I used to pay attention to Ed Zitron, exactly because he seemed to be someone who did the research, and looked at numbers. Until one day, I realized that seems to only look at numbers as long as it serves his agenda. When it doesn't, he looks away or makes the case for why the numbers are wrong, and shields himself from these "incorrect" numbers - or people who would point to anything suggesting that AI is not a total fad.
In April 2025, Zitron wrote [1] "I am sick and tired of everybody pretending that generative AI is the next big thing." This was at the time when AI coding tools had already gone "mainstream" with devs, and pretty much everyone was using tools like GH Copilot, Cursor, or something similar.
So I replied with this observation, saying how, at least within software engineers, AI is being adopted faster than any technology before [2]. To stay objective, obviously I brought receipts: sharing how, based on an older survey I ran, ~75% of devs in that survey said they used AI coding tools. Zitron made fun of the small sample size (216 people), blocked, and pretended like AI had zero PMF anywhere in the world.
I stopped taking him seriously since then. And I'm wondering ever since: does he deliberately only look at numbers and facts that he can tell the AI doomer story around? Or is it more that he finds that there's not many people who are "informed sceptics", and decided to play this role?
Ed's right about OpenAI being out over its skis, he's right about the numbers for everything being wildly optimistic and he's right about the circular financing/leverage aspect of a bubble. He hates AI though so he can't acknowledge its usefulness, it was pretty clear to see on his DOAC interview, Stephen kept bringing up real wins and clear progress, and Ed wouldn't engage in a substantive way.
Looking at it from an objective/historical perspective, when he first got on his high horse agents weren’t a thing, MCP was in its infancy at best, and half the code generated didn’t compile from frontier models.
From there everything split into two factions, which I’ll dub as believers and non-believers. From there, it has entirely been a cult following despite the evidence showing that models and agents have legs for software development.
FWIW my post history would show my extreme skepticism, and to an extent I still am. I think the real power of models isn’t the model at all but the harnesses.
Either way I think he lost the plot and runs on vibes himself. I also think this whole AI movement is going to have their 2000/2008 moment before the phoenix rises from the ashes.
I don't agree with the believers vs non-believers dichotomy. That makes this look like a religious war and it shouldn't have to be like that. To me AI is a very useful tool, no more no less, it isn't a 'make a wish' machine and it isn't a silver bullet for all of the issues that have plagued software so far. But when properly applied it's quite useful. "Non-believers" would have to be people that have yet to actually use AI, just like you can probably be a non-believer in peanuts until you've seen them, and most people that I know acknowledge the reality of AI tools and their use cases.
All tools, AI included, have a business end and as long as you point that away from yourself you will be fine. But to properly apply it (rather than as a faster way to make a huge mess) takes discipline and being methodical. It was never different.
AI enthusiasts have turned this into a religious war, like it or not. Far too many people here have tied their entire sense of self and worth to the success of AI, and will attack anyone who challenges their narrative that AGI is imminent. They refuse to accept the mere presence of critics, as you can see in this thread. It’s deeply irrational and unreasonable.
I think he deliberately looks at doomer facts. I'm not sure it's cynical so much as a genuine belief that it's all nonsense. But IMO the belief is based on a lack of understanding of AI.
> For any of his posts that I read, while there are numbers thrown around, the numbers don't actually connect to a coherent argument. In many cases, as we saw above, the numbers don't even really support his argument (such as an MAU decline in Facebook causing Meta financial problems which would then cause Meta to spuriously insert AI in places it doesn't belong). I suspect he's relying on people's eyes glazing over when they see numbers and just not thinking about what the numbers mean.
A frustrating thing about Ed Zitron is that he sometimes breaks news - gets hold of leaked numbers that nobody else has, for example - but he then wraps those numbers in his own extremely biased commentary. This makes it much harder to evaluate how credible the new information is.
The problem is that some numbers genuinely require a critical eye. If a source tells you that the OpenAI CFO told employees July ARR exceeded the Q2 total, there are a number of serious questions to be raised on how these numbers were calculated and what the point of such a confusing comparison is supposed to be. (I think the answer has to be that ChatGPT wrote the CFO’s script, no human financial expert would think to compare an annualized figure to the sum of three specific months.) But it’s hard to analyze the facts appropriately when they’re relayed by a guy who tells you that it’s all a giant scam and infers the worst possible answer to all the questions.
You're broadly right but I think you're dead wrong about the CFO statement — I think "July revenue alone exceeded all of Q2" is an extremely normal thing to say when revenue is going up at an insane rate. For example, if it was not true that June revenue exceeded March-May revenue, the point would be to show how their revenue growth has accelerated.
She, and we do not have the direct quote. It's a paraphrase, and the report can't seem to decide whether they're talking about revenue or growth. If she was talking about growth, "the ARR growth in July exceeded the ARR growth of Q2" is perfectly coherent.
Are you just speculating now? No source mentions growth.
The source you linked says: "Friar told staffers that annualized recurring revenue in July was higher than in the second quarter as a whole."
If we're charitable that means they annualized the quarter, i.e. multiplied by 4.
Which, if both measures are ARR, just means that a single month outperformed the average of three months, which is something that happens 50% of the time. Something you can opportunistically say whenever the coin flips the right way.
The less charitable reading is even worse, that one month's revenue times 12 is more than three months worth of revenue. Because duh.
The CNBC report I linked, which appears to be the primary source, mentions growth in the second paragraph. It is the main characterization of what Friar said.
“In an internal meeting with employees on Wednesday, finance chief Sarah Friar and board chair Bret Taylor touted OpenAI’s revenue growth and addressed competition with Anthropic, CNBC has learned.”
"July revenue alone exceeded all of Q2" would be a very normal thing to say, but you're skipping over some of the words. Unless revenue is sharply dropping, annualized recurring revenue in any month will exceed the revenue incurred in any quarter, because it's an annualized figure.
I think the most likely explanation is that the CFO misspoke and intended to say something more like your quote, or perhaps she spoke correctly and was misquoted. But without audited financial statements, all we have is speculation, and there have definitely been times in the past when executives of major companies made intentionally misleading statements about their revenue. (In fact, this thread began with a question about why you can't just look at the numbers, and here the answer is that nobody has reported the actual revenue numbers this comparison is based on.)
I'm not financially literate enough to trust my own analysis of the numbers. Ideally I'd like commentary from someone like Bloomberg's Matt Levine, a genuine expert in financial matters who is also extremely good at explaining them in terms non-finance-professionals like me can understand.
I didn’t even see the OP comment as a criticism of his analyses, just that he likes to do this rhetorical trick of dropping in numbers that are, at best, tangential to his point to prop up the argument’s credibility. Sometimes he does this in the middle of a real financial analysis which is even more maddening and confusing.
No, you can't. The numbers are genuinely confusing, and Zitron actively works to make them more confusing rather than explaining what they mean because he needs them to sound as bad as possible. He also buries the numbers in thousands of words of prose!
As someone else mentioned, ignore Ed’s personality and look solely at the balance sheets and capital analysis. Regardless of delivery, the math doesn’t math.
It's wild to me how much Ed is criticized while pathological liars like Musk and Altman are glossed over.
(Disclosure: I pay for Zitron's publication for the capital figures and exclusives he gets his hands on, I do not pay attention to his interviews, we are fact finding, if he wants to perform, I am not bothered, I've seen the performers on the other side, just keep the facts coming for ground truth)
I posted here a while ago when his bubble prediction lapsed and plenty of his fans were here. Telling me how I was wrong and the Q2 reports weren't complete yet because it was still early July and blah blah
As someone with a bit of understanding of accounts and finance I don't think Ed's analysis is very good. He's not someone who would pass a finance exam.
> For example, when Timothy B. Lee looked at a spreadsheet that Zitron used to create a projection of Anthropic's revenue, he found, "He doesn't count February 1-10, counts March 1-10 twice, counts August 21-October 21 as one month instead of two, and doesn't count October 21-November 1. [another commenter notes that his spreadsheet also contains February 30] ... Ed claims he tried to compute Anthropic's revenue for 2025 and came up with $3.6 billion, suggesting some funny business [but the numbers work out once you fix the errors]"
Yes, and his post about the exact same data was noticeably more confusing and less illuminating than theirs because he was primarily concerned to point to whatever the biggest number was and go "Ooh, big number!"
I don't find that frustrating about his writing-- on the contrary, his colorful (tinted with derision) use of language is very soothing to me, especially after encountering relentless LLM optimists like yourself on here every day, and dealing with people who have AI psychosis and are in positions of power over me, every week.
One of the best heuristics I've seen on whether there will be a fruitful outcome in taking someone seriously is how often they invoke labels (name calling, etc).
Even if Zitron is hopelessly wrong and a complete fool, he's an engaging writer. Like the original article itself states, it's not about the numbers, it's about catharsis. The transparent absence of AI tells in his articles is refreshing, and just rampaging against the AI boosters who lack empathy and self-awareness (which so many of them absolutely do) is cathartic. Even if it's all bunk, fake, and stupid, it's nice to have a way to self-soothe and manifest how much some of us want AI to have a reckoning.
Does that make his readers suckers? Possibly. But personally I also think it's a very human trait to seek comfort.
>He doesn't count February 1-10, counts March 1-10 twice, counts August 21-October 21 as one month instead of two, and doesn't count October 21-November 1.
Seriously, I expected more slight wrong before I read the article. Why do people trust someone who can't add numbers correctly to make predictions?
> gets hold of leaked numbers that nobody else has, for example - but he then wraps those numbers in his own extremely biased commentary.
You can think of yourself and draw your own conclusions (or we have completely lost the aptitude since "ai" started off?). You don't have to agree with Ed, or even follow his conclusions.
I still fail to see the problem here. If you disagree with numbers / conclusions he makes, I would be happy to hear/read your opinion on that (no sarcasm at all).
It makes me a bit concerned seeing how many people say oh he is this and that so I wouldn't even read or help him. For G'ds sake, what happened to healthy discussions? If you disagree, lets just share opinions and let people judge.
But if we just recuse ourselves from it, it will let opinions unchallenged.
Zitron has become the distorted reflection of the very AI boosters he criticizes and mocks.
I think the worst thing that happened to him was AI skepticism becoming a political position. This gave him a captive audience - as long as he says what they want to hear, which means that he can never ever concede that he might have been wrong or that AI might actually be progressing or having successes.
This is not conducive to good prediction long-term - rather it leads one to a state of cognitive dissonance where one's chosen enemies must be simultaneously terrifyingly powerful and incompetent dunces. The propagandist's disease.
He has built a following of people that want to hear his extra skeptical views. And even if he changed his mind about some things, he cannot admit it, as that is not what his followers want to hear from him.
It's rare that people make the news and build a following by saying a very balanced, down to earth opinions... unfortunately.
There are plenty of people on the far right whose positions are rational provided that you agree with their views. Conversely, there are plenty of people on the far left whose positions remain irrational despite that you might agree with their views. It is a serious cognitive error both to fail to see the irrationality of those you agree with as well as to fail to see the rationality of those you disagree with.
I almost wonder if it's even possible now that so many of us get our information through algorithmic feeds.
I listened to one Zitron interview on YouTube and my home page was immediately crammed full of similarly foamy-mouthed AI critics. (As well as plenty more Zitron.) And it's suggested some people with un-nuanced pro-AI takes, too. But what the algorithm has never, not even once, given me is a sober voice that calls attention to the nuances.
> And even if he changed his mind about some things, he cannot admit it, as that is not what his followers want to hear from him.
I don't think that's true - over the years he's changed his mind from "LLMs are useless slop machines" to "LLMs can be useful when applied wisely".
For instance in "The AI Hater's Manifesto" he says:
> It just came to me — the problem that I have with most people using LLMs is the delineation between outsourcing work and outsourcing thought. Those using LLMs to write little scripts or BQL code on a Bloomberg Terminal are inoffensive. [...] A tool being used as a tool to do tool things — in many cases involving the LLM writing a little 30-line Python script! — is not a problem, though it’s also not a trillion-dollar industry that needed to steal everybody’s art and writing.
Similarly, in "The More You Buy, The More You Lose":
> Sidenote: The only truly useful use case I’ve found is on the Bloomberg Terminal’s ASKB feature, which takes natural language and turns it into BQL code to make requests of Bloomberg’s datasets. It’s genuinely useful!
You can perhaps say that he underestimates what the technology is capable of (or, conversely, that other people overestimate LLMs) - but that's a different kind of conversation.
But even those concessions have been a shift: at every point, he's been as bearish as possible on LLMs, often past credibility.
And if you take it as a given that AI will never be any better than ~~now~~ a year ago, and that anyone who disagrees is an idiot or a liar, then that pretty much demands that the entire AI economy must be as fraudulent as he imagines. Which, while it serves his purpose of serving up AI-skeptic invective slop well, doesn't actually model what's going on, which is speculative investments that have the potential to generate extraordinary returns.
>It's rare that people make the news and build a following by saying a very balanced, down to earth opinions
Humans are built for fighting and killing the rival tribe, not for pondering whether the other tribe might actually be correct. We are exceptional for even being able to overcome ourselves enough that the outcome of "fighting and killing" can be minimized in the large.
At first I thought you were referring to Umberto Eco's laws of fascism: "Fascist societies rhetorically cast their enemies as at the same time too strong and too weak."
but it's subtly different, because incompetence is not weakness.
It's pretty clear in how people talk about billionaires: They must somehow conclude that the people most successful within a system that rewards certain things are not competent because the things the system rewards are not what they think the system should reward (it is fine to believe the system rewards the wrong things, but people walk straight into denying that those people have skill even within the system, or they claim that skill within the bounds of the system does not reflect any "real" skill - totally ignoring that elites tend to stay elites even through, say, Communist uprisings).
Hence how someone can say that Musk is "dumb" with a straight face.
> the system rewards are not what they think the system should reward
It's the grown-up version of nerdy kids hating "the jocks" in high school. We're all just a bunch of dumb kids. Maybe obsolete children, but still children when it counts.
That said, Elon Musk celebrated cutting funding that fed starving children and supported cancer research by waving around a chainsaw on stage. The richest man on earth did this because he wanted lower taxes... for himself.
When Anubis weighs his soul against the feather it's likely to completely destroy the scale.
I'm not saying any of the people in charge are good people at all. I'm just saying that they are skilled at achieving their own goals, regardless of the morality or lack thereof of those goals.
I don't even think that. I think half of them accidentally stumbled into success. Elon was kicked out of PayPal with a nice golden parachute so he wouldn't sue.
Regression to the mean is pretty nasty. Stumbling into success can make you a car dealership owner or something like that, but being consistently successful in business requires just about as much skill as being consistently successful in, say, chess.
To be clear, I'm just saying that I don't hate him because he's more skilled than I am. Not anymore than I hate olpymic athletes for being better at ice skating. He's certainly more skilled at Making Money.
I hate him because he is a cartoon villian, who when given enough wealth to feed the world chose to punch down instead of lifting up.
Things can look stupid when we don't understand them, but there has to be some intelligence somewhere at some point even if the arrival of wealth then suffocates it with sycophants. If Musk or Trump or GWB were all *merely* the idiots they're often mocked as being, chances are we'd never have even heard of them. They almost certainly weren't even "merely average".
Now, stupidity that arrives after the bank balance reads a billion dollars, that's terrifyingly powerful.
1. Trump was born wealthy. There was no "arrival of wealth" that smothered the intelligence you suppose he might have used to earn the wealth, because he didn't earn it.
2. You don't understand. Trump is successful because of his stupidity, selfishness, and vile behaviour. Not despite it. In 2016, the Republican primary field had 20 or so candidates on a wide range of reasonableness. Trump crushed them all, because he most represents the average Republican voter. If he were more intelligent, or if he cared more about anyone other than himself, they would not have voted for him. His arrogance, lack of intelligence and morals is quite literally his strength because it's what makes him so relateable to the American masses.
There are two possibilities. One is that half the country are total idiots and the people leading them, and leading the country, are also total idiots who only got into power because half the country are, again, complete idiots below average intelligence.
The other is that you are in a bubble and have been convinced that the leadership of the world’s premier superpower are below average intelligence, along with 70-80 million voters. The former is not something an intelligent person would believe, you have thrown out serious analysis for political theatre and memes and should rethink how you view the world. Now stop to consider that maybe actually half the country has a totally different worldview, and that their leadership, having reshaped global politics, is actually full of intelligent people who are cold and calculating. I know it’s harder to stomach, but just consider it for a moment.
Someone like Trump isn’t the first rise to the top of the country and be a con man or be as mad as a hatter in the end… He will not be remembered for his brain power.
I think he will be remembered for the decline and the loss of influence by America across the world that will be his legacy.
> below average intelligence, along with 70-80 million voters
Given what words mean, approximately 122.3 million US citizens eligible to vote ought to be below average.
When you say bubble, I'm inclined to agree, though for reasons that I suspect are uncooth to say out loud: we here are a bubble of smart people, so to us normal looks dumb.
There's rich, and they're enough money to rot your brain. I can believe Trump started with the latter, but the other two were less than that. IIRC GWB was "just" a multimillionaire even when he became president, about the same as Musk when Zip2 got sold.
Because it's not a single data point, intelligence is a wide surface area where you can be smart in some areas and stupid in others. Furthermore whether you are being intelligent is relative to ideals and goals, and not everyone has the same ones.
I find it difficult to imagine an intelligent person who finds it hard to get what they want. The trope of the oblivious genius who is incredibly competent in a single field of inquiry but completely unable to handle day-to-day life is more a creation of media than how reality works.
This isn't about people who just happened to make it big in a market rally, though. You're talking as if I'm speaking about people who just happened to buy some Bitcoin in 2012 rather than leaders of some of the most successful corporations on Earth.
They're not infallible, but modelling them as complete idiots that just happened to get lucky is also... Not really consistent with reality, as far as I can tell. These people might not be good people, but they're good at playing a certain kind of game.
Ed Zitron’s job is to convince people to pay him money to read what he writes, of course he’s going to preach to the choir, they’re paying him to do that and he knows it.
You're not wrong but for a lot of AI boosters it's also a political position - remember Anthropic and half of SV at this point is run by pie eyed effective altruists.
Ultimately one cannot separate technology or science from politics, it's inherently political
I'm not sure I have the force of will to not succumb to it. I think the only way to avoid it is to be so sure of the thing you are doing that the audience reaction is not the key metric by which you measure your success.
"Popular topic-expert" is a really cursed career to exist.
Ideally we would reward expert opinions on track-record, instead of how they make us feel. But we don't. Also ideally, the expert's conclusion wouldn't impact their ability to pay the rent (as long as it's correct), but again, that's not the world we live in.
> I think the worst thing that happened to him was AI skepticism becoming a political position.
It is a political issue in the sense that the discussion around AI can't avoid material and cultural issues. Ed Zitron definitely has an ax to grind and I'm personally turned off by his hyperbole, but I don't think something being a political issue means it's not worth discussing. I've also noticed the pro-AI crowd pivot to accusing people of not having actual reasons for hating AI, being victims of Chinese propaganda, etc as a way to avoid actually talking about specific issues that people have about AI and DCs.
I never said it's not worth discussing, I'm just saying that it becoming a charged issue was bad for Zitron specifically because his style of content fell in demand with the most irrational parts of the crowd.
How? AFAIK, his arguments based around the economics remains the same, he just seems increasingly dramatic and exacerbated, which is understandable when you realize industry + ecosystem (politics/reporting) is tulip mania delulu and insists 1+1=100, or 30 trillion, or whatever. His position isn't based on AI progress - it's based on AI economics, at this point he can be an obnoxious rationalist slamming flat earthers - just because he's annoying/smug doesn't mean he's less right on fundamentals which if anything is more clear now.
Are they? These companies have been caught tweaking their numbers. One example, not sure if cited by Zitron, or others, is that they build data centers through holding companies, who have to absorb the costs and massive capex based financial liabilites, so that the brand-name big-tech companies get to keep their expenses off their books. There have been trillions of debt discovered this way. Another issue is the apparently relentless progress of the hardware industry, needed to justify their super-high P/E ratios, measured against the fact, that to lessen the effect of HW amortization, hyperscalers opted to lengthen the depreciation timelines of their GPUs. So there is an apparent contradiction that new hardware needs to be both substantially better, and substantially the same, to make both stories true. I'm not a finance guy, and a lot of it is over my head, but even finance people keep asking the 'who's gonna pay for this' question. We're way past the belief that this is going to produce reasonable returns (as in a value for money kind of way), and hoping we can financially engineer ourselves out of this situation without having to feel the pain.
> These companies have been caught tweaking their numbers... they build data centers through holding companies, who have to absorb the costs and massive capex based financial liabilites, so that the brand-name big-tech companies get to keep their expenses off their books.
This is about as far from "tweaking their numbers" as you can get. It's a standard way infrastructure-heavy industries structure their investments and people would be asking questions if they didn't do this!
> hyperscalers opted to lengthen the depreciation timelines of their GPUs.
Yes and so they should! GPU depreciation timelines used to be 3 years!!
Google is famously still running 10 year old TPUs at 100% utilization, and 10 year old H100s are worth more now on the second hand market than they were when they were bought.
> This is about as far from "tweaking their numbers" as you can get. It's a standard way infrastructure-heavy industries structure their investments and people would be asking questions if they didn't do this!
Well, it's enough to throw off standard EBITDA accounting and allow firms to report fictional earnings numbers. A standard story has been that companies have beat their Q3 estimates, only for their stocks to go down.
To be fair to Ed, I’d describe his usual argument (at least currently) as saying that Meta, MS, google are “mature” companies trying to be maintain the high valuations and growth of a young company, which they no longer are.
If you take this to be his argument, then dan’s numbers are more consistent ed’s claim.
Of the three Google is in the best position. Meta and MS are in trouble. Zuckerberg will survive because he has control of his company, but Nadella is not going to survive Copilot if it don’t work.
When you predict a company is going to fail and instead it sets revenue records you're not "dramatic and exacerbated". You're refuted.
Luu's post includes a (long!) list of specific predictions that aren't "exacerbated"; they're simply wrong.
Luu's point isn't that AI is going to succeed or that the "AI bubble" will never pop. It's that these predictions are all wrong. If you agree "directionally" with Zitron, all that means is that you're skeptical of AI. That's a totally reasonable position to have, but it has nothing to do with whether Zitron's predictions are good or bad.
No, the whole thesis is XYZ likely fail because REVENUE RECORDS is not enough to dig out of hole relative to MAGNITUDE MORE SPEND. Saying Zitron wrong because XYZ made $2 for every $10 it spends revenue needed to justify spending. Fixtaing on the $1-$2 is misdirection/innumeracy, the thesis is in reaching the $10 relative to time, i.e. that $2 has to be $10 in X time, but the current velocity suggest it will not be.
I agree with Zitron directionally on accounting, I in fact disagree with him on AI... I am extremely AI pilled, i.e. I think there is a future where AI is worth trillions and will capture large swatch of economy. The transformation will be extreme, unlike any past revolutions... but the accounting suggest that future isn't coming in time to rescue current AI incumbents from finance blackhole, which some may survive, i.e. bail outs, nationalization... but the $$$ suggest however we get there, there will likely be massive $$$ corrections involved irrespective of adoption.
Did you read the article? It paints the picture of someone who does not in fact have a good track record on the 'fundamentals', even if his broad thesis of a bubble may prove correct.
e.g. when he accused Anthropic of fudging their revenue numbers, which amounted to him screwing up a spreadsheet
> you realize industry + ecosystem (politics/reporting) is tulip mania
It could be "tulip mania" or it could be "the internet".
Luu analyzed the numbers instead of just reacting to hype.
Specifically:
>> Oct 2024: OpenAI's forecast of $3.7B revenue in 2024 and $11.6B in 2025 and $100B in 2029 are absurd, "a statement so egregious that I am surprised it's not some kind of financial crime to say it out loud"
> Wrong (2025 goal exceeded, 2029 TBD but not an egregious financial crime level of implausible)
>> February 2025: Anthropic making $34.5B in revenue 2027 is "is laughable on many levels, chief of which is that OpenAI, which made around twice as much revenue as Anthropic did in 2024, barely made a billion dollars from API calls in the same year."
> Wrong (whether or not they make that in 2027, their 2026 ARR greatly exceeding that makes the 2027 estimate non-laughable)
On a spectrum between tulip wilting and fiber build out, we know deprecation cycle of DC hardware leans towards tulips i.e. <10 years (very generous) vs 20+ years for fiber layout, a lot of which is actually infra/earth works etc.
The numbers being cited is ~100B is well within accounting/ledger maxxxing tricks relative to current pool of investment. Luu is not analyzing number's he's just listing and believing numbers, and analytically entirely avoids the core Zitron thesis... once you tap out of easy investor $$$, FAANG warchest, accounting tricks... where is the rest of the order magnitude more $$$ that justifies existing spent relative to time frame coming from?
This narrow focus, of course some intermediaries in industrial chain is going to make $$$ selling/renting shovels - there is stupendous amount of $$$ being moved around, there will be some very phat winners, but even more losers in aggregate on broad ecosystem level. [1] is actually illustrative, there's a reason why opex low - capex premium is ridiculous right now, with almost everyone along compute industrial chain capturing 50%+ margins. Investors are burning $$$ and companies and pillaging warchests, intermediaries are raking in $$$, but that doesn't mean investors or companies doing all the spending will make more than they spend, i.e. the net ecosystem business model is not sustainable precisely because intermediaries are capturing crazy rent relative to actual monetization to sustain.
> but that doesn't mean investors or companies doing all the spending will make more than they spend, i.e. the net ecosystem business model is not sustainable precisely because intermediaries are capturing crazy rent relative to actual monetization to sustain.
You understand that this doesn't follow at all right?
The intermediaries margins can compress.
> opex low - capex premium is ridiculous right now
What does "capex premium" even mean?
Of course you spend more on capex when you build a data center than opex!
High capex matches the expected business model. If opex was high then everyone would be worried!
Investors exuberantly build $10 of housing when there is $5 of demand, builders extract $8, when they normally extract $2 under normal margins, builders raking it, but arrangement is net loses vs world where investors build same housing for $4 and make a profit. Intermediaries margins can compress but what they already extracted for current build out is already built in balance sheet.
>What does "capex premium" even mean?
>Of course you spend more on capex when you build a data center than opex!
No. Historically DC opex > capex, i.e. 60-80% goes towards power... because hardware costs were relative low % of TCO. Historically without delulu AI demand, IC producers capturing much less margin and TCO of DC was much lower than it is now. It's not opex vs capex it's TCO. AI is paying $10 vs $4, when demand is $5, $10 isn't sustainable, $4 is.
Now builders will be fine in case of crash, they'll compress margins for next round of buildouts, i.e. bubble bursts, current spend proves not sustainable. This is where the crux of argument is...
Future investors post crash when margins revert towards mean will be spending $4 to supply $5+ of demand. And due to nature of compute deprecatiion (i.e. tulips) they will have more efficient hardware with less opex/capex TCO per unit of compute, with much more sustainable balance sheet. The builders are still fine with their $2 margins, it sucks its not $8. But that leaves the current investors who spent $10 with stranded assets that are not competitive with more efficient $4 future build out, i.e. current investors have balance sheet black hole that cannot compete with none bubble market force.
This does not mean AI is doomed, it just means incumbents from current tranch of bubble driven, stupid high TCO build out is most likely doomed relative to future entrants. Unless incumbant has unassailable moat, or other hedge/cards (i.e. political bailout/intervention). That is the actual argument, Zitron is saying current ecosystem economics not sustainable, not that there is not a future model that isn't sustainable. But it does mean a lot of current players are balance sheet zombies, who _should_ die. But a reasonable disagreement is reality is size of bubble + contagion risk + influence of incumbents i.e. trillion dollar companies is such that they have non market lever (i.e. politics) to save themselves... but someone else is going to be doing the paying for a model that is net loss.
> On a spectrum between tulip wilting and fiber build out, we know deprecation cycle of DC hardware leans towards tulips i.e. <10 years (very generous) vs 20+ years for fiber layout, a lot of which is actually infra/earth works etc.
Counterargument: As advancements in transistor densities slow down, the rationale for increasing depreciation cycles makes more sense. As the performance gap between new & 5-year-old hardware continues to shrink, then the need to replace older hardware similarly shrinks, justifying longer depreciation cycles.
It's not just about node advancement, which is coupe de grace condition. Even if hardware advancement freezes, its about IC premium that fed current tranch of AI buildout. Current players paid $10 for a $2 hammer due to premium, a better future hammer might cost $3 but does twice the work of $2 hammer. That is like ball park the premiums we are talking about - from gpu to memory to other components getting inflated due to exuberate AI demand.
The economic logic is if current spend vs revenue gap is not sustainable... hardware prices / margins will revert towards mean. That $10 hammer will be compared against a $2 identical hammer (margin reversion/compression)... or worse, a $3 future hammer that does $4 / past $20 of work. The future player who only paid $2 can charge much less... i.e. simply paying $10 limits ability to price competitively. The future player who pays $3 has 50% more compute than incumbent who paid $10. The important DC TOC consideration, is in world where DC cost regress towards mean, opex > capex... so merely continuing to use that old $10 hammer is losing MORE than buying a $3 better hammer, i.e. the asset is economically stranded, it is COSTING MORE to run old hardware than simply buying new hardware. It's MORE than economically useless and $10 past purchase price not just sunk cost but dragging down balance sheet as amortized liability aka it is full write down / loss.
I'm not a huge fan of his style myself but fundamentally he isn't wrong. The whole "growth" so far has been all show and no go. I see people pouring in millions only to end up bankrupt a few months later. Evangelists portray those instances as rare and simply "skill issues" and "they don't know what they are doing but I am". Anyone that's ever had several servers running at home and seen the electricity bill at the end of the month knows it - I do as well. And we are talking about servers that use power, only a fraction of a server running 4x H100s at 100% 24/7. For those of us that have - we are talking servers that have one or two xeon silvers at best, 15% load on average. And I live in a country where the electricity is veeeeeeeeeeeery affordable. I'm not taking into account training, or investment to get it going, just inference.
20 bucks subscription a month, dario, you ain't fooling anyone with above-room-temperature IQ (Celsius). Even the expensive subscriptions - we know that's not the real price. And the limits, bumping up prices every two weeks and so on, just to keep the lights on while draining investors... Zitron's claims are pretty impossible to deny: the moment those companies go public, the real prices will need to come out of the shadows and end up on your monthly bill.
At the same time I fail to see the real world benefit to this, even in software: it's like comparing Lego (before they began their anti-consumer bs) to cheap Chinese toys. The moment you look at them side-by-side, you know which one is a premium product and which one is a cheap Chinese toy. While I have been a big supporter of open source since I was a child, I was never biblical about using open source - if some proprietary piece of software does a better job than the open source one - fine, take my money. Not anymore. I see the abysmal state of cyber-security as a direct consequence of slop. Slop-written code, slop-reviewed code, tests pass(also slop), ship it. Yeah, I'm not trusting you with my data, the hell with that, I'm self-hosting everything, adios. And I sure as hell don't trust the ai-bros for anything either.
And there's another thing: Microsoft was clear about it: github is constantly down because they can't handle the load. App stores are bumping up developer fees. Anyone can slop together a todo app(which was never hard to begin with). It's still a winner-takes-all economy - no one is going to install a todo app and migrate from Google/Apple just because. So the 1000 todo apps released daily will simply be a hole in the pocked of the people who slopped them together and another reason to be paying 500 bucks for 64 gigs of ram and another 500 bucks for a 2tb nvme. Raspberry pi's started off as educational platforms for 35 bucks and their commitment was to keep them at those prices. That's aged well. If I decide to upgrade my uconsole, I have to pull out another 350+ bucks for a raspberry pi. All thanks to the sloppification. The bubble can't pop soon enough and frankly I don't care what it takes down with it.
he makes a shit ton of money off this - in this attention economy we are doomed, people of every side find that going all in on something, no matter how intellectually dishonest, is much more profitable. Extreme AI bros or Extreme AI doomers seem dishonest. AI is an amazing technology - it is simultaneously not going to replace us in the next 3 years and it's not total garbage - the truth lies in the middle.
It is funny to see mainstream media finally catching up to how much of a grifter this guy really is. I believe I was the first at least on HN to make a list of his horrible predictions over the time [1]. Since then Kelsey Piper [2] also wrote about it and got some traction.
I knew that somewhere in the future, something has to give because you can't just go around saying anything without losing some credibility. You still have some ardent followers in his cult of a subreddit.
But just to be clear: Ed is part of a bigger problem in tech journalism which is characterised by extreme pessimism and excessive skepticism. It is not correct to view Ed in isolation rather to see it as a part of the culture in which he can thrive.
Zitron in general is representative of the conspiratorial thinking that has infected all spectra of the political space. Zitron obviously occupies more of the left space.
Good critique of LLMs & the companies behind them is hard to find, and it is harder when people gravitate towards this sort of thinking.
Upon further thought, I realized this should be counted as a correct prediction for Zitron! Zitron did indeed “keep writing this stuff until…proven wrong”. Zitron’s prediction says nothing about what happens after being proven wrong.
I've been doubting him myself (his recent articles just hedge on data centers more than anything else) but then I want to ask, are there any valid critics of AI? Not a "code is bad but it'll get better" but actual criticism in the nature of the financing, politics, etc. I think separating AI as a tool and it's capabilities and AI as product that needs to make a profit is helpful, however most people are really hedging in one camp or the other in their takes.
Cal Newport has been interesting on the topic. But it’s also possible to read Zitron and skip his personal opinions and just follow the discussion of financing, that’s what I personally do. I don’t understand why anyone would take the commentary of an internet pundit as a set of predictions to evaluate as gospel
I enjoy Newport myself, he's probably the most level-headed take I've come across. Everyone else has some agenda (or product) they're trying to get at and very much ruins the messaging (ex the agent 'civilization' piece is extremely overblown especially since..that's how multi agent systems coordinate already. Nothing happened that is unprecedented and isn't how the system is designed to work.
Newport does tend to be more level-headed, but honestly, the people that have gotten AI 'right' so far (in the true technical sense of how scaling laws held, etc) had their strong convictions set back in the mid-2010s, when Newport would've been far more skeptical. At least Newport's shown that he's capable of updating, but in general, he hasn't been someone I'd value for their predictive power.
The data center growth questions he raised have been where I found him interesting. I could never find any other articles to corroborate his predictions though. My question is when will the AI bubble burst?
I didn't know he'd had a long history of mispredicting AI quality and growth of companies like Microsoft and Google.
> I didn't know he'd had a long history of mispredicting AI quality and growth of companies like Microsoft and Google.
After watching this space for a long time, I think the growth angle is... untested.
All the FAANGs were slowing down after Covid. They boomed as the money printer went brrrr, then they plateaued. They've also kind of covered a lot of their potential their main total addressable markets, with the big exception of clouds, which probably have at least 5 or 10 years of growth as workloads are still being moved away from on-premise. Maybe ads, too, since TV is still around and big and there are probably a bunch of other holdouts I'm forgetting.
Then, FAANGS started reaccelerating around 2024.
Part of it was due to internal reforms, basically, layoffs shaking some things up.
But I suspect the bigger part has been AI, driven by CapEx and all sorts of other things. The problem with the AI growth is that it's highly opaque. We don't really know who the actual clients are. It is <<extremely>> likely that for Oracle (OCI), Microsoft (Azure), Google (GCP), Amazon (AWS) their direct customers are just OpenAI and Anthropic. That's it. It's likely that 70% of the AI growth is just 2 companies. That can't be healthy, it's also likely extremely risky.
Just the fact that the new cloud growth is so opaque is worrisome.
To shamelessly shill a passion project: A friend of mine and I try to be skeptical-but-reasonable on our podcast https://kairos.fm/muckraikers/ we aggregate papers and reporting and try to contextualize it with our own (hopefully useful) perspectives and takes
> however most people are really hedging in one camp or the other in their takes.
As with most things, most people are somewhere in the middle, as the silent majority. You only see the comments of the strongly aligned, which are also the most emotionally motivated to comment.
>I think separating AI as a tool and it's capabilities and AI as product that needs to make a profit is helpful, however most people are really hedging in one camp or the other in their takes.
There are a myriad of people out there who have brought up genuine issues that AI presents, or preexisting issues that AI is exacerbating.
This series of articles is perhaps the best long form critique of AI as both a technology and an industry I've encountered [1].
Like so many people posing as serious people these days, it's most accurate to think of Ed Zitron as an actor paid by his audience to tell them what they want to hear.
I haven't read every word but I don't recall him saying AI is bad or useless ever either. His refrain is that they cannot keep up with their endless spending on training and they're not reaching new markets.
I look at the rapid rise and decline of OpenClaw and get the impression AI has not really found any foothold among average people.
> I don't recall him saying AI is bad or useless ever either.
He does say that pretty often in interviews. That doesn’t change his thesis but that‘s one reasonable reason people dismiss him, he has often said that AI is useless when considering the externalities. And he will sometimes take a shortcut and just say „it’s useless, doesn’t do anything well“, which is of course way too simplified.
August 2024: "generative AI is a dead-end technology that has peaked”
July 2024: "Generative AI models aren’t getting more energy-efficient, nor are they getting more “powerful” in a way that would increase their functionality"
Nov 2025: "the fact we're running out of high quality training data and we're hitting the walls of scaling laws, in the training paradigm, these models aren't getting better. What we're seeing today is pretty much what they're always gonna be like"
There are other examples too from his prose of talking about how they are barely useful but I don't want to dig it up
I don't know about Chegg, but the fundamental problem with Duolingo is their product does nothing to teach people a foreign language. The nasty reality of that product is if you go through their entire learning tree in a language, you might be at a CEFR level A1 for that language. And, you spent 10x the time and 10x the money you would have spent via something like Lingoda to get the same outcome.
This claim is just not true. And gets tiresome. You roughly end up where those courses claim to be in reading and listening - depending on language it can be over B1. (No course finishes B2, some do have B2 content). I ended up being able to watch some (not all) netflix series in foreign langue and I was in early B1 section. I clearly learned.
I speak 4 foreign languages, at (fully tested through exams like the DALF, DELE, Goethe Institute) levels of C2, C1, C1 and B2. Take an actual test proving your knowledge and get back to me.
I've spent a lot of time with Duolingo learners. They're all pre-A1. You're just lying to yourself to make yourself think your phone addiction isn't "that bad." Take a real language class.
It is simple. I do not want to and have absolutely no reason to take the language class. So, why would I? They costs money, time and are not something to look forward to. I got actual visible measurable results. Why would I changed?
Besides, language classes also fail. Going to language classes for years and not being able to use the language in any practical situation was always fairly common result.
> I've spent a lot of time with Duolingo learners. They're all pre-A1.
I know people who tested B2 after doing nothing but Duolingo German. And like I said, I can watch Netflix shows and read books in language I used Duolingo for. That is definitely something.
Hate on him all you want (he's gotten very repetitive for the sake of subscribers and reads), but the basic premise that the modern AI industry is a circular-dealing, point-of-diminishing-returns grift still holds. The bubble is here and the longer it inflates, the worse the pop will be. Sure the goalposts have moved, but the basic numbers don't math.
The current cross-deals aren't purely circular; quite a bit of revenue is, in fact, flowing into the AI industry from the outside (you know, via customers). What's more notable is the shared risk, as the deals tie multiple companies across the chain to a set of shared bets.
If the frontier labs suddenly found themselves unable to compete with cheap open weight models running on widely available compute, then one might expect the frontier labs to be the only ones exposed to that risk, while a chip-manufacturer like nvidia could thrive in either environment. A partnership or commitment from the chip manufacturers to the frontier labs could change that. Whether that kind of inescapable connection exists is hard to predict without a lot of specific modeling, and at this point I'm inclined to think that neither chip nor datacenter demand is going to drop any time soon, and that the commitments would be unwound before a company like nvidia is threatened.
Posting revenue numbers as an example of a company not dying seems rather foolish. Microsoft's gaming side is floundering and dying and Linux has been growing at an unprecedented rate as a result of Microsoft's decisions. Not to mention the geopolitical aspects at play here as countries make a serious consideration to drop Windows.
These companies are all heavily buoyed by their investments in AI which is essentially an oroboros of money.
If we sample from Zitron’s claims, a lot of them are wrong. Probably most of them.
I think his contribution to the discourse is the bigger picture. These companies are going into potentially economy-wrecking debt over a speculative future that looks much foggier than the last two previous technological booms.
i relate to zitron cuz of how much he seems to genuinely hate the people in charge of this AI bubble. he, like me, seems to wish that when this all falls apart they get proportionate harm to go.
that'll obviously never happen. it's wishful thinking and venting tbh. that's why i like listening to his stuff.
it's better than reading the wholly AI slop docs my CEO keeps sending out
Only recently came across this guy and if u ignore the standard social media hyper hype hype he has, I dont think he is wrong even his 2024 predictions.
US has 1.5x the money supply since Covid. This means everything has to go up 1.5x to reach the same parity from before. eg inflation. All corporate earnings are going to inflate as 1b yesterday is 1.5b today.
There is no whete for these mega corps to go. They dont know how to grow. meta becoming well meta with the vr crap is case and point. meta blowing billions on a few hyped AI people is the nail.
Other than cloud, microsft/google aimt doing anything. xbox? nope, hardware? nah. It is all attention economy or selling the pick axes for it.
the only mega corp that still seems to be moving the needle is apple which i think speaks volumes.
Eds point is that AI tam or expectations are insane full stop and he admits coding has a use just not as big of a tam. Gen ai in art, music, etc did not take off at all esp compared to code. this makes sense, no one pays an artist 500k/year. their time is worth little compared to coding.
Where he is wrong tho is that this is not new. there is always a hype cycle of nonsense that screws the little people. what is new is the level of polarization. it feels like ai psychosis right now in ways that remind me of crypto but worse because ai is actually useful so people are even more insufferable.
I think the most enjoyable part of this is to have written it in the characteristic long-form Zitron wall of text style while still managing to pack almost the whole text with meaning, which is entirely the opposite of Zitron style.
As always, shooting down commentary asking for more and more evidence is very simple.
Zitron has a somewhat ranting style. Luu's passage on Google search for example just nitpicks on the person that Zitron allegedly named as responsible for the decline of Google search.
The real issue of course is that search quality clearly has gone down. What does Luu want? A "study"? Everyone can see that. The person is not the issue.
This is a highly selective analysis of Zitron's writings that focuses on a couple of mistakes. People need to understand that any normie will understand the difference between rants, numbers and speculation in Zitron's essays. They are not written for autists who talk of "priors" and "evidence".
"This is a highly selective analysis of Zitron's writings that focuses on a couple of mistakes"
While reading this I was thinking it would be interesting to see just a few examples of Zitron predictions that he got right. Since you appear to follow his work, do you know of any?
The issue is the conspiracy. By pinning the whole problem on a single villain rather than the incentives the larger organization has created he 1. doesnt help solve the problem 2. creates an environment ripe for tribalism. Even if you ignore the racist undertones(fine its a stretch) he's still misleading people about what is happening.
Being thorough and accurate might make you a lot of money in the stock market, but it's not a good way to get any media presence, because that demands being in front of people quite often, and basically nobody can be insightful and well researched in all topics of the day. Once you become a pundit, whether on politics or tech, and rely on eyeballs to feed you, you are going to be throwing stinkers. And at that point, you might as well just align with an audience and not care too much about whether you are predicting anything accurately.
We have people telling us the next recession being imminent all the time. Others told us that NFTs were key to the future of art. That's the pundit way. If you want accuracy, you become like Charlie Munger, have maybe 2, 3 really good insights througout your career, and hope do to well using the insights yourself, not by being a pundit
This is exactly right. Now, that does mean something about Zitron: he’s a pundit, not an expert or a forecasting genius. You could point to any number of analogous booster types. Twitter/X somehow loves pushing these people onto my recommended feed. I remember reading breathless threads about how o3 was going to single-handedly end white collar work. There are just more of that sort of person, so none of them in particular gets the same amount of attention as Zitron, who seems to be the only person willing to go on the record against AI. Maybe his overall worldview is sound, and maybe it isn’t. But he’s not really making confident, specific predictions about the future.
Which leads me to a criticism of the piece: several assertions are described as “Wrong,” with no explanation or citation, which are not obviously wrong to my mind. For example, the assertion that “DeepSeek has commoditized the [LLM]” is at least debatable. It is a prospect that the major labs seem to have at least considered.
In my experience, people who make fun of AI's ability to eliminate white-collar work are exactly the same people who are white-knuckling it, hoping they can make another $300k next year to buy that Rivian.
People who don't need a salary look at the situation more objectively and, in general, can see that a lot of white-collar work is in peril.
I think that is tied to the general trend that people see themselves as far less replaceable than others. And I’m not sure that people who don’t need a salary are exactly objective about it - unless you mean people who live off the state or something, those are people who need their capital reserves to continue to appreciate in order to sustain their lifestyles, and are thus relying on the destruction of white collar work to justify the valuations of the technology companies which are underpinning the strength of the markets. It’s like saying in the 19th century that only the propertied classes should be allowed to make decisions about labor policy, because the laborers themselves have their judgment clouded by their position.
But I take your point, and I only brought up the o3 example to emphasize that people on both sides of the booster/doubter debate can be prisoners of the moment. There are boosters eternally convinced that the utopia (or armageddon, for the doomer-inclined) is either already here or imminent, and there are doubters eternally convinced that we have reached the peak. One of them will eventually be right, but neither has been yet. You can’t fault one of them for calling their shot if you’re not willing to see it happening on the other side.
The flip side of that is that at least half of all white-collar work was already useless before AI. If it hasn't already been eliminated, there's no reason to assume a-priori that AI can eliminate it.
Not completely. It just means you can have an AI do the useless busywork instead of a human. Something can be economically useful but still have proponents in a firm.
Call your project Gas Town, write a blog about it, and investors will be swarming all over it thinking managing agents is even more impressive than managing people.
>In my experience, people who make fun of AI's ability to eliminate white-collar work are exactly the same people who are white-knuckling it, hoping they can make another $300k next year to buy that Rivian.
In my experience, people who believe that some white collar industry can easily be automated with AI usually don't work that same role or even that industry, and suffer from Dunning-Kruger bias. In my line of work, I'm desperate for Claude to actually do a better job at not producing a big mess, and it's failing terribly.
I don't work in law, but I feel like I barely need a lawyer if I can ask an LLM to interpret a contract for me. I don't work as a physician, but why can't I just cut and paste an MRI report and ask Claude what to do. I'm no plumber, but it I take a picture of some fixtures I am thinking about replacing, Claude will give me some items to add to the shopping cart from a plumbing supply store.
In all cases Claude was super confident and it wrote what felt really rational.
But in my experience it gets things right, in some amazing ways, but it gets things wrong, in shocking ways.
Everyone thinks it's someone else's job that it'll automate.
> We have people telling us the next recession being imminent all the time. Others told us that NFTs were key to the future of art.
That's quite the conflation, and a very subtle way to tie a claim that's still unresolved to one that is to try and make the first one seem wrong with no evidence. Recessions are almost guaranteed in our system, they're a part of a cycle. Unless you think that the Great Recession was the last one in history, saying the next one is 'imminent' would be correct for any other viewpoint. We just don't know when it'll officially start.
...get any media presence, because that demands being in front of people quite often, and basically nobody can be insightful and well researched in all topics of the day
What? Zitron is getting a lot of media attention by repeating the same takes hundreds of times. He's not a general pundit.
I just remembered that before AI, I knew Zitron from ranting about IoT and consumer tech. He just appeared one day in my Twitter Timeline, claiming to be the "man on the ground" at CES (or so my memory at least)
So from that standpoint, I guess it would make sense that a personality like that would now have ended up at the AI topic.
I would like to see a comparison with predictions of the pro-AI bubble and their hit rates though. This writing feels as selective as it tries to paint Zitron as.
All of the grifters (and lots of capital) can't wait to jump on the bandwagon because they see it as their chance to clean up. The same happened during the last 4 tech revolutions. (in my life so far: semiconductors, the internet, mobile communications, smart phones, crypto, EVs, probably forgot a few). And of course, as always, the longer term will be more amazing than the short term hype.
> Given how fast these companies are growing (in terms of revenue and profit), it doesn't seem that AI is, as Zitron implied, some kind of desperation move they're reaching for because "they don't know how to grow" and are all out of ideas
Growth isn't a valid rebuttal, unless we can also sus out how much of that growth is tied up in circular financing of AI projects. We have a pretty good idea how much of Nvidia's valuation is tied up in the AI craze, it's a bit harder to tell with the megascalers....
Can you provide any examples of any of the megascalers publishing any detailed financials that touch on their AI spend or revenue or profit? The only one I’m aware of that comes close is Microsoft and they have still buried it in barely related line items which still leave us making assumptions.
There’s speculation on both sides and certainly Zitron is on the extreme end of the anti-AI side with the most cynical speculation but it is indisputable that none of the megascalers are open about their AI financials. Hence, we are all speculating endlessly. If only there were published financials then the speculation could end!
The obfuscation of financials doesn’t necessarily mean something bad is happening, it could be a competitive advantage for Google to be secretive about how cost effective their TPUs are or for Microsoft to hide how much revenue uplift they’ve experienced by adding AI to 365.
They don't publish their AI spending, revenue, and profit - but they do tend to break out other non-AI segments of their companies which can demonstrate that at least part of their growth isn't relevant to AI.
> They don't publish their AI spending, revenue, and profit - but they do tend to break out other non-AI segments of their companies which can demonstrate that at least part of their growth isn't relevant to AI.
So the previous statement that "As public companies, the megascalers publish pretty detailed financial reports" is incorrect and irrelevant to the question that was asked.
Financial reports alone don’t paint the whole picture when it comes to valuations. For example, theoretically amazon has committed to invest 25 billion in anthropic, and anthropic has committed to spend 100 billion on aws compute. As far as we can tell, no real money has actually changed hands in either direction, but both valuations are being buoyed by their prospective investments…
Because of accounting tricks, quarterly financials don't accurately reflect the size of this fiery money pit.
The datacenter build-outs are all majority (>50% ownership) financed by other companies, with a shell company owned by the hyperscaler as a minority owner. The data center then grants the hyperscaler an exclusive leasing agreement, and because the shell company is a minority owner, legally, it's not their debt.
The only reason this has worked is because there's such a long delay taking delivery on GPUs. When these capital allocators start paying for GPUs for data centers which haven't yet broken ground, then we'll see a very visceral market reaction. Yes, some of this has happened already, but there's enough momentum that it can be absorbed and dismissed as an anomaly. But governments are unexpectedly passing moratoriums on data centers everywhere, so it's only a matter of time before there's no data center to offload those GPUs to. That's when the music stops.
I believe that was Zitron's central thesis and why he started reporting on this. This mirrors the mortgage-backed securities situation that led to the 2008 GFC.
The thing about $1.65T being "hidden from the balance sheets" is overblown. How do we know about that money if it's hidden? Because it isn't - it's in other public SEC documents, which the $1.65T claim also slightly misinterprets.
A startup raises $50 million from OpenAI and Anthropic to finance API calls to OpenAI and Anthropic that they are using at a loss who in turn spend that money on compute with Microsoft and Google who in turn invest in Anthropic and OpenAI who then invest the startup using the startup’s revenue to value it… the cycle repeats.
There are multi-billion dollar valued startups invested in by OpenAI and Anthropic with hundreds of millions in ARR that are spending 90% of their revenue with Anthropic and OpenAI.
Situational Awareness, the fund that recently imploded, invested tens of billions into AI companies using their holdings in Anthropic to help finance the investments…
This could all work out fine in the long term, we’re all just speculating at this point, but the circular financing is absolutely making it to revenue because capital invested into startups is used to fund growth which is achieved by subsidizing costs.
The vast majority of startups are not funded by OpenAI or Anthropic. They are not a significant source of venture capital. Meanwhile, OpenAI is pulling in $40B+ per year and Anthropic $65B+ per year.
You are mixing up valuations with liquid cash and you're also making sweeping statements about how those startups are spending their cash. A majority of a raise is not spent on AI compute.
Situational Awareness blew up because they used leverage to invest, and leverage is a great way to blow up any fund even if they were directionally correct about AI.
You’re applying pre-AI investing to a post-AI world. Yes, a decade ago, a startup raised money and spent 90% of it on people. The people built software which had incredible margins. Build it and then print money for ever more. That’s not the case any more, these startups no longer have incredible margins, they’re not collecting $100/m per user and banking $99 of it. They’re collecting $1000 and sending $999 of it to Anthropic and OpenAI.
Revenue numbers are vastly inflated compared to pre-AI but these startups aren’t keeping the money. Profits are worse than ever before. Startups with 30 employees that reach $100m ARR in 6 months are not banking $90m or $80m or… they’re just passing that money straight through to OpenAI and Anthropic.
If startups aren’t just funnelling all their funds raised straight through to OpenAI and Anthropic, where is this combined $100bn in revenue coming from? Who is paying for it? My spend on software certainly hasn’t gone up in a post-AI world. My company is spending less on software now.
OpenAI have stopped being so reckless with their cash investments which is why they appear to have slowed down but they’re still investing millions in huge numbers of startups through token allowances. They invest $2 million in every YC startup (or did a few months ago). There’s an entire market of reselling these things tokens!
This is so completely wrong and deluded I’m not sure where to start.
I work with AI startups and scale ups on a regular basis as well as plenty of more old school companies, all of whom are spending money on AI models, because they are getting insane value from them.
This idea of the revenue for OAI and Anthropic coming from “circular financing” is just bizarre wishful thinking coming from AI doomers with zero financial literacy.
The revenue numbers reported by AI companies (not just OAI and Anthropic) isn’t being driven by Nvidia at all, in fact, the numbers wouldn’t add up if you thought that was the case. The revenue being brought in by AI companies is far, far higher than the sum of any investments from Nvidia.
The AI doomers just can’t handle the idea that AI is actually incredibly valuable and every company is using it and increasing their use of it every month.
And yes, I see this every day in my job and with every company I work with.
I also contract in the startup space, and many of these startups have pretty much their entire infra bill covered by AWS/Azure/GCP credits, and all of their AI spend covered by Anthropic/OpenAI credits.
Theoretically they'll spend real money on those things down the line, assuming they find product-market fit, but who knows how many of the current crop of startups will reach that point
Wow, thanks for your perspective, it’s lucky to find someone on Hacker News who works with technology every day!
You presume to know my position but you do not. AI is an innovative new technology that is radically changing how we build and use technology and will continue to do so. That doesn’t mean that trillions of dollars is going to be spent on it. Despite the penetration all technology has in our lives, most companies are barely using technology from 20 years ago because implementation is a nightmare. Businesses are risk and cost averse, better the line item you know. And so, most companies could be radically improved not by human-level intelligence, or even dog level intelligence, most companies just need macros that are easy to implement. Most companies could 10x their productivity without AI! After all that’s what startups have been doing for the 20 years pre-AI, that’s been the YC investment thesis (which has worked very well).
My position is that AI is a radical step forward in technology that pragmatic businesses will benefit from handsomely by using cost effective models. A middle of the road local model that can trigger tools is more than most companies need. The frontier models by the frontier labs are a complete waste of money outside of the most extreme edge cases.
Conflating “the technology is incredible” with “companies will spend trillions per year on the technology” is ridiculous. Your argument about usage says absolutely nothing about the financials yet you’re dismissing the AI “doomers” (people who are pessimistic about the financials, not the technology) on that basis.
If you look at what we know of the financials of OpenAI and Anthropic it is impossible to come up with a financial case to justify the trillions of dollars in revenue needed for the AI booster’s vision of the future.
How much money does The JavaScript Company make? How much money did Docker make? It’s like the AI boosters who argue for the financial case have forgotten the last 20 years. The world of technology is built on open source, it’s built on companies that made a huge impact and failed financially. Docker led the way with containerization, one of the most influential technologies of the last 20 years, and the company almost went under multiple times. We constantly gripe about how unsustainable open source is. Why is all this suddenly different? Why is making an innovative new technology suddenly guaranteeing trillions in revenue?
How many trillions of dollars were invested in data centres to build Docker containers?
If you think I lack financial literacy, please explain where the money is going to come from. Please make the financial case for trillions of dollars being spent on AI over the next few years. Keep in mind that the reason technology has been so profitable over the last 20 years is because of the margins, software is basically free money. AI is not free money. AI is very expensive money. Also keep in mind that the current (rumored) revenue of Anthropic is primarily made up of the most expensive use case (generating millions of lines of code) being paid for by rich tech companies which does not represent the wider economy.
Bets are meaningless but feel free to stake a claim here to how you think things will be 4 years from now. I’ll stake my claim: AI will be more impactful than ever while Anthropic + OpenAI will have less revenue than today.
It's coming from regular companies spending their own money on using AI. I.e. Profitable companies deciding they want to use AI for various reasons and spending their own revenue on said AI, whether it be Claude Cowork, OpenAI ChatGPT Work, OpenRouter, Nebius Tokenfactory, models hosted on BaseTen, Fireworks, etc, tools like Lovable. Or every piece of cyber software which are ALL using AI heavily these days.
It's really not as complicated as AI-doomers like to make out, they seem so confused somehow that existing profitable, successful companies are spending larger and larger amounts of their revenue on AI. It's not circular by any definition.
So anyway, this whole worry about "where the money comes from", is kind of funny. Where does the money come from to hire employees? Where does the money come from to pay for Cloud bills? The money for AI will come from the same place, it's not some big mystery. Total cloud/compute spend in the world is well over 5T per year, including all cloud and colo spend.
AI is basically both taking up software spend, dev salary spend, white collar officer worker spend, hardware spend, general IT spend, etc. And if you sum up all the budget associated with all company software, personnel, white collar workers, etc, you end up with a much bigger number (probably 10-20T or more most likely).
So the idea of total AI revenue being in the trillions, it's pretty simple, and will happen over the next couple of years, just like happened with regular servers and cloud.
People like yourself downplaying AI reminds me of people both downplaying the internet ("it will never make money") and also the original launch of cloud providers (AWS originally), "no one will ever trust the cloud not to lose your data, why would any pay for AWS" etc etc).
Both sets of folk were radically wrong, and the AI-doomers will be wrong this time too. Capabilities will increase across the board, amazing applications will be built (already happening), and people will want to pay for these products. People are ALREADY paying huge amounts of money for these products.
Who do you think is paying for Lovable? They are probably the fastest growing startup ever from 0 to 1B valuation because less technical people LOVE using it and have zero issue paying for it. But AI doomers will somehow dismiss Lovable as somehow getting "circular financing", when loads of small business owners I know love the product and spend 100s of dollars a month on it!
It's going to be funny watching the doomers over the next two years when none of the big AI companies goes bankrupt and keep growing in revenue. But but but the circular revenue!
> existing profitable, successful companies are spending larger and larger amounts of their revenue on AI
Apart from the brief "tokenmaxxing" craze among the big tech firms a while back, is there any evidence that profitable companies are cutting their own margins in order to spend on AI?
Lovable's success is the perfect example. Lovable has a large number of users who do not pay for the platform who have been subsidized by investors while inference costs were high. Lovable know that long term, sending all their revenue to Anthropic and OpenAI and Google is very bad for business, especially if that revenue is subsidized by investors, which is why they have trained their own model. Lovable's long term success is in conflict with OpenAI and Anthropic! Lovable succeeds when it stops sending $0.50 of every $1.00 to OpenAI and Anthropic and Google, Lovable succeeds when it drives down the costs of inference to as little as possible.
Regarding the cloud infrastructure comparison, it is not at all comparable. During the time I spend writing this comment, my device will make thousands of requests and connections to different servers for all sorts of reasons. During the time I spend writing this comment, my device has interacted with an LLM exactly zero times. The throughput of internet infrastructure is not even in the same universe as the throughput of LLMs at their most wildly successful. How many times does Lovable's AI run per month for their average customer? A few times? The repeated, continued value Lovable delivers to their customers is in the interactions that occur between their customer's customers and their customer's apps. A Lovable customer can love Lovable and have huge success with their Lovable app while using zero tokens per month.
You should be comparing AI to a product that eventually became commoditized, not comparing it to an entire category, e.g: shared website hosting. Shared website hosting was very expensive to set up 30 years ago. Over time, it got cheaper and cheaper, now today it is commoditized, the major brands have all consolidated, companies have gone under, and technological innovations have completely reshaped how websites are hosted. Who still uses shared website hosting today? Websites are bigger than ever, web servers underpin the economy, Stripe alone has web servers that process trillions of dollars... how much money is there in web servers?
> Capabilities will increase across the board, amazing applications will be built (already happening), and people will want to pay for these products. People are ALREADY paying huge amounts of money for these products.
You're so caught up in the technology that you're oblivious to the economic reality. The capabilities, the amazingness, the excitement, that isn't how money is made. The most cheap and boring technology (like web servers) are fundamental to our economy. AI can be all of these things, it can have incredible capabilities and be amazing and have so much excitement and radically reshape our economy and be fundamental to every business... and make no money.
You, like so many nerds, cannot seem to separate technology from business. Business is boring and simple and based on principles that have stood the test of time. Business doesn't run on excitement, it runs on numbers. Shopify powers most ecommerce, Shopify is one of the most important companies in ecommerce, Shopify is wildly successful, Shopify's revenue... $12bn. Stripe's revenue, on trillions of dollars in payments... less than $10bn. Shopify and Stripe are wildly successful and very important companies that are involved in trillions of dollars flowing through the economy and you're suggesting that AI is going to do 100x more revenue than them?
Tailwind CSS is used on probably half of all major websites today. The creators of Tailwind recently announced they had to lay off everyone because the company was struggling to make any money despite usage growing every single day. WordPress, which (supposedly) powers half of all websites is operated by a company that is struggling too. Google and Meta, some of the most profitable companies in the world, almost all of their revenue is still from advertising that has barely changed in 25 years. Google make hundreds of billions of dollars from... showing...
$800M a year is less than 1% of their quoted revenues at $100B+ / year. Claiming credits as revenue would be tax fraud. Credits to clients for services are counted as debits against
There are zero serious companies collecting $1000 on revenue and sending $999 as a cost of goods sold to Anthropic/AI. It would be unprofitable to even run a proxy to Anthropic on such thin margins. But I digress.
No company was banking $100 and keeping $99 in the "before times" either. These are fantasy numbers not even the most highly optimized software company produced. As an example, Slack famously went public in 2019 and it had revenue of $401M with a gross margin of ~79%, meaning they were pulling in $316M in gross profit. That is the figure before labor, administration, R&D, sales & marketing, etc. They actually operated on a net loss after factoring for those expenses, despite their high gross margin, which is common in high growth startups (Amazon famously ran losses or marginal profits until decades after their founding because they continuously reinvested in expansion).
Credits reduce revenue by all basic accounting standards. You can accuse these companies of fraud, it is within the realm of possibility, but it would also be <1% of their total quoted revenue, so not really worth the heat at the same time.
You are making conflicting arguments at the same time. There exist startups that are able to generate gross profit with some consumption of AI services, they are also able to invest nearly 100% of their capital into AI to generate those profits without needing to spend on traditional labor, and yet AI is not sustainable. By your own circular logic it is of course sustainable, but by grounded logic, you have to understand any business that goes from zero 4 years ago to $100B+ in annual revenue today with double digit growth rates is offering the world something of value. Anyone who has tried AI sees some value in it. There is some revenue and profit to be made here. Betting against that in the long term will just lose you money and sanity.
> No company was banking $100 and keeping $99 in the "before times" either.
Yes they were and are. The marginal cost of software as a service or data as a service is near zero. Slack is a good example. A new Slack customer costs Slack nothing. Free money! Slack had a high valuation because of the margins. Slack and other traditional high-growth technology companies were valued highly despite being loss making because there was an understanding that paying for growth early returns a lot more later on. Slack (pre-acquisition) could turn off their expensive growth engine and start making money hand over fist.
(Look at what Bending Spoons are doing now, they're picking up "zombie" technology companies that have incredible margins but no growth. Bending Spoons are cutting these companies to the bone, giving up on growth, running on a skeleton staff, and making money hand over fist, cashing out on the incredible margins of software.)
Someone shared up thread an example of Harvey, a legal AI company, who regularly post about their token consumption. They're consuming trillions of tokens per month for their product. Harvey's investors include OpenAI. Harvey has raised more than $1bn and is currently valued at $11bn (and raising again at $15bn apparently). As of last month, Harvey's revenue was reported to be $30m/month on 13 trillion tokens per month.
Let's be conservative and assume their average spend per million tokens with OpenAI is $2. That's $26 million in token spend per month, on $30 million per month revenue. $2 is lowballing it of course, they're surely using one of the frontier models. That's pretty close to every dollar coming in going straight out to OpenAI. Considering the capital they're raising and burning (seems like $50m a month) while relatively small (<1k employees) I would guess they're spending at least double their revenue with OpenAI.
Of course, long term, this is fine for Harvey, as model costs come down and businesses mature they are going to be spending a lot less. Maybe they'll start running their own hardware, offloading certain workloads to cheap models, using scripts for routine tasks where AI is overkill. Great for Harvey and Harvey's investors, an absolute disaster for OpenAI.
> By your own circular logic it is of course sustainable, but by grounded logic, you have to understand any business that goes from zero 4 years ago to $100B+ in annual revenue today with double digit growth rates is offering the world something of value. Anyone who has tried AI sees some value in it. There is some revenue and profit to be made here. Betting against that in the long term will just lose you money and sanity.
You're making a leap from "useful" to "profitable". Yes, there is absolutely revenue and profit to be made for companies building products, for the companies providing technology, not for the companies providing inference. There are not software margins in inference, it's a commodity, the only reason OpenAI and Anthropic went "from zero 4 years ago to $100B+ in annual revenue" is because nobody cares about the money today.
Right now, we're in a gold rush, we're in the growth-engine phase, we're in the "spend a billion to make a million as long as you're growing" phase. Right now, people at Harvey aren't worried that every dollar in is at least a dollar out to OpenAI, who cares, investors are funding it, they're growing, they're taking over the legal world, that's all that matters, they can balance the books later... and when they do start to balance the books, when they convert that growth-at-all-costs into profit (as every company eventually does) OpenAI are going to get absolutely eviscerated.
The circular financing problem doesn't mean that startups building on AI aren't generating revenue from normal companies, it means that the money going into Anthropic and OpenAI is...
839 comments
[ 0.24 ms ] story [ 23.0 ms ] threadIn April 2025, Zitron wrote [1] "I am sick and tired of everybody pretending that generative AI is the next big thing." This was at the time when AI coding tools had already gone "mainstream" with devs, and pretty much everyone was using tools like GH Copilot, Cursor, or something similar.
So I replied with this observation, saying how, at least within software engineers, AI is being adopted faster than any technology before [2]. To stay objective, obviously I brought receipts: sharing how, based on an older survey I ran, ~75% of devs in that survey said they used AI coding tools. Zitron made fun of the small sample size (216 people), blocked, and pretended like AI had zero PMF anywhere in the world.
I stopped taking him seriously since then. And I'm wondering ever since: does he deliberately only look at numbers and facts that he can tell the AI doomer story around? Or is it more that he finds that there's not many people who are "informed sceptics", and decided to play this role?
[1] https://x.com/edzitron/status/1916903519594156407?s=20
[2] https://x.com/GergelyOrosz/status/1916906481686921483?s=20
From there everything split into two factions, which I’ll dub as believers and non-believers. From there, it has entirely been a cult following despite the evidence showing that models and agents have legs for software development.
FWIW my post history would show my extreme skepticism, and to an extent I still am. I think the real power of models isn’t the model at all but the harnesses.
Either way I think he lost the plot and runs on vibes himself. I also think this whole AI movement is going to have their 2000/2008 moment before the phoenix rises from the ashes.
All tools, AI included, have a business end and as long as you point that away from yourself you will be fine. But to properly apply it (rather than as a faster way to make a huge mess) takes discipline and being methodical. It was never different.
A frustrating thing about Ed Zitron is that he sometimes breaks news - gets hold of leaked numbers that nobody else has, for example - but he then wraps those numbers in his own extremely biased commentary. This makes it much harder to evaluate how credible the new information is.
https://www.cnbc.com/2026/07/29/openai-cfo-sarah-friar-tells...
OP is way overinterpreting something we don't even have verbatim in a way that unfortunately resembles what they're rightly accusing Zitron of.
The source you linked says: "Friar told staffers that annualized recurring revenue in July was higher than in the second quarter as a whole."
If we're charitable that means they annualized the quarter, i.e. multiplied by 4.
Which, if both measures are ARR, just means that a single month outperformed the average of three months, which is something that happens 50% of the time. Something you can opportunistically say whenever the coin flips the right way.
The less charitable reading is even worse, that one month's revenue times 12 is more than three months worth of revenue. Because duh.
Neither of these makes any sense.
“In an internal meeting with employees on Wednesday, finance chief Sarah Friar and board chair Bret Taylor touted OpenAI’s revenue growth and addressed competition with Anthropic, CNBC has learned.”
But the figures she is comparing are the revenue proper, not the growth per month.
I think the most likely explanation is that the CFO misspoke and intended to say something more like your quote, or perhaps she spoke correctly and was misquoted. But without audited financial statements, all we have is speculation, and there have definitely been times in the past when executives of major companies made intentionally misleading statements about their revenue. (In fact, this thread began with a question about why you can't just look at the numbers, and here the answer is that nobody has reported the actual revenue numbers this comparison is based on.)
In the face of AI, should be try to be even handed by an equally debased reality takes form?
Seems like one of those memes about the overton window. Critics are suppose to be even handed, but not the materialistic capitalists!
It's wild to me how much Ed is criticized while pathological liars like Musk and Altman are glossed over.
(Disclosure: I pay for Zitron's publication for the capital figures and exclusives he gets his hands on, I do not pay attention to his interviews, we are fact finding, if he wants to perform, I am not bothered, I've seen the performers on the other side, just keep the facts coming for ground truth)
No. For the same reason listening to Jim Kramer to get market data is a bad idea.
There are other, better sources for those data.
> For example, when Timothy B. Lee looked at a spreadsheet that Zitron used to create a projection of Anthropic's revenue, he found, "He doesn't count February 1-10, counts March 1-10 twice, counts August 21-October 21 as one month instead of two, and doesn't count October 21-November 1. [another commenter notes that his spreadsheet also contains February 30] ... Ed claims he tried to compute Anthropic's revenue for 2025 and came up with $3.6 billion, suggesting some funny business [but the numbers work out once you fix the errors]"
https://www.wheresyoured.at/exclusive-openai-financials/
https://www.ft.com/content/e15b0d7e-ff6b-4f16-ba7a-4068feddb... (https://archive.ph/pAIEa)
Hasn't failed - both in real life and online.
Does that make his readers suckers? Possibly. But personally I also think it's a very human trait to seek comfort.
This attitude is at the root of many of the world's problems, IMHO
Seriously, I expected more slight wrong before I read the article. Why do people trust someone who can't add numbers correctly to make predictions?
You can think of yourself and draw your own conclusions (or we have completely lost the aptitude since "ai" started off?). You don't have to agree with Ed, or even follow his conclusions.
It makes me a bit concerned seeing how many people say oh he is this and that so I wouldn't even read or help him. For G'ds sake, what happened to healthy discussions? If you disagree, lets just share opinions and let people judge. But if we just recuse ourselves from it, it will let opinions unchallenged.
I think the worst thing that happened to him was AI skepticism becoming a political position. This gave him a captive audience - as long as he says what they want to hear, which means that he can never ever concede that he might have been wrong or that AI might actually be progressing or having successes.
This is not conducive to good prediction long-term - rather it leads one to a state of cognitive dissonance where one's chosen enemies must be simultaneously terrifyingly powerful and incompetent dunces. The propagandist's disease.
He has built a following of people that want to hear his extra skeptical views. And even if he changed his mind about some things, he cannot admit it, as that is not what his followers want to hear from him.
It's rare that people make the news and build a following by saying a very balanced, down to earth opinions... unfortunately.
The other...
Just gets you disliked by both sides, "certainty sells".
Some folks would prefer a confident wrong/simple answer over a "well it depends/is nuanced" answer.
I listened to one Zitron interview on YouTube and my home page was immediately crammed full of similarly foamy-mouthed AI critics. (As well as plenty more Zitron.) And it's suggested some people with un-nuanced pro-AI takes, too. But what the algorithm has never, not even once, given me is a sober voice that calls attention to the nuances.
I don't think that's true - over the years he's changed his mind from "LLMs are useless slop machines" to "LLMs can be useful when applied wisely".
For instance in "The AI Hater's Manifesto" he says:
> It just came to me — the problem that I have with most people using LLMs is the delineation between outsourcing work and outsourcing thought. Those using LLMs to write little scripts or BQL code on a Bloomberg Terminal are inoffensive. [...] A tool being used as a tool to do tool things — in many cases involving the LLM writing a little 30-line Python script! — is not a problem, though it’s also not a trillion-dollar industry that needed to steal everybody’s art and writing.
Similarly, in "The More You Buy, The More You Lose":
> Sidenote: The only truly useful use case I’ve found is on the Bloomberg Terminal’s ASKB feature, which takes natural language and turns it into BQL code to make requests of Bloomberg’s datasets. It’s genuinely useful!
You can perhaps say that he underestimates what the technology is capable of (or, conversely, that other people overestimate LLMs) - but that's a different kind of conversation.
And if you take it as a given that AI will never be any better than ~~now~~ a year ago, and that anyone who disagrees is an idiot or a liar, then that pretty much demands that the entire AI economy must be as fraudulent as he imagines. Which, while it serves his purpose of serving up AI-skeptic invective slop well, doesn't actually model what's going on, which is speculative investments that have the potential to generate extraordinary returns.
Humans are built for fighting and killing the rival tribe, not for pondering whether the other tribe might actually be correct. We are exceptional for even being able to overcome ourselves enough that the outcome of "fighting and killing" can be minimized in the large.
I've got some bad news for you.
but it's subtly different, because incompetence is not weakness.
It's pretty clear in how people talk about billionaires: They must somehow conclude that the people most successful within a system that rewards certain things are not competent because the things the system rewards are not what they think the system should reward (it is fine to believe the system rewards the wrong things, but people walk straight into denying that those people have skill even within the system, or they claim that skill within the bounds of the system does not reflect any "real" skill - totally ignoring that elites tend to stay elites even through, say, Communist uprisings).
Hence how someone can say that Musk is "dumb" with a straight face.
It's the grown-up version of nerdy kids hating "the jocks" in high school. We're all just a bunch of dumb kids. Maybe obsolete children, but still children when it counts.
That said, Elon Musk celebrated cutting funding that fed starving children and supported cancer research by waving around a chainsaw on stage. The richest man on earth did this because he wanted lower taxes... for himself.
When Anubis weighs his soul against the feather it's likely to completely destroy the scale.
I hate him because he is a cartoon villian, who when given enough wealth to feed the world chose to punch down instead of lifting up.
Things can look stupid when we don't understand them, but there has to be some intelligence somewhere at some point even if the arrival of wealth then suffocates it with sycophants. If Musk or Trump or GWB were all *merely* the idiots they're often mocked as being, chances are we'd never have even heard of them. They almost certainly weren't even "merely average".
Now, stupidity that arrives after the bank balance reads a billion dollars, that's terrifyingly powerful.
2. You don't understand. Trump is successful because of his stupidity, selfishness, and vile behaviour. Not despite it. In 2016, the Republican primary field had 20 or so candidates on a wide range of reasonableness. Trump crushed them all, because he most represents the average Republican voter. If he were more intelligent, or if he cared more about anyone other than himself, they would not have voted for him. His arrogance, lack of intelligence and morals is quite literally his strength because it's what makes him so relateable to the American masses.
The other is that you are in a bubble and have been convinced that the leadership of the world’s premier superpower are below average intelligence, along with 70-80 million voters. The former is not something an intelligent person would believe, you have thrown out serious analysis for political theatre and memes and should rethink how you view the world. Now stop to consider that maybe actually half the country has a totally different worldview, and that their leadership, having reshaped global politics, is actually full of intelligent people who are cold and calculating. I know it’s harder to stomach, but just consider it for a moment.
Being propagandized into extremism does not require anyone to be a total idiot.
Everyone is susceptible to propaganda and populism.
Subset, propaganda is not strictly required for most of it, only for why e.g. Jan 6 didn't disqualify him.
I think he will be remembered for the decline and the loss of influence by America across the world that will be his legacy.
Given what words mean, approximately 122.3 million US citizens eligible to vote ought to be below average.
When you say bubble, I'm inclined to agree, though for reasons that I suspect are uncooth to say out loud: we here are a bubble of smart people, so to us normal looks dumb.
“The biggest single, best predictor of how someone’s going to vote in American politics now is education level.”
I’ll note you took “idiots” rather than uneducated - that demographic our President loves.
Instead, the smarter rich people stay out of the press. And the only reason we hear about them is because they were always very stupid.
Irrational means stupid.
Stupid can win for a very long time, and in the end the "loss" is born mostly by bag-holders.
They're not infallible, but modelling them as complete idiots that just happened to get lucky is also... Not really consistent with reality, as far as I can tell. These people might not be good people, but they're good at playing a certain kind of game.
Ultimately one cannot separate technology or science from politics, it's inherently political
I'm not sure I have the force of will to not succumb to it. I think the only way to avoid it is to be so sure of the thing you are doing that the audience reaction is not the key metric by which you measure your success.
Ideally we would reward expert opinions on track-record, instead of how they make us feel. But we don't. Also ideally, the expert's conclusion wouldn't impact their ability to pay the rent (as long as it's correct), but again, that's not the world we live in.
It is a political issue in the sense that the discussion around AI can't avoid material and cultural issues. Ed Zitron definitely has an ax to grind and I'm personally turned off by his hyperbole, but I don't think something being a political issue means it's not worth discussing. I've also noticed the pro-AI crowd pivot to accusing people of not having actual reasons for hating AI, being victims of Chinese propaganda, etc as a way to avoid actually talking about specific issues that people have about AI and DCs.
I'm not even sure it's political. I think it might be a religion now. He doesn't think that AI is bad, he believes that AI is bad.
This is about as far from "tweaking their numbers" as you can get. It's a standard way infrastructure-heavy industries structure their investments and people would be asking questions if they didn't do this!
> hyperscalers opted to lengthen the depreciation timelines of their GPUs.
Yes and so they should! GPU depreciation timelines used to be 3 years!!
Google is famously still running 10 year old TPUs at 100% utilization, and 10 year old H100s are worth more now on the second hand market than they were when they were bought.
H100 spot prices have only dropped from $5 in May 24 to $3.20 now despite the release of the B200: https://semianalysis.com/gpu-pricing-index/
I'm pretty sure your claim about TPUs is similarly exaggerated, only a v1 (barely) qualifies and would have no utility today.
Well, it's enough to throw off standard EBITDA accounting and allow firms to report fictional earnings numbers. A standard story has been that companies have beat their Q3 estimates, only for their stocks to go down.
If you take this to be his argument, then dan’s numbers are more consistent ed’s claim.
Luu's post includes a (long!) list of specific predictions that aren't "exacerbated"; they're simply wrong.
Luu's point isn't that AI is going to succeed or that the "AI bubble" will never pop. It's that these predictions are all wrong. If you agree "directionally" with Zitron, all that means is that you're skeptical of AI. That's a totally reasonable position to have, but it has nothing to do with whether Zitron's predictions are good or bad.
No, the whole thesis is XYZ likely fail because REVENUE RECORDS is not enough to dig out of hole relative to MAGNITUDE MORE SPEND. Saying Zitron wrong because XYZ made $2 for every $10 it spends revenue needed to justify spending. Fixtaing on the $1-$2 is misdirection/innumeracy, the thesis is in reaching the $10 relative to time, i.e. that $2 has to be $10 in X time, but the current velocity suggest it will not be.
I agree with Zitron directionally on accounting, I in fact disagree with him on AI... I am extremely AI pilled, i.e. I think there is a future where AI is worth trillions and will capture large swatch of economy. The transformation will be extreme, unlike any past revolutions... but the accounting suggest that future isn't coming in time to rescue current AI incumbents from finance blackhole, which some may survive, i.e. bail outs, nationalization... but the $$$ suggest however we get there, there will likely be massive $$$ corrections involved irrespective of adoption.
e.g. when he accused Anthropic of fudging their revenue numbers, which amounted to him screwing up a spreadsheet
It could be "tulip mania" or it could be "the internet".
Luu analyzed the numbers instead of just reacting to hype.
Specifically:
>> Oct 2024: OpenAI's forecast of $3.7B revenue in 2024 and $11.6B in 2025 and $100B in 2029 are absurd, "a statement so egregious that I am surprised it's not some kind of financial crime to say it out loud"
> Wrong (2025 goal exceeded, 2029 TBD but not an egregious financial crime level of implausible)
>> February 2025: Anthropic making $34.5B in revenue 2027 is "is laughable on many levels, chief of which is that OpenAI, which made around twice as much revenue as Anthropic did in 2024, barely made a billion dollars from API calls in the same year."
> Wrong (whether or not they make that in 2027, their 2026 ARR greatly exceeding that makes the 2027 estimate non-laughable)
The numbers being cited is ~100B is well within accounting/ledger maxxxing tricks relative to current pool of investment. Luu is not analyzing number's he's just listing and believing numbers, and analytically entirely avoids the core Zitron thesis... once you tap out of easy investor $$$, FAANG warchest, accounting tricks... where is the rest of the order magnitude more $$$ that justifies existing spent relative to time frame coming from?
[1] is a reasonable discussion of DC cost models, which calculates depreciation as part of the annual cost.
> here is the rest of the order magnitude more $$$ that justifies existing spent relative to time frame coming from?
That money comes from long term debt (ie bonds by public companies[3]) and new investment into neo-cloud companies (ie, IPOs like 4).
The justification comes revenue. Eg, the NScale IPO above[4] has $51B in long term contracted revenue with an annual run rate of $500M.
[1] https://epoch.ai/data-insights/ai-datacenter-cost-breakdown
[2] https://www.cushmanwakefield.com/en/united-states/insights/d...
[3] eg https://www.yondrgroup.com/newsroom/press-release/yondr-secu... (but you'll find lots of similar bonds issued)
[4] https://dealroom.co/news/143730-nscale-eyes-september-us-ipo...
You understand that this doesn't follow at all right?
The intermediaries margins can compress.
> opex low - capex premium is ridiculous right now
What does "capex premium" even mean?
Of course you spend more on capex when you build a data center than opex!
High capex matches the expected business model. If opex was high then everyone would be worried!
Investors exuberantly build $10 of housing when there is $5 of demand, builders extract $8, when they normally extract $2 under normal margins, builders raking it, but arrangement is net loses vs world where investors build same housing for $4 and make a profit. Intermediaries margins can compress but what they already extracted for current build out is already built in balance sheet.
>What does "capex premium" even mean? >Of course you spend more on capex when you build a data center than opex!
No. Historically DC opex > capex, i.e. 60-80% goes towards power... because hardware costs were relative low % of TCO. Historically without delulu AI demand, IC producers capturing much less margin and TCO of DC was much lower than it is now. It's not opex vs capex it's TCO. AI is paying $10 vs $4, when demand is $5, $10 isn't sustainable, $4 is.
Now builders will be fine in case of crash, they'll compress margins for next round of buildouts, i.e. bubble bursts, current spend proves not sustainable. This is where the crux of argument is...
Future investors post crash when margins revert towards mean will be spending $4 to supply $5+ of demand. And due to nature of compute deprecatiion (i.e. tulips) they will have more efficient hardware with less opex/capex TCO per unit of compute, with much more sustainable balance sheet. The builders are still fine with their $2 margins, it sucks its not $8. But that leaves the current investors who spent $10 with stranded assets that are not competitive with more efficient $4 future build out, i.e. current investors have balance sheet black hole that cannot compete with none bubble market force.
This does not mean AI is doomed, it just means incumbents from current tranch of bubble driven, stupid high TCO build out is most likely doomed relative to future entrants. Unless incumbant has unassailable moat, or other hedge/cards (i.e. political bailout/intervention). That is the actual argument, Zitron is saying current ecosystem economics not sustainable, not that there is not a future model that isn't sustainable. But it does mean a lot of current players are balance sheet zombies, who _should_ die. But a reasonable disagreement is reality is size of bubble + contagion risk + influence of incumbents i.e. trillion dollar companies is such that they have non market lever (i.e. politics) to save themselves... but someone else is going to be doing the paying for a model that is net loss.
Counterargument: As advancements in transistor densities slow down, the rationale for increasing depreciation cycles makes more sense. As the performance gap between new & 5-year-old hardware continues to shrink, then the need to replace older hardware similarly shrinks, justifying longer depreciation cycles.
The economic logic is if current spend vs revenue gap is not sustainable... hardware prices / margins will revert towards mean. That $10 hammer will be compared against a $2 identical hammer (margin reversion/compression)... or worse, a $3 future hammer that does $4 / past $20 of work. The future player who only paid $2 can charge much less... i.e. simply paying $10 limits ability to price competitively. The future player who pays $3 has 50% more compute than incumbent who paid $10. The important DC TOC consideration, is in world where DC cost regress towards mean, opex > capex... so merely continuing to use that old $10 hammer is losing MORE than buying a $3 better hammer, i.e. the asset is economically stranded, it is COSTING MORE to run old hardware than simply buying new hardware. It's MORE than economically useless and $10 past purchase price not just sunk cost but dragging down balance sheet as amortized liability aka it is full write down / loss.
20 bucks subscription a month, dario, you ain't fooling anyone with above-room-temperature IQ (Celsius). Even the expensive subscriptions - we know that's not the real price. And the limits, bumping up prices every two weeks and so on, just to keep the lights on while draining investors... Zitron's claims are pretty impossible to deny: the moment those companies go public, the real prices will need to come out of the shadows and end up on your monthly bill.
At the same time I fail to see the real world benefit to this, even in software: it's like comparing Lego (before they began their anti-consumer bs) to cheap Chinese toys. The moment you look at them side-by-side, you know which one is a premium product and which one is a cheap Chinese toy. While I have been a big supporter of open source since I was a child, I was never biblical about using open source - if some proprietary piece of software does a better job than the open source one - fine, take my money. Not anymore. I see the abysmal state of cyber-security as a direct consequence of slop. Slop-written code, slop-reviewed code, tests pass(also slop), ship it. Yeah, I'm not trusting you with my data, the hell with that, I'm self-hosting everything, adios. And I sure as hell don't trust the ai-bros for anything either.
And there's another thing: Microsoft was clear about it: github is constantly down because they can't handle the load. App stores are bumping up developer fees. Anyone can slop together a todo app(which was never hard to begin with). It's still a winner-takes-all economy - no one is going to install a todo app and migrate from Google/Apple just because. So the 1000 todo apps released daily will simply be a hole in the pocked of the people who slopped them together and another reason to be paying 500 bucks for 64 gigs of ram and another 500 bucks for a 2tb nvme. Raspberry pi's started off as educational platforms for 35 bucks and their commitment was to keep them at those prices. That's aged well. If I decide to upgrade my uconsole, I have to pull out another 350+ bucks for a raspberry pi. All thanks to the sloppification. The bubble can't pop soon enough and frankly I don't care what it takes down with it.
I knew that somewhere in the future, something has to give because you can't just go around saying anything without losing some credibility. You still have some ardent followers in his cult of a subreddit.
But just to be clear: Ed is part of a bigger problem in tech journalism which is characterised by extreme pessimism and excessive skepticism. It is not correct to view Ed in isolation rather to see it as a part of the culture in which he can thrive.
[1] https://news.ycombinator.com/item?id=48447549
[2] https://www.theargumentmag.com/p/ais-biggest-critic-has-lost...
Good critique of LLMs & the companies behind them is hard to find, and it is harder when people gravitate towards this sort of thinking.
Lol, I appreciate the sprinkling of well timed humor in an article that is mostly straightforward facts and analysis.
I didn't know he'd had a long history of mispredicting AI quality and growth of companies like Microsoft and Google.
After watching this space for a long time, I think the growth angle is... untested.
All the FAANGs were slowing down after Covid. They boomed as the money printer went brrrr, then they plateaued. They've also kind of covered a lot of their potential their main total addressable markets, with the big exception of clouds, which probably have at least 5 or 10 years of growth as workloads are still being moved away from on-premise. Maybe ads, too, since TV is still around and big and there are probably a bunch of other holdouts I'm forgetting.
Then, FAANGS started reaccelerating around 2024.
Part of it was due to internal reforms, basically, layoffs shaking some things up.
But I suspect the bigger part has been AI, driven by CapEx and all sorts of other things. The problem with the AI growth is that it's highly opaque. We don't really know who the actual clients are. It is <<extremely>> likely that for Oracle (OCI), Microsoft (Azure), Google (GCP), Amazon (AWS) their direct customers are just OpenAI and Anthropic. That's it. It's likely that 70% of the AI growth is just 2 companies. That can't be healthy, it's also likely extremely risky.
Just the fact that the new cloud growth is so opaque is worrisome.
https://www.baldurbjarnason.com/2023/ai-position/ https://illusion.baldurbjarnason.com/
As with most things, most people are somewhere in the middle, as the silent majority. You only see the comments of the strongly aligned, which are also the most emotionally motivated to comment.
The internet is not real life.
There are a myriad of people out there who have brought up genuine issues that AI presents, or preexisting issues that AI is exacerbating.
This series of articles is perhaps the best long form critique of AI as both a technology and an industry I've encountered [1].
[1] https://aphyr.com/posts/420-the-future-of-everything-is-lies...
It's interesting how he has pivoted from denying that LLMs are useful to accusing frontier labs of Enron-scale fraud.
I suppose it's the natural pivot you have to make when your whole business is an anti-AI newsletter that costs $7 per month.
I look at the rapid rise and decline of OpenClaw and get the impression AI has not really found any foothold among average people.
He does say that pretty often in interviews. That doesn’t change his thesis but that‘s one reasonable reason people dismiss him, he has often said that AI is useless when considering the externalities. And he will sometimes take a shortcut and just say „it’s useless, doesn’t do anything well“, which is of course way too simplified.
August 2024: "generative AI is a dead-end technology that has peaked”
July 2024: "Generative AI models aren’t getting more energy-efficient, nor are they getting more “powerful” in a way that would increase their functionality"
Nov 2025: "the fact we're running out of high quality training data and we're hitting the walls of scaling laws, in the training paradigm, these models aren't getting better. What we're seeing today is pretty much what they're always gonna be like"
There are other examples too from his prose of talking about how they are barely useful but I don't want to dig it up
You also dont have to pay.
I've spent a lot of time with Duolingo learners. They're all pre-A1. You're just lying to yourself to make yourself think your phone addiction isn't "that bad." Take a real language class.
It is simple. I do not want to and have absolutely no reason to take the language class. So, why would I? They costs money, time and are not something to look forward to. I got actual visible measurable results. Why would I changed?
Besides, language classes also fail. Going to language classes for years and not being able to use the language in any practical situation was always fairly common result.
> I've spent a lot of time with Duolingo learners. They're all pre-A1.
I know people who tested B2 after doing nothing but Duolingo German. And like I said, I can watch Netflix shows and read books in language I used Duolingo for. That is definitely something.
> Why would I changed?
Exactly.
If the frontier labs suddenly found themselves unable to compete with cheap open weight models running on widely available compute, then one might expect the frontier labs to be the only ones exposed to that risk, while a chip-manufacturer like nvidia could thrive in either environment. A partnership or commitment from the chip manufacturers to the frontier labs could change that. Whether that kind of inescapable connection exists is hard to predict without a lot of specific modeling, and at this point I'm inclined to think that neither chip nor datacenter demand is going to drop any time soon, and that the commitments would be unwound before a company like nvidia is threatened.
These companies are all heavily buoyed by their investments in AI which is essentially an oroboros of money.
I think his contribution to the discourse is the bigger picture. These companies are going into potentially economy-wrecking debt over a speculative future that looks much foggier than the last two previous technological booms.
that'll obviously never happen. it's wishful thinking and venting tbh. that's why i like listening to his stuff.
it's better than reading the wholly AI slop docs my CEO keeps sending out
Comments should get more thoughtful and substantive, not less, as a topic gets more divisive.
When disagreeing, please reply to the argument instead of calling names. "That is idiotic; 1 + 1 is 2, not 3" can be shortened to "1 + 1 is 2, not 3."
Please don't fulminate. Please don't sneer...
https://news.ycombinator.com/newsguidelines.html
US has 1.5x the money supply since Covid. This means everything has to go up 1.5x to reach the same parity from before. eg inflation. All corporate earnings are going to inflate as 1b yesterday is 1.5b today.
There is no whete for these mega corps to go. They dont know how to grow. meta becoming well meta with the vr crap is case and point. meta blowing billions on a few hyped AI people is the nail.
Other than cloud, microsft/google aimt doing anything. xbox? nope, hardware? nah. It is all attention economy or selling the pick axes for it.
the only mega corp that still seems to be moving the needle is apple which i think speaks volumes.
Eds point is that AI tam or expectations are insane full stop and he admits coding has a use just not as big of a tam. Gen ai in art, music, etc did not take off at all esp compared to code. this makes sense, no one pays an artist 500k/year. their time is worth little compared to coding.
Where he is wrong tho is that this is not new. there is always a hype cycle of nonsense that screws the little people. what is new is the level of polarization. it feels like ai psychosis right now in ways that remind me of crypto but worse because ai is actually useful so people are even more insufferable.
Zitron has a somewhat ranting style. Luu's passage on Google search for example just nitpicks on the person that Zitron allegedly named as responsible for the decline of Google search.
The real issue of course is that search quality clearly has gone down. What does Luu want? A "study"? Everyone can see that. The person is not the issue.
This is a highly selective analysis of Zitron's writings that focuses on a couple of mistakes. People need to understand that any normie will understand the difference between rants, numbers and speculation in Zitron's essays. They are not written for autists who talk of "priors" and "evidence".
While reading this I was thinking it would be interesting to see just a few examples of Zitron predictions that he got right. Since you appear to follow his work, do you know of any?
We have people telling us the next recession being imminent all the time. Others told us that NFTs were key to the future of art. That's the pundit way. If you want accuracy, you become like Charlie Munger, have maybe 2, 3 really good insights througout your career, and hope do to well using the insights yourself, not by being a pundit
Yeah uhm so I'm not sure if you've like seen the world recently, but uh.
Yea
Which leads me to a criticism of the piece: several assertions are described as “Wrong,” with no explanation or citation, which are not obviously wrong to my mind. For example, the assertion that “DeepSeek has commoditized the [LLM]” is at least debatable. It is a prospect that the major labs seem to have at least considered.
People who don't need a salary look at the situation more objectively and, in general, can see that a lot of white-collar work is in peril.
But I take your point, and I only brought up the o3 example to emphasize that people on both sides of the booster/doubter debate can be prisoners of the moment. There are boosters eternally convinced that the utopia (or armageddon, for the doomer-inclined) is either already here or imminent, and there are doubters eternally convinced that we have reached the peak. One of them will eventually be right, but neither has been yet. You can’t fault one of them for calling their shot if you’re not willing to see it happening on the other side.
What if (almost) everyone is right, that their work really is more difficult to automate than others realize?
This could be the lesser minds problem. Based on it I am doubtful most knowledge workers are as immediately replaceable as everyone suspects.
How can it do that?
In my experience, people who believe that some white collar industry can easily be automated with AI usually don't work that same role or even that industry, and suffer from Dunning-Kruger bias. In my line of work, I'm desperate for Claude to actually do a better job at not producing a big mess, and it's failing terribly.
I don't work in law, but I feel like I barely need a lawyer if I can ask an LLM to interpret a contract for me. I don't work as a physician, but why can't I just cut and paste an MRI report and ask Claude what to do. I'm no plumber, but it I take a picture of some fixtures I am thinking about replacing, Claude will give me some items to add to the shopping cart from a plumbing supply store.
In all cases Claude was super confident and it wrote what felt really rational.
But in my experience it gets things right, in some amazing ways, but it gets things wrong, in shocking ways.
Everyone thinks it's someone else's job that it'll automate.
That's quite the conflation, and a very subtle way to tie a claim that's still unresolved to one that is to try and make the first one seem wrong with no evidence. Recessions are almost guaranteed in our system, they're a part of a cycle. Unless you think that the Great Recession was the last one in history, saying the next one is 'imminent' would be correct for any other viewpoint. We just don't know when it'll officially start.
What? Zitron is getting a lot of media attention by repeating the same takes hundreds of times. He's not a general pundit.
They’re not the same take. Predicting collapse thirty days from now for three years running isn’t the same take, it’s a series of wrong takes.
Peter Zeihan
So from that standpoint, I guess it would make sense that a personality like that would now have ended up at the AI topic.
I would like to see a comparison with predictions of the pro-AI bubble and their hit rates though. This writing feels as selective as it tries to paint Zitron as.
Growth isn't a valid rebuttal, unless we can also sus out how much of that growth is tied up in circular financing of AI projects. We have a pretty good idea how much of Nvidia's valuation is tied up in the AI craze, it's a bit harder to tell with the megascalers....
There’s speculation on both sides and certainly Zitron is on the extreme end of the anti-AI side with the most cynical speculation but it is indisputable that none of the megascalers are open about their AI financials. Hence, we are all speculating endlessly. If only there were published financials then the speculation could end!
The obfuscation of financials doesn’t necessarily mean something bad is happening, it could be a competitive advantage for Google to be secretive about how cost effective their TPUs are or for Microsoft to hide how much revenue uplift they’ve experienced by adding AI to 365.
Page 24 of Amazon's 2025 report for example https://www.sec.gov/Archives/edgar/data/1018724/000101872426... separates AWS from the rest of the company.
Or Google/Alphabet's 2025 report https://www.sec.gov/Archives/edgar/data/1652044/000165204426... page which breaks out search revenue and YouTube revenue.
So the previous statement that "As public companies, the megascalers publish pretty detailed financial reports" is incorrect and irrelevant to the question that was asked.
The datacenter build-outs are all majority (>50% ownership) financed by other companies, with a shell company owned by the hyperscaler as a minority owner. The data center then grants the hyperscaler an exclusive leasing agreement, and because the shell company is a minority owner, legally, it's not their debt.
The only reason this has worked is because there's such a long delay taking delivery on GPUs. When these capital allocators start paying for GPUs for data centers which haven't yet broken ground, then we'll see a very visceral market reaction. Yes, some of this has happened already, but there's enough momentum that it can be absorbed and dismissed as an anomaly. But governments are unexpectedly passing moratoriums on data centers everywhere, so it's only a matter of time before there's no data center to offload those GPUs to. That's when the music stops.
I believe that was Zitron's central thesis and why he started reporting on this. This mirrors the mortgage-backed securities situation that led to the 2008 GFC.
https://youtu.be/HXlcMbxzz0U?is=XdvcNJKGJxEwlB7I
https://finterm.ai/blog/big-tech-hidden-debt-fact-check.html is one good explainer on that.
The predictions were predictions of revenue. Circular financing of AI projects does not create revenue for OpenAI, Meta, or Google only for Nvidia.
A startup raises $50 million from OpenAI and Anthropic to finance API calls to OpenAI and Anthropic that they are using at a loss who in turn spend that money on compute with Microsoft and Google who in turn invest in Anthropic and OpenAI who then invest the startup using the startup’s revenue to value it… the cycle repeats.
There are multi-billion dollar valued startups invested in by OpenAI and Anthropic with hundreds of millions in ARR that are spending 90% of their revenue with Anthropic and OpenAI.
Situational Awareness, the fund that recently imploded, invested tens of billions into AI companies using their holdings in Anthropic to help finance the investments…
This could all work out fine in the long term, we’re all just speculating at this point, but the circular financing is absolutely making it to revenue because capital invested into startups is used to fund growth which is achieved by subsidizing costs.
You are mixing up valuations with liquid cash and you're also making sweeping statements about how those startups are spending their cash. A majority of a raise is not spent on AI compute.
Situational Awareness blew up because they used leverage to invest, and leverage is a great way to blow up any fund even if they were directionally correct about AI.
Revenue numbers are vastly inflated compared to pre-AI but these startups aren’t keeping the money. Profits are worse than ever before. Startups with 30 employees that reach $100m ARR in 6 months are not banking $90m or $80m or… they’re just passing that money straight through to OpenAI and Anthropic.
If startups aren’t just funnelling all their funds raised straight through to OpenAI and Anthropic, where is this combined $100bn in revenue coming from? Who is paying for it? My spend on software certainly hasn’t gone up in a post-AI world. My company is spending less on software now.
OpenAI have stopped being so reckless with their cash investments which is why they appear to have slowed down but they’re still investing millions in huge numbers of startups through token allowances. They invest $2 million in every YC startup (or did a few months ago). There’s an entire market of reselling these things tokens!
https://mlq.ai/news/openai-and-anthropic-pour-up-to-800m-a-y...
I work with AI startups and scale ups on a regular basis as well as plenty of more old school companies, all of whom are spending money on AI models, because they are getting insane value from them.
This idea of the revenue for OAI and Anthropic coming from “circular financing” is just bizarre wishful thinking coming from AI doomers with zero financial literacy.
The revenue numbers reported by AI companies (not just OAI and Anthropic) isn’t being driven by Nvidia at all, in fact, the numbers wouldn’t add up if you thought that was the case. The revenue being brought in by AI companies is far, far higher than the sum of any investments from Nvidia.
The AI doomers just can’t handle the idea that AI is actually incredibly valuable and every company is using it and increasing their use of it every month.
And yes, I see this every day in my job and with every company I work with.
Real money, or credits?
I also contract in the startup space, and many of these startups have pretty much their entire infra bill covered by AWS/Azure/GCP credits, and all of their AI spend covered by Anthropic/OpenAI credits.
Theoretically they'll spend real money on those things down the line, assuming they find product-market fit, but who knows how many of the current crop of startups will reach that point
You presume to know my position but you do not. AI is an innovative new technology that is radically changing how we build and use technology and will continue to do so. That doesn’t mean that trillions of dollars is going to be spent on it. Despite the penetration all technology has in our lives, most companies are barely using technology from 20 years ago because implementation is a nightmare. Businesses are risk and cost averse, better the line item you know. And so, most companies could be radically improved not by human-level intelligence, or even dog level intelligence, most companies just need macros that are easy to implement. Most companies could 10x their productivity without AI! After all that’s what startups have been doing for the 20 years pre-AI, that’s been the YC investment thesis (which has worked very well).
My position is that AI is a radical step forward in technology that pragmatic businesses will benefit from handsomely by using cost effective models. A middle of the road local model that can trigger tools is more than most companies need. The frontier models by the frontier labs are a complete waste of money outside of the most extreme edge cases.
Conflating “the technology is incredible” with “companies will spend trillions per year on the technology” is ridiculous. Your argument about usage says absolutely nothing about the financials yet you’re dismissing the AI “doomers” (people who are pessimistic about the financials, not the technology) on that basis.
If you look at what we know of the financials of OpenAI and Anthropic it is impossible to come up with a financial case to justify the trillions of dollars in revenue needed for the AI booster’s vision of the future.
How much money does The JavaScript Company make? How much money did Docker make? It’s like the AI boosters who argue for the financial case have forgotten the last 20 years. The world of technology is built on open source, it’s built on companies that made a huge impact and failed financially. Docker led the way with containerization, one of the most influential technologies of the last 20 years, and the company almost went under multiple times. We constantly gripe about how unsustainable open source is. Why is all this suddenly different? Why is making an innovative new technology suddenly guaranteeing trillions in revenue?
How many trillions of dollars were invested in data centres to build Docker containers?
If you think I lack financial literacy, please explain where the money is going to come from. Please make the financial case for trillions of dollars being spent on AI over the next few years. Keep in mind that the reason technology has been so profitable over the last 20 years is because of the margins, software is basically free money. AI is not free money. AI is very expensive money. Also keep in mind that the current (rumored) revenue of Anthropic is primarily made up of the most expensive use case (generating millions of lines of code) being paid for by rich tech companies which does not represent the wider economy.
Bets are meaningless but feel free to stake a claim here to how you think things will be 4 years from now. I’ll stake my claim: AI will be more impactful than ever while Anthropic + OpenAI will have less revenue than today.
It's coming from regular companies spending their own money on using AI. I.e. Profitable companies deciding they want to use AI for various reasons and spending their own revenue on said AI, whether it be Claude Cowork, OpenAI ChatGPT Work, OpenRouter, Nebius Tokenfactory, models hosted on BaseTen, Fireworks, etc, tools like Lovable. Or every piece of cyber software which are ALL using AI heavily these days.
It's really not as complicated as AI-doomers like to make out, they seem so confused somehow that existing profitable, successful companies are spending larger and larger amounts of their revenue on AI. It's not circular by any definition.
So anyway, this whole worry about "where the money comes from", is kind of funny. Where does the money come from to hire employees? Where does the money come from to pay for Cloud bills? The money for AI will come from the same place, it's not some big mystery. Total cloud/compute spend in the world is well over 5T per year, including all cloud and colo spend.
AI is basically both taking up software spend, dev salary spend, white collar officer worker spend, hardware spend, general IT spend, etc. And if you sum up all the budget associated with all company software, personnel, white collar workers, etc, you end up with a much bigger number (probably 10-20T or more most likely).
So the idea of total AI revenue being in the trillions, it's pretty simple, and will happen over the next couple of years, just like happened with regular servers and cloud.
People like yourself downplaying AI reminds me of people both downplaying the internet ("it will never make money") and also the original launch of cloud providers (AWS originally), "no one will ever trust the cloud not to lose your data, why would any pay for AWS" etc etc).
Both sets of folk were radically wrong, and the AI-doomers will be wrong this time too. Capabilities will increase across the board, amazing applications will be built (already happening), and people will want to pay for these products. People are ALREADY paying huge amounts of money for these products.
Who do you think is paying for Lovable? They are probably the fastest growing startup ever from 0 to 1B valuation because less technical people LOVE using it and have zero issue paying for it. But AI doomers will somehow dismiss Lovable as somehow getting "circular financing", when loads of small business owners I know love the product and spend 100s of dollars a month on it!
It's going to be funny watching the doomers over the next two years when none of the big AI companies goes bankrupt and keep growing in revenue. But but but the circular revenue!
Apart from the brief "tokenmaxxing" craze among the big tech firms a while back, is there any evidence that profitable companies are cutting their own margins in order to spend on AI?
Regarding the cloud infrastructure comparison, it is not at all comparable. During the time I spend writing this comment, my device will make thousands of requests and connections to different servers for all sorts of reasons. During the time I spend writing this comment, my device has interacted with an LLM exactly zero times. The throughput of internet infrastructure is not even in the same universe as the throughput of LLMs at their most wildly successful. How many times does Lovable's AI run per month for their average customer? A few times? The repeated, continued value Lovable delivers to their customers is in the interactions that occur between their customer's customers and their customer's apps. A Lovable customer can love Lovable and have huge success with their Lovable app while using zero tokens per month.
You should be comparing AI to a product that eventually became commoditized, not comparing it to an entire category, e.g: shared website hosting. Shared website hosting was very expensive to set up 30 years ago. Over time, it got cheaper and cheaper, now today it is commoditized, the major brands have all consolidated, companies have gone under, and technological innovations have completely reshaped how websites are hosted. Who still uses shared website hosting today? Websites are bigger than ever, web servers underpin the economy, Stripe alone has web servers that process trillions of dollars... how much money is there in web servers?
> Capabilities will increase across the board, amazing applications will be built (already happening), and people will want to pay for these products. People are ALREADY paying huge amounts of money for these products.
You're so caught up in the technology that you're oblivious to the economic reality. The capabilities, the amazingness, the excitement, that isn't how money is made. The most cheap and boring technology (like web servers) are fundamental to our economy. AI can be all of these things, it can have incredible capabilities and be amazing and have so much excitement and radically reshape our economy and be fundamental to every business... and make no money.
You, like so many nerds, cannot seem to separate technology from business. Business is boring and simple and based on principles that have stood the test of time. Business doesn't run on excitement, it runs on numbers. Shopify powers most ecommerce, Shopify is one of the most important companies in ecommerce, Shopify is wildly successful, Shopify's revenue... $12bn. Stripe's revenue, on trillions of dollars in payments... less than $10bn. Shopify and Stripe are wildly successful and very important companies that are involved in trillions of dollars flowing through the economy and you're suggesting that AI is going to do 100x more revenue than them?
Tailwind CSS is used on probably half of all major websites today. The creators of Tailwind recently announced they had to lay off everyone because the company was struggling to make any money despite usage growing every single day. WordPress, which (supposedly) powers half of all websites is operated by a company that is struggling too. Google and Meta, some of the most profitable companies in the world, almost all of their revenue is still from advertising that has barely changed in 25 years. Google make hundreds of billions of dollars from... showing...
There are zero serious companies collecting $1000 on revenue and sending $999 as a cost of goods sold to Anthropic/AI. It would be unprofitable to even run a proxy to Anthropic on such thin margins. But I digress.
No company was banking $100 and keeping $99 in the "before times" either. These are fantasy numbers not even the most highly optimized software company produced. As an example, Slack famously went public in 2019 and it had revenue of $401M with a gross margin of ~79%, meaning they were pulling in $316M in gross profit. That is the figure before labor, administration, R&D, sales & marketing, etc. They actually operated on a net loss after factoring for those expenses, despite their high gross margin, which is common in high growth startups (Amazon famously ran losses or marginal profits until decades after their founding because they continuously reinvested in expansion).
Credits reduce revenue by all basic accounting standards. You can accuse these companies of fraud, it is within the realm of possibility, but it would also be <1% of their total quoted revenue, so not really worth the heat at the same time.
You are making conflicting arguments at the same time. There exist startups that are able to generate gross profit with some consumption of AI services, they are also able to invest nearly 100% of their capital into AI to generate those profits without needing to spend on traditional labor, and yet AI is not sustainable. By your own circular logic it is of course sustainable, but by grounded logic, you have to understand any business that goes from zero 4 years ago to $100B+ in annual revenue today with double digit growth rates is offering the world something of value. Anyone who has tried AI sees some value in it. There is some revenue and profit to be made here. Betting against that in the long term will just lose you money and sanity.
Yes they were and are. The marginal cost of software as a service or data as a service is near zero. Slack is a good example. A new Slack customer costs Slack nothing. Free money! Slack had a high valuation because of the margins. Slack and other traditional high-growth technology companies were valued highly despite being loss making because there was an understanding that paying for growth early returns a lot more later on. Slack (pre-acquisition) could turn off their expensive growth engine and start making money hand over fist.
(Look at what Bending Spoons are doing now, they're picking up "zombie" technology companies that have incredible margins but no growth. Bending Spoons are cutting these companies to the bone, giving up on growth, running on a skeleton staff, and making money hand over fist, cashing out on the incredible margins of software.)
Someone shared up thread an example of Harvey, a legal AI company, who regularly post about their token consumption. They're consuming trillions of tokens per month for their product. Harvey's investors include OpenAI. Harvey has raised more than $1bn and is currently valued at $11bn (and raising again at $15bn apparently). As of last month, Harvey's revenue was reported to be $30m/month on 13 trillion tokens per month.
Let's be conservative and assume their average spend per million tokens with OpenAI is $2. That's $26 million in token spend per month, on $30 million per month revenue. $2 is lowballing it of course, they're surely using one of the frontier models. That's pretty close to every dollar coming in going straight out to OpenAI. Considering the capital they're raising and burning (seems like $50m a month) while relatively small (<1k employees) I would guess they're spending at least double their revenue with OpenAI.
Of course, long term, this is fine for Harvey, as model costs come down and businesses mature they are going to be spending a lot less. Maybe they'll start running their own hardware, offloading certain workloads to cheap models, using scripts for routine tasks where AI is overkill. Great for Harvey and Harvey's investors, an absolute disaster for OpenAI.
> By your own circular logic it is of course sustainable, but by grounded logic, you have to understand any business that goes from zero 4 years ago to $100B+ in annual revenue today with double digit growth rates is offering the world something of value. Anyone who has tried AI sees some value in it. There is some revenue and profit to be made here. Betting against that in the long term will just lose you money and sanity.
You're making a leap from "useful" to "profitable". Yes, there is absolutely revenue and profit to be made for companies building products, for the companies providing technology, not for the companies providing inference. There are not software margins in inference, it's a commodity, the only reason OpenAI and Anthropic went "from zero 4 years ago to $100B+ in annual revenue" is because nobody cares about the money today.
Right now, we're in a gold rush, we're in the growth-engine phase, we're in the "spend a billion to make a million as long as you're growing" phase. Right now, people at Harvey aren't worried that every dollar in is at least a dollar out to OpenAI, who cares, investors are funding it, they're growing, they're taking over the legal world, that's all that matters, they can balance the books later... and when they do start to balance the books, when they convert that growth-at-all-costs into profit (as every company eventually does) OpenAI are going to get absolutely eviscerated.
The circular financing problem doesn't mean that startups building on AI aren't generating revenue from normal companies, it means that the money going into Anthropic and OpenAI is...