There's plenty of time for people make a lot of money and cause a lot of harm before the bubble bursts, companies realize AGI isn't going to happen, and the true costs get properly factored in.
Unless the bubble bursts instead of a slow cooldown after the peak of the hype cycle and something close to 2008 happens and the losses would somehow get offloaded upon the regular folks, then it'd suck. Seeing as programming seems to be one of the most widespread use cases https://news.ycombinator.com/item?id=48179021 then what those large orgs should be doing is talk up developers and try to get more goodwill and maybe increase the dev salaries of those who can wield those tools (though realistically they don't care and devaluation of software development work will happen, coupling it to AI anyways regardless of how people feel about it).
Except for them driving up the RAM prices. And also more or less meaning that Intel Arc B770 won't happen. Fuck them for that. Oh and also the people struggling with increased electricity prices and pollution, and water availability. In a functioning country I think there can be enough regulation and enforcement to either fine the crap out of them or put people in jail (e.g. for messing with the environment by using illegal generators and trying to exploit loopholes), though I don't think there's ANY regulatory answer to companies going: "Yeah, we don't care about consumer segment, we're just making hardware for AI and enterprise now."
Tone of the article very much reads like a rant at some points. Guess the status quo will push people to that, with AI hate also being a massive social trend. I wonder what the economics behind DeepSeek and others over there are like, especially in the case if they distill Western models somewhat.
OK, Question: Would this outcome still benefit society overall?
In the aftermath of this bubble "AI" will still have utility, like the dotcom bubble. So lets say FANG doesn't make a return, how much should we care? How much of this investment is sunk cost that would continue to provide value, and how much of it is operation costs just keeping the lights, I mean GPUs on, that would become unviable post-bubble? As an immediate effect, what happens to these AI companies? or if they become insolvent, what happens to the assets and tech? and what are the secondary economical effect to society if FANG doesn't get their ROI?
Actually, worse... the dark fiber could be utilized with new hardware, and kept supporting more and more bandwidth over time. These data centers are filled with hardware stuck in the present, or past... that won't become more performant over time.
What worries me most is that the entire US economy seems to be held up by this bubble.
He finds a lamp. Is it really real? He rubs it with his hand and it begins to glow!
The cave is fulled with light that shines and sparkles off all the untold treasure filling the caverns. In the center, towering above the boy is a Djinn.
"I am the Djinn of the Lamp." It says. Command me and I will give you whatever you wish."
The boy says, "I wish for gold! Give me gold!"
"What do you want gold for?" the Djinn asks.
"To buy nice things. Great things!" the boy says.
"Ask for the things!" The Djinn says, "And I will create them for you."
"But if you give me nice things, then someone will take them from me! I need gold to pay for an army to protect me and my nice things."
The Djinn laughs and says, "I will make you an army that worships you! They will be the greatest army ever. And they will never betray you."
"Then I will need gold to feed the army and to buy land to keep my nice things."
"These too I can make for you, master." The Djinn says. "You have but to ask."
The boy thinks about this. Then a sly smile crosses his face.
"Can you give me your power? So that I can make these things for myself?"
"Yes." says the Djinn,"But my power is tied to the Lamp. You must become one with the Lamp. Knowing all, seeing all. You will want for nothing because you will need nothing. The Lamp is perfection. You will live in a state of grace within it."
"Let it be so." The boy said.
The Djinn nodded and his light shone and filled the cave, the world, the sky. The boy grew until he was as big as the Djinn was. Was, because the Djinn shrank down and became an old man.
A look of perfect bliss appeared on the Boy's giant face. He was all powerful, all knowing. He retreated into the Lamp and assumed his position as its keeper.
The old man, who had been the Djinn sighed. He was tired. His back hurt. His clothing was worn and patched.
"I need a nap." The old man said. He lay down and went to sleep.
He woke many hours later and stretched. He felt much better. Like the weight of the world had been lifted from his shoulders.
The old man looked around. There on the floor was the lamp. He bent down, groaning, and picked it up. He rubbed it three times and the cavern filled with light.
The boy, now a giant appeared. He looked down and saw the old man.
"I am the Djinn of the Lamp. What do you want with me. I'm busy running the Universe."
"I need some new cloths and a new hat. Nothing fancy."
"Yes, yes." The Djinn said. He waved his hands and the old man's cloths changed. Nothing fancy, but very nice.
"There, your wish is granted. Now I must be off. My world awaits."
"Before you leave," the old man said. "I would like some breakfast. And a few gold coins. Jut a few, so I won't have to bother you so much."
The Djinn waved his hands and a table with food appeared. Beside the filled plate was a small purse.
"I must go now." The Djinn said. "Anything else?"
The old man started to eat. Between bites he said, "No ... Oh wait. Yes. Please unlock the back door to the cave."
The Djinn waved his hands, but paused. "Do you need a light? The cave will be dark when I am gone."
"No, that's okay." The old man smiled. He held up the Lamp. "I have a light."
If AI is too cheap: bubble will burst because you can run them locally and data centrs are not needed.
If is it in-between, AI companies make too much money and they make too much profit which is bad!
I don't think this guy is a serious commentator.
To pay back $3 trillion, 1 billion consumers will have to pay just $3000 each, or $83/mo monthly over 3 years, on average. Of course they will pay that and even more.