Posted by jatins 11 hours ago
Regardless of whether you love LLM’s as technology, the financial realities of Anthropic and especially OpenAI really does not look good. They have a massive expenditure that they need to keep going in order to make profits, which at least in terms of OpenAI are horrendously behind. Zitron published these numbers together with Financial Times, so you gotta give him that at least. Meanwhile the CEO’s talk all kind of nonsense and give their own predictions to get more investors money to fund what might or might not be the biggest bubble in the history of finance. I certainly do not hope this happens, since the consequences would be horrific. But there is likely to be a some sort of correction in horizon, since the models will plateau and they will run out of money at some point.
Let me finish with my own prediction. I think a lot of people are going to lose a lot of money some time next couple of years.
Whenever you'd look up anything pertaining to China's future, you'd inevitably find your screen plastered wall to wall with thumbnails of a photoshopped Xi Jinping, tears streaming down his face, next to large impact font text reading "CHINA WILL COLLAPSE IN X DAYS", with X varying from 1 to 30. Much like Ed Zitron's predictions, these events obviously never occur.
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.
https://danluu.com/futurist-predictions/
Yet, in this case, he takes the marketing numbers at face value, not being an expert in AI financing, while those who know more, can spot the sleight of hand, just like he can when it comes to semiconductors.
What I’ve stopped doing is reading him regularly. It feels hard to parse the factual from the obviously exaggerated.
I get he’s frustrated. We all are. But I’m not sure letting it out that much helps making the very urgent case he’s making.
I think something similar is happening with e.g. Amodei's predictions of mass unemployment due to AI. It won't happen in the immediate term, because deficit spending removes the economic incentive for firms to pare down their workforces.
All that said, I don't think these people are "wrong," per se. They're just early. When the sh*t hits the fan on all this, it's going to be a big problem. And, for example, companies whose primary business is collecting money for Internet ads will come face to face with the reality of how low value their products are. I have some insight into this, as I work for such a firm, and I know the true extent of the bot traffic out there.
Its silly to say hes just early, when his predictions give specific timelines that don't work at all
- "so egregious that I am surprised it's not some kind of financial crime to say it out loud" — on OpenAI forecasting $11.6B for 2025 (https://www.wheresyoured.at/exclusive-openai-financials/) actual: $13.07B. source is his own scoop (https://www.wheresyoured.at/exclusive-openai-financials/)
- "artificial intelligence has three quarters to prove itself before the apocalypse comes" — Mar 2024
- "If OpenAI doesn’t either reduce their $8.5bn operating costs to $1bn or less and raise at least $5bn in the next year, they will die." — Jul 2024 2025 costs: $34B
- Generative AI “isn’t getting much more efficient” (Jul 2024 ). OpenAI’s frontier-model API price fell from $30/$60 per million input/output tokens for GPT-4 to $4/$20 for GPT-5.6 Sol, alongside major capability gains.
- “I would be shocked if [Musk’s] wealth doesn’t return to something more like he had in 2019 or 2020” (Dec 2022 ). Musk is now worth approximately $873 billion , several times his wealth when Zitron wrote this.
He goes on to specifically discuss how unrealistic $100B by 2029 is given that OAI is structurally unprofitable.
The fact that you’re skewing your misinterpretation of what he said so much shows your own bias I’m afraid.
I left Microsoft during the windows 8 cycle in large part because I could tell nobody knew what the hell they were doing, which is an objectively true statement about the time and place. My dad happened to buy Microsoft stock at that same time and did very well with it.
That's the paradoxical problem that I think all big tech has. Poor decisions by incompetent people, met with inexplicable financial success.
You might choose a different timeline for Ed’s outcomes. Okay: those are now your predictions, not his.
Zitron has done us the favor of including timelines with his predictions, so that Dan can invalidate almost all of them.
The US typically adds that much annual GDP every 16-24 months at this point. The notion that somehow the gigantic ~$31 trillion economy will fall apart if the outsized deficits don't continue, is very absurd.
The exact same things were said of the Bush deficits. The US economy was supposedly dead in 2009-2010. Here in 2026 the economy is 50% larger inflation adjusted and it has left most of Europe in the dust. The housing bubble contagion was much worse than anything we're sitting on now with AI spend.
Why do I say $1 trillion instead of $2 trillion? There are plausible scenarios where increased taxes bring down the deficits (the Dems will take the House + Senate + Presidency, we'll see how much taxes go up), there's no plausible scenario where the deficits go away completely.
That means his predictions were either wrong, or meaningless.
on this specific point: tokens aren't fungible between models right? Like the argument Zitron makes is that improvements in output are due to, glibly, using more tokens to get there. We see people turn on new models and instantly use up all their tokens.
Like the actual measure is more something like "for this specific task, did it cost less now to do it than it did 3 years ago with these AI pipelines" right? The token pricing isn't actually relevant in that discussion.
By that metric, so was Nostradamus. The apocalypse is coming for sure, we're just quibbling about the timeline.
Realistically, timing is everything. You don't need to get it precisely right, but you also don't get a pass if you, for example, keep predicting an imminent recession through a decade of unprecedented growth.
This was said in February 2024. This is months before GPT-4o released. GPT 4.5 was released a year later.
I don't know anyone holding on to models of GPT 4.5 caliber, yet know 4o. ChatGPT 4o and 4.5 score 8 and 14 points on Artificial Analysis benchmarks. [1]
For perspective, Qwen 3.6 27B, which can be ran on single GPU setups, scores 3-5x that on modern benchmarks.
I don't know why anyone would even make a claim like that in the first place. It's like saying "Computers are never going to get faster". I feel like we could run out of sand and still have faster machines over time. Just a silly thing to say.
[1] https://artificialanalysis.ai/?models=muse-spark-1-2%2Cgemin...
But there are enough signs of trouble that could happen soon: Oracle debt is junk. OpenAI might not be on a viable trajectory to IPO. One or both of those could collapse the lower quality data center companies. Zitron is probably overconfident about a crash in the short term. Probably.
Zitron is fairly clear that nothing is going to happen for a year or so — he says himself that he thinks there's another round of funding possible for both OpenAI and Anthropic.
Markets don't just pre-plan their behaviour three years earlier and act it out lock-step. Other circumstances can change. By now, the world's financial press has covered some of the scariest aspects of this, and the situation has evolved.
There are other moves (the OpenAI/Blackrock AI debt securitization idea for one) that could delay it even further.
Zitron, I get the impression, has moved on to talking about the horsemen of the bubble apocalypse — talking about banner events that would need to happen for his predictions to be true. This is safer ground for a forecaster, because every forecast affects the future.
His shorter term predictions have not all failed by any means: he described Oracle's woes before the ratings agency downgraded them specifically because of OpenAI.
But most of his predictions will be irrelevant if the insane securitization plan happens. Because it will stop being about an AI bubble then; the worst risk will be the collapse of the entire US economy. It will need a different kind of analyst.
Me, I don't really care either way. I'm not on the cloud AI hype train, I don't work for a YC company, I'm not an American taxpayer so I will not be directly on the hook, and as Americans like to point out, the UK economy is behind on the whole AI thing so (unlike Ireland, which the USA will 100% leave to fail) we are ironically insulated. Maybe a couple of small British investment banks will fail and a pension fund or two will default.
For the most part we'll just watch the flames.
I do enjoy watching a Brit — albeit an ex-pat — upset a bunch of po-faced AI evangelists. It’s like “Itanic” all over again.
I think he is directionally correct. My own impression is that the bubble will burst at the worst time, and so my assessment is that, given the way this is entangled with the functioning of the USA as an economic power and with the future of the US political hard right, it will therefore darkly but poetically burst sometime around Labor Day 2028, which is the worst possible time.
But predictions need to be specific and falsifable. If not, its just rag-chewing over a beer (luv that shit, but i aint predicting on taco tuesday). If they arent falsifable, then its not a prediction.
I think a lot of the HN comments generally can be described as one camp which cares about and enforces the rigor of predictions and trying to direct limited ear-time to voices which tend to get predictions right, vs the other camp that puts more weight towards directional accuracy.
Can you give an example of what a good/proper prediction might be, even in a hypothetical universe, in this schema? Does one "predict" when they play blackjack? Or is there a different concept for that kind of thing?
This is a spectrum of course: a prediction that OAI will collapse is probably right as _eventually_ all companies come to an end, but under that interpretation, the prediction is useless. It's more signal / useful / falsifable to say OAI is going to collapse around/at <year> due to <thesis>.
Anyway thats my two cents. Its fine to outline forces and trends, but when you make predictions there are useful (better, falsifiable) and useless.
Bad (form) prediction: OAI is gonna be wobbly in a bit Good (form) prediction (could be totally wrong): OAI as we know it today is going to collapse due to running out of money around/at 20xx.
And sure we can be pedantic about detail, but the litmus test is: is the prediction useful if you had a crystal ball and you could know if it was true/false a priori?
One pony’s trash is another pony’s treasure, so I guess here one persons pedanticism is another persons hobby.
Vs of course Nvidia: $302b sales, $197b op income.
Oracle was never as big of a deal as that brief market cap run implied. The herd pushed it up for no reason.
The comparison to Apple, Microsoft, Google and Amazon are pretty similar. Oracle fits in their pockets too. Who cares if their debt is junk. Ellison has nearly wrecked that ship on numerous occasions over the decades. He went on an elaborate acquisition binge in the previous epoch, buying his way to the next stage (preventing Oracle from being market-eliminated, or acquired), and that was an incredible mess that took a long time to sort. He's doing the same move now, trying to spend to stay on the board as the world rapidly changes under his feet.
Non-rhetorical answer to your rhetorical question, but:
The near future of the political right wing of the USA is dependent on the Ellisons staying afloat to create an impervious right-wing media sphere that would survive the end of Fox.
The Paramount-Skydance/Warner merger is now delayed until 2027. Trump/the GOP needs that merger to go ahead, but if Larry's debt position worsens it really might not.
So you can expect the executive branch to push for the USA to backstop Oracle's debt in some way, whether directly or indirectly (taking some sort of stake in OpenAI to allow it to guarantee Oracle gets most of its money, for example — anything to get the credit rating back up).
It will be the first stage of this becoming a problem for the American taxpayer.
The whole point of making predictions is timing. I can tell you the US dollar will continue to devalue (a 100% accurate prediction). But it's a worthless statement unless I can tell you when and how much.
FTA:
> Note that I didn't attempt to catalogue statements that are nonsensical or were simply factually incorrect statements at the time, such as his December 2024 claim that “Generative AI's products have effectively been trapped in amber for over a year.” January 2026 claim that "[models are] basically the same as they were a year ago. They have the same efficacy". Zitron has not only made forward-looking statements that AI capabilities will not improve, he's also consistently made backwards-looking statements that capabilities have not improved which, while obviously false at the time, seem to play well to his base (along with his other false statements). If you connect all his statements together, it's implied that AI had the same capabilities in January 2026 as they did in December 2023 (and if you connect later statements, it's actually implied that capabilities in August 2026 are the same as in December 2023, though to be fair to Zitron he frequently contradicts himself and has also admitted to limited improvement at times).
For example saying in 2024 that LLMs had peaked. That’s not early, that is already, definitively wrong.
If his issue is that he is accurate on something shady going on with the finances of AI companies but the effects are mitigated due to government intervention, he should say.
Your explanation gives him more credit than I think he deserves. He has been unambiguously incorrect many times over the past years in the time scales over which he makes predictions.
In general, claiming that a prediction is just early in an unfalsifiable claim. It allows no admission of being incorrect whether you are correct or incorrect in a given moment. If you are right, then good, you've made a correct claim. If you're wrong, just claim there are clandestine corrective forces keeping the disaster you are predicting at bay. Either way you make it out clean and still have some sense of legitimacy.
Google's 500 million Gemini-user goal had an end-of-2025 deadline. Zitron called it so unrealistic that Sundar Pichai should be fired. Google reported more than 650 million monthly users by October.
"AI had already peaked" is also a claim about the state of the technology at that time. Agent and coding benchmarks moved sharply after it. The fact that every technology eventually peaks does not make a past claim that it already peaked correct.
The bubble may still burst. That would validate the bubble thesis. It would not retroactively fix the prediction record.
btw Zitron has the only fan base that has come after me with doxxing and death threats so far. Truly misery loves company!
If it does not burst it will be because specific efforts have been taken to deflate it in the face of concerns like those he is raising.
The bubble may not burst for example if the securitization of AI debt really happens. Then when it happens it won't be an "AI bubble" that bursts, it will be a full-on collapse of the US economy. Like when 2008 happened it wasn't really about mortgages anymore.
https://www.crisesnotes.com/sigh-no-ed-zitron-ai-bond-issuan...
The problem is you have the vise of a stock-market decline on one side (something basically everyone thinks is impossible), and AI-induced unemployment on the other side (something a lot of commentators, including Zitron sometimes, seem to think won't happen). Those two things in tandem would be worse than 2009 by a multiple. I doubt the US government will be able to bail it out.
I think it might have helped a little but I never fully bought this argument. I don't think I've ever bought a product which was advertised to me online and I was using some of these platforms for years. I'm probably a liability to them; using up compute but not clicking on ads or buying anything.
Also when I ran social media ads many years back, I never got any users out of it. It literally seemed like mostly bot traffic back then; I can't imagine how bad the situation would be now with LLMs.
https://www.wheresyoured.at/measures/
Now ORCL is back to $140.
I found this in two minutes via a search engine, but the star blogger Dan Luu apparently cannot handle that. I'm not a regular Zitron reader, but incidentally this blog post that came up in the search is several levels above Luu's post.
Zitron gets the big picture right.
> Although Zitron's past predictions have generally been wrong, maybe he'll be right about something in the future. Perhaps some of these companies will have valuations decline for some reason. But, even if there's some kind of massive AI crash and OpenAI and Anthropic go to zero, in terms of the societal impact, if on top of that, some other event occurs that prevents further progress in models beyond whatever AI labs have internally right now, that's still going to result in a fair amount of change. Which companies are successful will change who gets rich, but particular companies failing won't stop changes that fall out of current or next generation model capabilities from happening; it just moves around who benefits the most.
> If Zitron ends up being right about some company or other collapsing, that's pretty uninteresting to me compared to how capabilities have developed and will develop, where he's been wrong to date. It also happens that he's been wrong about the financial predictions he's made to date, but that doesn't really interest me, though I included a number of financial predictions for completeness.
Not sure how you can "get the big picture right" while having many egregiously wrong predictions.
And for personal use, I can count on a single hand people I know who pay for an AI subscription, and of those people nobody pays for more than the $20/month plan. And they all just use it as search or maybe to vibecode some one-off party game they use once and then throw away.
I don't know a single person who has done something like run Openclaw or leaves coding agents running on their laptop open all day.
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.
I didn't know he'd had a long history of mispredicting AI quality and growth of companies like Microsoft and Google.
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...
- safety critics who think AI can take over the world like Yud (I find this the least credible but still valid)
- Bernie type of critics who think AI can cause widespread job losses
- Ruxandra Teslo who thinks AI can remove meaning which I feel is the most serious one [1]
What are not valid
- environmental like emissions and water usage
- AI is useless and it will take the economy with it because it is a bubble
- AI spreads misinformation and causes societal damage
- AI is trained on copyright (are we really on this side of the debate ?!)
Water usage, sure. But emissions has plenty of reasonable concern.
There are the various xAI data centers have/are running using mobile gas turbines.
In general, the extreme amount of power is going to put pressure on the grids. I hope this leads to the world doubling down on renewables to offset it all, but is that going to happen? Hell, the US actively paid [0] to stop a turbine project.
These companies subsidizing green energy expansion, because it's now the cheapest power to expand.
These companies are going to help "correct" the widespread problem with utility monopolies, in the US. Datacenters are deploying their own power generation, partly because power cost no longer aligns with power generation costs. This mismatch is motivating research into local nuclear power generation [1] (which is almost certainly an effort to force the monopolies, rather than actually deploy).
[1] https://www.reuters.com/legal/litigation/big-tech-puts-finan...
Unless you consider natural gass green energy, it's quite the opposite for the moment. This is because the AI build-out is about speed, not cost-effectiveness. That is why xai's Colossus I & II were illegally running gas turbines, and Google recanted its pledge for data center renewables targets.
This point is about power generation companies accommodating the new power demand. It's true that the power companies are keeping coal and gas peaker plants running, that they planned to retire. But, actual additions are all green [1].
And, step functions are inefficient, always, so "for the moment" isn't so unreasonable. Dataceners are something like 100 million tonnes of CO2 per year, where cars are 1,800! So, this isn't some world ending emissions, within this moment. Everything is trending green. Why? Because it's cheaper (thank you China).
Xai using gas turbines is closer to my second point of side-stepping monopolies (improperly in that case), but is also just one example. See previous link for very long term, and all the datacenters using green energy, mid term.
Everyone will do what's cheapest. Luckily, China has made that solar power. We just need to get the monopolies off their asses, and put some of their record profits into actual power expansion. Lucky for them, everyone blames the data centers for all of this. Where I am, the single power company has already raised rates so much, in preparation for electric, that it's more expensive to fast charge an electric car at noon than to buy gas, with a planned 10% increase over the next few years. So, they're dodging criticism too!
https://www.siliconvalley.com/2026/09/01/us-battery-installs...
I fundamentally don't think it is wrong to increase emissions as long as you consider the tradeoffs and externalities. Every single action you do in life has externalities - if you start opposing all of them then what really is your point? You just hate people doing stuff.
If you think emissions are so harmful, try putting a number to it. I implore you to do the exercise and convince yourself or me or others and suggest that the pros don't outweigh the cons. I'm half predicting that the conversation will end in dubious claims about tail risks and world itself collapsing (I hope you don't do that).
I don't think AI is uniquely harmful for the environment given the value it provides. Hell, it can even contribute to accelerating renewable resources and increasing efficiencies overall. There's way more to lose by slowing down AI because of emission control than to gain by reducing emissions.
You can just flip this around and ask why is slowing down AI to control emissions harmful? What numbers that don't make tail risk claims are you providing for this?
>You can just flip this around and ask why is slowing down AI to control emissions harmful? What numbers that don't make tail risk claims are you providing for this?
Slowing down reduces the value it provides to humans, that's clear to me and you I hope.
Because climate change stands to be the biggest market failure in history. You need to show that the rapid scale out of data centers is going to reasonably offset the current trajectory we're on.
Proof?
If AI works out as proponents would make you believe, then either you'll see masses of people out of a job - for which society is unprepared. Or you'll see a big, society-wide productivity boost. Read: consuming non-renewable resources on this planet even faster (not just energy).
Society as a whole would benefit if rollout would slow down. Give us time to reflect on those 2nd-order effects & how to address them. But instead we have the opposite: a crazy race with big AI labs trying to out-spend & out-do the other guy, ignoring externalities everywhere.
- AI centralizes power in the hands of capital, rendering those who can afford compute hardware vastly more capable than those who cannot and thus increasing social stratification
- AI is generally trained on the creative output of humanity without those who train it giving back proportionally (copyright "rules for thee, not for me")
- AI breaks social processes built around the idea that TRYING something is inherently a cost in time or effort, such as filing a legal claim or sending someone a threatening letter. We haven't made the social changes to punish or charge people for using every appeal/option/application, so this makes asymmetric-effort tasks like applying for a job really bad in the interim
- AI use makes it harder to develop the ability to critically think for yourself, especially among those who most need to develop that ability
- AI produces large amounts of mid-tier content, making it harder to discover exceptional human-created creative content produced after AI started to exist
- AI leads to distrust in remote communication, increasing cynicism and breaking social bonds generally. This is NOT your point about "AI spreads misinformation" - no matter whether it's true or not that AI can be used to produce misinformation, having people doubt each other is a harm
- AI demand crunches hardware and time availability for other adjacent markets, such as computer gaming, construction, 3D graphics production, etc. This harms both hobbies and professions in those fields having to cope with rising prices and lower availability of materials
- Everyone is using the same or similar AI, leading to a homogenization of culture and process across humanity. This is perhaps a mixed blessing, because humans are capricious, but less variety can be viewed as a harm
I will also say I personally hate seeing "job loss" said to mean "wealth loss" or "people starving". The goal of life isn't to have a job, it's to be well and happy. If you can be well and happy without a job, great, so it's really painful to me how people don't even see those things are not the same.
Looks to me like I can run Claude Code without being able to afford my own datacenter.
> - AI produces large amounts of mid-tier content, making it harder to discover exceptional human-created creative content produced after AI started to exist
That it does, but I think the real social evil is bad recommender systems. eg YouTube is basically on a mission to drive me insane because it literally only recommends me a) reviews of espresso machines b) video essays by autistic people about Mario 64 c) PBS scienceslop about how quantum physics is super mysterious. None of these are even what I watch, but they're also not what I want to watch.
Until Anthropic bans you from using their data centers, at which point you cannot run Claude Code at all. Welcome to being a have-not (at least in a world where only genAI-assisted coding is acceptable).
He who controls the GPUs controls the world.
Those are two very different things, and the former should be a serious concern. If AI does indeed become a double-digit percentage of electricity usage as the AI labs themselves predict, then it becomes a significant contributor to emissions, full stop.
> - AI spreads misinformation and causes societal damage
> environmental like emissions and water usage
There is real crisis with warming this year, yes the plan to consume staggering anounts of energy and make environment worst in the process is valid criticism.
> AI is useless and it will take the economy with it because it is a bubble
If it turns out to be true, a lot of innocent people get hurt. Valid.
> AI spreads misinformation and causes societal damage
As valid as criticism of facebook was valid the whole time. And yes, facebook made world into worst place.
> AI is trained on copyright (are we really on this side of the debate ?!)
100% valid.
> safety critics who think AI can take over the world
Not valid, that is bullshit. If you think the word is wrong suggest polite word that says the same.
I'm sure I can make a lot of accurate predictions. It doesn't mean I'm worth listening to, if people can't distinguish the accurate from the inaccurate in the moment.
Reminds me of a tactic I've seen often amongst both critics and shills on fads. An OpenClaw fanatic on Youtube comes to mind. He makes opposing claims in different videos. One of them has to turn out to be true, and he trumpets his successes ("Look, I predicted this!"). Only a few notice he also predicted the opposite.
Just look at the other thread about Zitron and his Enron comparisons.
That's relevant because if you make a thousand predictions, a few of them might turn out to be true. That doesn't mean you're good at making accurate predictions.
1. Zitron claims model capability has peaked 2. Zitron claims AI lab growth (user and revenue) has stalled.
In the first case, TFA refutes by claiming 'wrong' repeatedly, which does not convince me of anything. If anything, Zitron is probably right in this regard, since the majority of progress in recent LLM tech has been setting up of guardrails to cajole the models using 'agents'.
In the second case, numbers are given showing growth, directly refuting Zitron. However, I'm giving Zitron the benefit of the doubt, given the old saying - market can remain irrational far longer than you can remain solvent. As long as people can be convinced that the sky is falling, rational predictions rarely pan out.
You don't have to like them, use them, or consider them "good enough", but the idea that models haven't gotten better in the last two years is ridiculous.
Is there any objective measure that shows this?
Would we use examples such as his Feb 2024 claim "I believe we're reaching the upper limits about what generative AI can do"?
> 2. Zitron claims AI lab growth (user and revenue) has stalled.… I'm giving Zitron the benefit of the doubt …
Is there some date by which you'd say it'd be fair to evaluate whether Z’s claims are true (without the benefit of the doubt)?
You mentioned revenue - would we use claims such as his 2024 claim that the companies no longer knew how to grow? But that in 2024, 2025, and 2026 both the companies revenues and profits have grown at double-digit rates each period?
You also mentioned users - would we use claims that "Sundar Pichai wants Gemini to be 'used by 500 million people before the end of 2025, 'a number so unrealistic that someone at Google should have been fired, and that someone is Sundar Pichai", where Gemini then hit 750 M users?
Or by what measures should we evaluate whether Z's claims are true?
I am certain that plugging a circa 2023 model into a 2026 harness would be a pretty frustrating experience. Yes, you could code a bit with AI in 2023, but models are just much better at it than they used to be. And smaller open models are leaps and bounds better at it than they were three years ago.