If you liked this piece, you should subscribe to my premium newsletter. It’s $70 a year, or $7 a month, and in return you get a weekly newsletter that’s usually anywhere from 5,000 to 18,000 words, including vast, detailed analyses of NVIDIA, Anthropic and OpenAI’s finances, and the AI bubble writ large. My Hater's Guides To Private Credit and Private Equity are essential to understanding our current financial system, and my guide to how OpenAI Kills Oracle pairs nicely with my Hater's Guide To Oracle.

This week, I’ll publish the final part of my ongoing series (“What If…We’re In An AI Bubble?”) about the factors and events that will cause the AI bubble to finally pop, focusing on what consequences might follow the collapse of OpenAI and the wider data center 

Subscribing to premium is both great value and makes it possible to write these large, deeply-researched free pieces every week. 

Today I’m going to speak from the heart, and tell you that we’re ruled by fucking imbeciles.

AI is a perfect storm of failed concepts and organizations, and the apex of the Era of the Business Idiot, an epoch where we’re ruled by people so thoroughly disconnected from the actual workforce that it was inevitable that a technology would be created specifically to grift them.

Just ask Aaron Levie, CEO of Box: 

CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI.

LLMs are dangerous for many, many reasons, but the under-discussed one is how well they play to a certain kind of executive imbecile. Generative AI is — to quote Mo Bitar — really good at doing an impression of work, much like most managers and c-suite executives, and even if it’s completely incapable of doing something, it’ll absolutely say it can and tell you you’re amazing for suggesting it.

And that’s why Business Idiots love it. 

Where regular human beings would say annoying things like “that’s not possible within that timeline” or “we don’t have the resources to do it,” AI will say “of course, right away!” and burn as many tokens as possible. When it makes mistakes, it’ll apologize — as it should because it failed you — but then promise to do better next time, all while costing so much less, at least in theory, than a regular, stinky human being. 

It’ll create a PRD (product requirements document) of a theoretical software project with the confidence and vigor that you need to take it immediately to a software engineer and say “build this immediately,” and when the software engineer tells you a bunch of bullshit about it not being possible, it’ll spit out several convincing-sounding responses. Fuck, why even bother talking to that engineer at all? Claude Code can mock up a prototype that you can then shove in their fucking face before you fire them for not using AI to do it themselves.

I realize I sound a little churlish and dismissive of those who may or may not actually get something out of AI, but this entire industry feels like a mixture of kayfabe and ignorance, slathered with a kind of angry desperation that reflects the distance between reality and fantasy, driven by people that don’t do any fucking work. 

Any executive-level fuckwit you’ve met in your life now has a seemingly-powerful tool that can burp up mimicry of open source software and, if you constantly prompt it, eventually get something half-functional onto some sort of web server. When you face bugs, it’ll try and fix them, sometimes also “fixing” (adding or deleting code) from elsewhere to be helpful, like when Cursor using Anthropic’s Claude Opus 4.6 model deleted an entire production database and all its backups. It will never, ever say no, even if it’s incapable, even if it has no thoughts, even if what you are asking is equal parts impossible and unreasonable in both its timescale and scope.

A Business Idiot, given his druthers, can sit there and fuck around and make an LLM spit out something that makes him feel like he’s coding, which in turn makes him feel that you, a lazy and stupid engineer, could do even more with the power of AI. It doesn’t matter that it costs an absolute shit-ton of money, or that there’s no way to measure its efficacy. The Lion does not concern himself with things like “efficacy” or “productivity,” and the Lion is increasingly tired of your whining! The Lion doesn’t even understand what it is you do every day other than not doing what The Lion is asking for!

You laugh, but this is genuinely how the majority of managers and executives think and act, and now they have a special chatbot that can fart out functional-enough prototypes to convince a Business Idiot they can do anything, because executives and managers do not regularly do much work. As a result, they have little idea what work looks like other than when they look over your shoulder, which is why they wanted you back in the office, and their distance from production is why the same people who were anti-remote work are now aggressively trying to shove AI down your throat. 

Organizations aren’t burning millions or hundreds of millions of dollars a year on AI because it’s good, they’re doing it because they are run by people who do not know what the fuck they’re doing. 

Generative AI is catnip for hall monitors, snitches, toadies, and any other group that hates work and loves talking down to others. Put another way, it ingratiates losers who believe that learning to do or being good at something is a waste of time, because they deserve to just do what they want without any of that messy “effort.” 

While I’m not saying every LLM user is an imbecile, they’re built to convince the mediocre and incurious that they’re remarkable, and it turns out that a great many of them run venture capital firms and Fortune 500 companies.

I also want to be clear that while there are sane and normal people who use these things, they’re mostly drowned out by a crowd of people that oscillate between bootlicking and regurgitating capitalist mythology in a way that makes it hard to trust anybody who spends significant amounts of time using an LLM. 

One thing you’ll notice about the most moistened AI boosters is that they lack much degree of pride in their work. Everything they say must, at some point, compliment the mindless, unprofitable, unreliable tool underneath it — how “incredibly powerful” it is, how it’s “only getting better,” how it’s “only the beginning” of something that’s eaten over a trillion dollars and absorbed the majority of venture capital. 

It isn’t about the work, or the craft, or the thought behind it. Everything is a numb, mindless death march toward saying “job done” and burping out some sort of pseudo product, if one even exists.

I’m not even being sarcastic! Per Bloomberg, Salesforce has been marketing “powerful AI products” that don’t actually exist:

Patients at the University of Chicago Medicine featured in a promotional video seamlessly refill prescriptions, book appointments and even get parking tips with the help of Agentforce, Salesforce Inc.’s flagship AI tool.

All this is possible, according to the video released in October, with the new technology. But the scenes are largely aspirational — little of that AI functionality is live. Patients who call the hospital system today are greeted with keypad-selection menus and routed to human schedulers. Its chatbot is still being tested and not visible to most web visitors.

The problem is that chunks of the capabilities shown in presentations – including from top customers like Williams-Sonoma Inc. and Finnair Oyj. – are actually mock-ups that aren’t yet widely in use.

In a rational society, Salesforce’s stock would take a beating and the SEC would open an immediate and brutal investigation. 

Sadly, our society is oriented around the power fantasies of the mediocre and spiritually-dead losers, people bereft of pride or joy in the things they create that believe that they’re owed everything. 

They’re Business Idiots, and they are your enemy. Even those who believe they’re aligned with the Business Idiots by supporting and using Large Language Models are the enemy, because The Business Idiots believe that “AI” will simply remove anybody else from the picture, automating work, creativity, communication, friendship, and that includes anyone that helped its ascent. 

And yet none of it’s really working, because Business Idiots don’t really know how anything works.

LLMs Are Built To Coddle Losers

As I said back in the original piece, think of The Business Idiot as a kind of con artist, except the con has become the standard way of doing business for an alarmingly large part of society.  Salesforce, one of the most-prominent hypesters behind the AI bubble has spent millions of dollars on advertising and marketing to promote a product that doesn’t exist in the way that it’s being sold. 

Sidenote: Now, I want to preface this by saying that not every LLM user is a loser. There are plenty of people bullied into using these models by their bosses, who use them as a glorified search engine, or who use them to write quick scripts or coding snippets to speed up their work. I think this is pretty clear in the copy, but I’m anticipating somebody reads that title and then stops reading entirely.

Only an economy oriented around coveting and coddling losers would have let AI get this far. Every single story about AI has to either directly gloss over the obvious financial and technological issues or start speaking in the kinds of vague theoreticals reserved for cults and multi-level marketing scams. Even Bloomberg’s piece — which is pretty critical! — helps gaslight Salesforce’s customers by quoting an executive blaming their own processes for Salesforce’s outright lies:

Successful deployment of more advanced uses of Agentforce is as much about a customers’ processes and internal compliance as it is about technology, said Madhav Thattai, Salesforce executive vice president of Agentforce. “You’re going to see customers start with simpler use cases,” he said. “As they start to develop and implement more complex things, they kind of realize this full autonomous vision.”

What the fuck does that mean? What’re you talking about, Madhav? What “autonomous vision”? What complex things? Do you even know? Hello?

Even in this very critical piece, the endless pursuit of “fairness” — the Business Idiot’s favourite weapon when they don’t want to be graded on their actual work — means that we have this slop-adjacent explainer that mostly amounts to “yeah you know sometimes their shit needs to be better and then one day, wow, boom! We’re gonna have all sorts of stuff happening.”

But this is the world the Business Idiots have created, as I described last year:

I, however, believe the problem runs a little deeper than the economy, which is a symptom of a bigger, virulent, and treatment-resistant plague that has infected the minds of those currently twigging at the levers of power — and really, the only levers that actually matter. 

The incentives behind effectively everything we do have been broken by decades of neoliberal thinking, where the idea of a company — an entity created to do a thing in exchange for money —has been drained of all meaning beyond the continued domination and extraction of everything around it, focusing heavily on short-term gains and growth at all costs. In doing so, the definition of a “good business” has changed from one that makes good products at a fair price to a sustainable and loyal market, to one that can display the most stock price growth from quarter to quarter. 

This is the Rot Economy, which is a useful description for how tech companies have voluntarily degraded their core products in order to placate shareholders, transforming useful — and sometimes beloved — services into a hollow shell of their former selves as a means of expressing growth. But it’s worth noting that this transformation isn’t constrained to the tech industry, nor was it a phenomena that occurred when the tech industry entered its current VC-fuelled, publicly-traded incarnation. 

This naturally created a tech industry (and a larger economy) dominated by executives that were rewarded for growth, which meant that our tech products are inherently oriented around that growth: 

When the leader of a company doesn't participate in or respect the production of the goods that enriches them, it creates a culture that enables similarly vacuous leaders on all levels. Management as a concept no longer means doing "work," but establishing cultures of dominance and value extraction. A CEO isn't measured on happy customers or even how good their revenue is today, but how good revenue might be tomorrow and whether those customers are paying them more. A "manager," much like a CEO, is no longer a position with any real responsibility — they're there to make sure you're working, to know enough about your work that they can sort of tell you what to do, but somehow the job of "telling you what to do" doesn't come with it any actual work, and the instructions don’t need to be useful or even meaningful.

The problem with an economy dominated by Business Idiots is that it eventually loses its connection to the wider concept of production or solutions to customers’ problems, because that might cause management to interact with the real world and, by extension, have actual problems themselves. The problems that Microsoft, Google, Meta and Amazon solve on a daily basis are those related to its shareholders. How do we keep growing? How do we keep people engaged with our products? How do we convince our customers to pay more for our customers? And how do we keep people buying our stock?

Thankfully, The Business Idiots have captured both the media and the markets, twisting the definition of a “good company” into one measured by these very same questions. It doesn’t matter that Facebook is deliberately broken or Google Search’s results were intentionally made worse because number go up, and that’s all The Business Idiot cares about! It doesn’t even matter that 10% of Meta’s 2024 revenue came from scams or that its Kylie Jenner-branded chatbot led a man with dementia to his death or that its John Cena-branded chatbot would roleplay about having sex with children or that it wants to spend $125 billion or more on AI in 2026 because Meta’s ad sales have yet to slow down. 

It doesn’t matter that Meta CTO Andrew “Boz” Bosworth has overseen multiple unprofitable, unpopular products or is hated by basically every single person I’ve ever talked to at Meta — The Wall Street Journal will still write a glowing profile saying he’s a “blunt, outspoken provocateur” that’s “transforming Meta” by “unleashing AI.” One can be a colossal fucking loser that everybody hates, lay off thousands of people, fail to make anything of note, oversee multiple failures, and the Business Idiot’s consent-manufacturing machine will help wall you off from reality. 

“But Ed,” I hear you cry. “You can’t call somebody like Andrew Bosworth a loser. He’s a huge success! He made lots of money!” You’re falling for the Business Idiot’s biggest trap: that having wealth or being a C-suite executive is proof that you’re not a disconnected loser. 

Boz, like every other oaf destroying your favourite tech products, is the ultimate loser — he’s succeeded by taking credit for other people’s ideas, firing people when his own ideas fail, and repeating the cycle as many times as he wants because that’s what being an executive means to him. Boz has no pride in his work. If he did, he’d have resigned over the failures of both the metaverse and Meta’s wasteful, directionless AI efforts, or even over how fucking awful Facebook has become.

The sad truth is that he doesn’t care! He doesn’t give a shit. Boz, like every other Business Idiot, exists to extract value from others and get rewarded by shareholders. As he said in 2018, to Boz, “all the work [Meta does] in growth is justified.” That includes deliberately making notifications less useful, injecting clickbait and AI slop into your feed, and hiding chronological feeds behind an Escher painting of different menu options. 

Boz is indicative of the vast majority of CEOs and upper-level management of most of the world’s organizations. If you read this and feel self-conscious, it’s because you secretly know I’m talking about you or somebody you know. One can be incredibly-rich and well-known and yet a huge, unbelievable loser, because being a loser is deep within your soul. A loser is somebody who takes from others, claims others’ work as their own, and demands more credit for having done so. A loser is somebody who believes work and creation is beneath them, and that they are owed the fruits of labor regardless of their actual contributions to the world.

This is why so many people have such an abnormal reaction to AI, promoting and defending it like it’s their religion or nation state. While many people use LLMs and see them as a kind of word calculator or search engine, so many more see within it the chance to ascend above the proles who “work” or “create,” because they find the process of labor or effort so utterly loathsome. When somebody badmouths AI, the Business Idiot must defend it with everything they have, because attacking LLMs is attacking the output of an LLM, which is in turn a judgment on those who are tolerant of its mediocrity and impossible-to-avoid hallucinations.

You see, if you demand good work with intention, that might mean the Business Idiot actually had to do something, and that’s not what The Business Idiot signed up for. We are slaves to middle management and the middle management mindset, we are living in their world, and it will collapse because they never really understood anything to begin with.

LLMs impress the writers who do not want to write, the coders who don’t want to code, the researchers who don’t want to research, and the lawyers that don’t want to actually understand case law. Those that desperately tell you how powerful AI is and that you simply must use it are looking for you to validate their own laziness or distaste for effort, and those who are impressed with LLMs’ outputs tend to be people with low standards. 

The aggression with which AI boosters and executives act toward those who aren’t impressed suggests a genuine intellectual and moral weakness. Nobody who’s this insistent, aggressive and violative with their language of “it’s here and if you don’t adopt it you’re stupid and dead” has ever been right about anything. Nobody this desperate, insistent and forceful has ever had good intentions, good vibes or brought good omens — they are always bearers of some kind of con. 

Most technology is sold on elevating and ascending human beings. AI cheapens every interaction by creating a work-shaped product from a person that doesn’t respect you enough to give you work that’s barely fit for a human because it wasn’t made for one. 

This is why being an AI booster requires you to debase yourself. You must accept becoming a dogshit dealer that loves accepting and receiving low quality goods. You must celebrate intentionless and decaying slop, and defend it and the machine that made it with your entire being. You must sully yourself — treat its unexceptional, sloppy and unreliable outputs as signs of sentience, or at least the proof that digital sentience is possible. You must defend horrible, abrasive, ugly, loud monoliths of steel full of $50,000 graphics cards. You must say they are necessary, and you must aggressively antagonize those who do not. 

That’s because they’re not defending LLMs so much as they are the greater form of Rot Economic capitalism. The Business Idiots have successfully changed our experience of buying and using software from one of “paying for a service” to “accessing powerful technology,” reframing every mistake as the necessary pain of new innovations and every mediocre output as proof that the tech industry can still innovate, because critiquing these things — asking for them to be anything approaching the autonomous, reliable and powerful technology that everybody claims they are — is considered “improper” or “biased” or “skeptical.” 

Oh yes, they use “skeptical” as a pejorative.

This aggression only proves that the management sect is scared. LLMs were meant to be the thing that replaced all workers, but the actual outputs and outcomes don’t seem to be resulting in anything changing other than lots of things becoming worse or more-expensive. Every AI booster will say “AI is writing all the code at some organizations,” but never seem to explain what happens as a result, such as whether software is being shipped faster, or better software is being made, or, well…anything. 

The answer is simple: they don’t know because nothing has actually changed. Organizations writing massive amounts of code using LLMs are facing massive product stability issues and, in the case of Zillow, spending millions of dollars a year to turn their codebase into a confusing, intentionless slop and increasing software reviewer loads by 29,000 hours a month. 

This is only made possible in an economy run by people who don’t do any work, and a tech and business media that exists to ingratiate them.

The AI Industry Is A Grifting Machine

I want to lead with a surprising comment: I don’t think LLMs, as a tool, are a grift. There are use cases, though those use cases are miniscule compared to the egregious promises and extrapolations made by the majority of the media and the executive sect, and absolutely nothing about them warrants the amount of money invested in them. 

That being said, I think LLMs lend themselves perfectly to grifting.

Sam Altman helped propagate a technology perfect for conning people with potential, a larger extrapolation of Altman’s own life of taking dogshit — Loopt, for example! — and parlaying it into larger opportunities. It can make a really half-hearted demo of a lot of things, and that’s good enough to sell to Business Idiot. 

Dario Amodei took this grift and perfected it. Anthropic is a company purpose-built to con people into giving it by money by making people feel smart. LLMs can do work-shaped stuff, sometimes, as long as you debase yourself to accept mediocre and often-broken stuff that you have to keep a vigilant eye on, and either use a subsided product that loses Anthropic money or pay a shit ton of money as an enterprise to Anthropic and it still loses money. 

The media was also primed for the grift. Reporters are never incentivized or supported to actually spend meaningful time understanding technology, meaning that the vast majority lean toward access journalism or, at best, the most kindly, “objective” (read: pro-business) takes that result in “wow, isn’t the future great?” no matter how good the thing they’re using actually is. Editors are, in many cases, entirely disconnected from the process of reporting or writing, let alone the underlying technology their reporters cover, which leads them to at best live in a world of “I sure don’t trust these CEOs but their technology sure is powerful.” 

As a result, all a technology has to do is either look or sound plausible. Can LLMs write all code? Not really! But because they can write some code and there are lots of eager people on Twitter saying it’s powerful, that’s all it takes to write the sentence “software engineers are writing most of their code using LLMs.” Can Anthropic actually take down Figma? God no, but the mere existence of Claude Design is enough to write that it might. All it takes is the hint of something to be true for it to be written about as gospel. Each statement adds another bullet point to Anthropic’s investor deck so that it can raise another $30 billion in funding, which in turn validates any journalist’s beliefs in Anthropic’s ability to destroy other companies with a product the journalist has not and never will use. 

Business Idiots did well to pressure modern journalism into conflating scrutiny with a lack of curiosity. To ask too many questions is “unfair.” To not immediately assume that LLMs are getting “exponentially better” is to be an ignorant luddite. To not assume that everything will work out like it did with Uber or Amazon Web Services is to “ignore history.” 

Grifters took advantage of this industrialized intellectual weakness using a tool purpose-built to do enough of an impression of something to impress the media and executives.

It worked, because both are sold to in much the same way — by telling a plausible-enough story that ingratiates somebody who is never the end user of the product in question. 

If a journalist gets curious, an LLM can make a good-enough impression of somebody writing software to fool somebody who doesn’t really know what they’re doing, and if you prompt it again and again and again, it can get something functional out the door. This is all it takes for somebody — a reporter or an executive — to extrapolate that because they were able to do something (even though the LLM did it), a subject-matter expert would be able to do even more. 

As a result, LLMs are fantastic tools for grifters. Somebody that doesn’t really like doing anything other than getting applause for other people’s work can now run multiple concurrent agents, endlessly tweak prompts and tell everybody that they’re an “AI specialist,” with their LLMs making them seem busy in a way that’s hard to argue with because there’s so much bullshit going on. 

An ethically-questionable “AI beat reporter” (though this is not across the board) can easily become prominent by simply writing up whatever it is the companies are excited about and reporting on leaks of Slack conversations, creating the appearance of “scrutiny” without ever scrutinizing or questioning the ethics or underlying economics. An oafish product manager with terrible ideas can now pump out half-functional scripts and software that sort of does something, and when their manager — somebody who also doesn’t do any work — sees what they’re doing, they can happily report to their manager that the person in question is “AI-first.”

And when you’ve oriented your entire economy around middle managers, vice presidents, and executives that don’t do any real work, this shit seems magical.

AI companies are natural grifting instruments. Because AI startups are so capital-intensive, they naturally require tons of money, which means that venture capitalists have something to invest in, and because there’re always so many rounds, valuations are constantly being pumped. Because AI models can be plugged into anything, by extension any AI founder can pretend that any industry can be automated using AI, and because venture capitalists don’t build stuff or really know stuff anymore, they’re naturally impressed by basically any demo or plausible-sounding promise, especially when an LLM can make something that looks like software. Because AI data centers are so capital-intensive, they require endless amounts of risky debt, but that risk allows private credit to take investments from insurance and pension funds desperate for yield, and because everybody involved is a Business Idiot, nobody has actually thought about what happens if these things don’t work out. 

AI allows everybody to grow as long as everybody ignores the big, obvious problems with its efficacy and underlying economics. All you have to do is keep up the kayfabe that the problems aren’t problems and the solutions are imminent, or if you want to pretend to be a critic, you can also suggest that all of this is inevitable. Don’t worry about the fact that data centers aren’t getting finished, or that OpenAI and Anthropic make up 75%+ of all AI compute capacity, or that they make up more than 50% of Amazon, Google and Microsoft’s revenue backlog, or that both of them are horrendously unprofitable outside of brazen accounting tricks that would only work on a business and tech media intent on believing everything they say. Don’t worry about it! Stop asking. Don’t worry about Claude deleting entire databases in seconds, they’re gonna fix that somehow, some day. 

That ignorance is a sign of laziness, and of the dominance of the Business Idiot mindset. Everybody wants this to be Uber (it isn’t) or Amazon Web Services (it isn’t) because it allows them to avoid learning stuff or making informed decisions. If it’s like Uber or Amazon, you can just throw your hands up and say “it’ll work itself out!” which is way, way harder than explaining to me how an industry that loses billions of dollars with no path to profitability doesn’t run out of money at some point. 

This is, again, part of the grifter’s toolkit. When you don’t want somebody to think about what’s actually happening, you point them toward something that ingratiates them. Somebody who is rude and mean and asking about those billions of dollars of losses is a hater — somebody who says “well, Amazon Web Services lost a lot of money!” is historically-aware and erudite, even if actual history tells you that Amazon Web Services cost around $50 billion before becoming profitable, or around a quarter of Amazon’s 2026 capex.

This isn’t to say everybody making this argument is lazy, just that they’re unwitting pawns in a larger grift where mythology is used to support the biggest waste of capital in human history. 

And really, that’s the larger LLM grift: encouraging people to accept or sell lazy, half-baked shortcuts instead of fundamental units of labor or production, all while making them feel smart for doing so. It is a technology that perfectly fits the grifter strategy of giving people as little proof as possible to prove something is real, then letting them fill in the blanks with whatever will make them feel like they’re “ahead,” even if being “ahead” means "mournfully accepting that their job might be automated.”

Yet I challenge you any time you hear somebody saying that “AI is here, and it’s transformative” to ask them what the fuck that means, because while “it” might be real, it’s unclear what they actually mean by “it.” The grifters want you to immediately start filling in the blanks, assuming that CEOs saying they’re laying people off because of AI aren’t blatantly lying and that AI has done something, somewhere, that remotely warrants any of this waste and endless propagandizing. 

And they want you to do that because they’re losing.

The AI Industry Is Losing

If you’ve ever been in a bad relationship — romantic or otherwise — you’ll know the feeling of trying to find any way to prove that things will improve, and the amount of times you’ve ignored something glaringly, obviously wrong. “They’re going through a lot,” “they don’t need to tell me what I need to hear, I know they feel it inside,” “they’re busy right now,” and every other rationalization of somebody not being good to you or interested in you is an exercise in self-deception to avoid dealing with an uncomfortable truth. 

Any time you’ve ever found yourself looking for shreds of proof that things are going well is the exact time you should be leaving somebody, yet you’ve likely stayed and sought them out like Sherlock Holmes before he spends thousands of dollars on therapy.

Every time you stick around a little longer, you do so based on increasingly-questionable data and the knowledge that changing course will require a brutal reckoning with reality. Sometimes you stick around forever, because making more bad decisions is sometimes harder than making one good yet difficult one.

People are making the same mistake with AI. 

Right now, everybody is ignoring many, many warning signs at once, all because of short-term thinking. Because hyperscalers’ existent businesses have yet to slow down, everybody assumes — without any actual proof — that AI is somehow driving growth. Conversely, nobody seems to have an answer as to how big tech makes the $2 trillion to $3 trillion of brand new revenue it needs to justify its trillions of dollars of planned capex, and even the Financial Times only sees Amazon making any kind of return on hyperscaler AI investment by 2030: 

And even here, in a piece called “the impossible maths of the AI boom,” The Financial Times deliberately finds a way to make things look better by removing every single operating cost! 

Those covering Anthropic’s so-called “profitable” second quarter are intentionally ignoring that Musk deliberately discounted the months of May and June in an obvious attempt to engineer a headline. They’re also ignoring the obvious mismatch between Anthropic CFO Krishna Rao’s sworn affidavit from March 6 2026, when he said it had “exceeding” $5 billion in lifetime revenue, which doesn’t line up with any of its previously-reported or stated annualized revenues. The answer, in the end, is that it’s just easier to ignore this stuff, because taking it seriously would require thinking about Anthropic in skeptical terms, which would, in turn, require you to start questioning the fundamental stability of the AI industry.

And they need you to do that because they’re fucking losing.

OpenAI had a negative 122% non-GAAP operating margin in Q1 2026, and ChatGPT growth has stalled. Despite its so-called profitability, Anthropic has had to raise a combined $75 billion (between Google, Amazon and investors) since the beginning of the year. Both OpenAI and Anthropic had to lower their gross margin projections at the end of 2025. 

Sidenote: As I get into in my coverage, Anthropic’s costs scale linearly with its revenues, and the only reason Q2 was profitable was said costs were artificially suppressed. 

Anthropic and OpenAI — neither of whom have any path to sustainability or profitability outside of accounting shenanigans and willing co-conspirators — now make up 50% of all upcoming hyperscaler revenues, and the only way either of them can pay is if somebody, either a venture capitalist or hyperscaler, chooses to give them the money. Nobody has an explanation as to how that works or who funds it, other than that “hyperscalers are some of the most-profitable, cash-rich companies in the world (as their cashflows drop to their lowest levels in history),” and that “both of these companies are growing incredibly fast.”

Anthropic’s growth is a direct result of Business Idiots controlling a large portion of our economy. Nobody — not a single company — has been able to express in clear-set terms based on their actual bottom line a conversion of “I spent this much money and got this in return.” 

In fact, it seems like the opposite is happening. As I’ll mention in greater depth later, Andrew Macdonald, the Chief Operating Officer of Uber, recently gave an interview where he said that the company’s ballooning AI costs are “harder to justify,” in part because there’s no way to link its token spend to useful new features. 

Everybody spending millions of dollars on AI tokens is experimenting. As I’ve discussed previously, nobody really knows how to measure the ROI of AI, and the naturally-chaotic nature of LLMs makes it impossible to measure how much it might even cost:

To be very clear about what I mean, I think there is currently an AI token binge across both Anthropic and OpenAI. Enterprises do not know the actual value of AI, and do not know how much they should actually be budgeting, which is why Uber and others are running through their token budgets but not, it seems, spending less. We’re currently in an abundance phase — one where nobody is truly thinking about the costs outside of their fear of missing out — but there’s this nasty undercurrent of “wait, how much does that cost?” followed by “oh, fuck, well…you know I love AI but…”

Marc Benioff isn’t spending $300 million a year on Anthropic tokens because it’s doing something. He’s doing it because he, like every Business Idiot, has no idea what to do other than spend money, hire people, or fire people. Spending lots of money on AI allows him to say that Salesforce is an “AI-first organization,” and then blatantly lie for two years running that he’s “not hiring any more engineers” despite the many, many job listings on Salesforce’s website for engineering positions. 

It’s kayfabe that exists to distract you from the fact that Agentforce only has $800 million in annualized revenue, or around $66 million a month for a company that makes $11 billion or more a quarter. 

Seriously, somebody please show me a company spending millions of dollars on AI tokens that can also express a clear, indisputable return on investment. Show me the actual returns. Show me the processes automated and what those processes being automated do to offset these remarkable costs. All of this fucking bloviating about how AI is inevitable and real and so powerful never seems to result in a profit. While companies can vaguely say “oh we saved X number of hours from this,” I am still waiting for somebody to say “we saved this much money and this is how investing in these tokens is profitable.”

It’s always something vague, like when Klarna said it estimated ChatGPT would “drive $40 million in profit improvement in 2024,” a stat that it never explained or returned to. Klarna CEO Sebastian Siemiatowski once told Sam Altman to use Klarna “as his guinea pig” — only to have to hire back the humans it tried to replace with LLMs after a massive wave of customer complaints. Klarna once said that its chatbots did the work of 700 people, a blatant lie that it got away with because the media doesn’t want to scrutinize an era built on deception.

That’s because underneath the puffery and the propaganda and the pervasive sense of inevitability, the AI industry is losing. Anthropic and OpenAI’s revenue growth is only possible thanks to a near-perpetual misinformation campaign that overstates both the current and future capabilities of LLMs, and a near-society-wide ignorance at the executive level. Every story about Anthropic’s customers burning millions of dollars’ worth of tokens comes back to one unfortunate fact: nobody knows how much it’s costing but whatever it costs today isn’t sustainable. 

For example, and as I mentioned earlier, Uber COO Andrew Macdonald said this weekend that it was becoming “harder to justify AI costs within the company”:

"That link is not there yet, right?" he said. "I think maybe implicitly there is more that is getting shipped, but it's very hard to draw a line between one of those stats and, 'Okay, now we're actually producing 25% more useful consumer features.’

Macdonald added that AI can seem free if you're "just a user sitting there coming up with interesting use cases" without paying for it. But ultimately, the company foots the bill.

I believe that Uber’s experience is indicative of effectively every company’s experience with AI. Business Idiots, disconnected entirely from production, demand their workers burn as many tokens as possible, incentivizing them to do so for reasons that only make sense to somebody who doesn’t do any work. 

And burn as many tokens as they could, Uber’s engineers did. Four months into the year, Uber had exhausted its entire AI budget — in part because it created a leaderboard of the biggest AI users, giving employees an incentive to run wasteful tasks and prompts, if not for bragging rights, then at least to show the higher-ups that they’re onboard with the new direction. . 

AI is meant to be this ultra-powerful streamlining tool that changes the workplace forever, yet the practical result appears to be “we’ve spent a bunch of money on something that makes our least-sentient managers excited.”

Too many members of the media work overtime to find ways to either ignore or explain away these problems. Stories about how Anthropic and OpenAI have agreed to a combined $1.048 trillion in compute commitments fail once to ask how they might get that money, other than to suggest that both may become cash flow positive by either 2028 or 2030, again with no discussion as to how other than “they will.” 

For them to do so, they will need…well, a trillion dollars over the next four years, either through revenue or funding. That’s an insane amount of money — more than any startup or even public company has had to raise in history — and the fact that more people aren’t talking about that suggests that they either don’t care or don’t want to. 

The same goes for those covering NVIDIA and other semiconductor companies. While the largest company on the stock market once again beat analyst expectations and raised guidance, few (other than JustDario, it seems) noticed that despite all that extra revenue, NVIDIA only saw its cash and equivalents grow by $600 million quarter-over-quarter. 

Why? Because it’s investing tens of billions of dollars investing in AI data center companies like IREN, CoreWeave, and, of course, both Anthropic and OpenAI, and has agreed to spend an unbelievable $30 billion on cloud service agreements in the next six years, quite literally paying its customers to buy its products in the most blatant circular financing since the dot com bubble.

This is what an industry does when it’s in distinct, existential distress. NVIDIA is now the fifth-largest purchaser of AI compute behind OpenAI, Anthropic, Microsoft and Meta at a time when AI compute is meant to be facing a supply crunch, which suggests that while demand may exist on a low level for those trying to pick up a few hundred H100s, the only customers for data centers full of Blackwell GPUs (at least, those that actually exist) are Anthropic, OpenAI, an organization with no clear AI strategy and a CEO that can never be fired, and the company selling the GPUs.

That’s a big fucking problem considering that there are tens of gigawatts of data centers being developed that will require around $380 billion in annual revenue to substantiate.

Data Center Deals Are Catnip For Dullards, Taking Advantage of The World’s Dumb, Desperate Money

There is, at this point, little proof that the AI data center “boom” is anything other than the largest real estate speculation in history. 

Some will point to the difficulty one might have finding GPUs, carefully ignoring how the majority of capacity is taken up by OpenAI and Anthropic, leaving the vast majority of customers to fight for scraps thanks to the extremely slow pace of data center construction. Others will say that guidance from companies like NVIDIA and Samsung prove that “the demand is there.”

Forgive me, I’m going to be a little stern.

• That “demand” is almost-entirely funded by debt.
• That “demand” is not an extrapolation of demand for AI services, but for debt’s hunger to invest in something private credit and banks believe in.
• Private credit and banks believe in data centers based on flawed maths and magical thinking, because they too are run by Business Idiots.
• AI data centers are themselves a grift, convincing investors that they’re backing large infrastructure projects akin to housing developments or factories, rather than warehouses full of expensive hardware for customers that may or may not actually exist.

I know, I know, you’re gonna say “Ed, you can’t paint with such a broad brush!” but I can find no data center debt deal that makes me feel like anybody was really thinking too hard when they put it together. Blue Owl agreed to invest up to $10 billion in Stargate Abilene after a single fifteen-minute conversation, despite the only tenant being OpenAI, a company that couldn’t afford to pay for the compute it committed to, and nobody ever having built a gigawatt-scale data center in history. This was likely because Blue Owl took advantage of the Business Idiots who run Crusoe:

To protect its investment in Abilene, Blue Owl negotiated a completion guarantee that meant it wouldn’t be responsible for finishing the data center should Crusoe go bankrupt. It also got regular payments on its investment during the construction phase, according to a Blue Owl marketing document. (More typically, a firm in Blue Owl’s position would need to wait until the data center is up and running to get paid.)

This is, to be clear, a huge scam, and something that should’ve horrified investors, except said investors are also Business Idiots that saw a big number and said “whoopie!” Money men with little connection to how long stuff takes to build, let alone the underlying technology being sold or the companies that might actually pay for the compute saw the potential to “back the next industrial revolution” and fell over themselves to take part. 

Like every greedy dullard, Business Idiots backing data centers are easily won over by the blatant lie that a data center is an “AI factory,” conjuring up images of large buildings that print money with little human labor needed. In reality, data centers are vast, labor-intensive construction projects connected to large, labor-intensive power projects, filled with GPUs that are so expensive that they require billions in debt that are upgraded on a yearly cycle, with customers that may or may not exist by the time it actually turns on. Calling them “AI factories” is an intentional attempt to simplify projects that have more in common with building cities than any kind of modern factory.

These Business Idiots are too informed by other Business Idiots, like the sell-side and buy-side analysts that have no interest in talking about what might happen in the distant future when they can conjure up plausible-sounding statements that pump their bags. Every single buy and sell-side analyst should have said CoreWeave, IREN, and other NVIDIA-backed neoclouds are dangerous investments fueled by circular finance. Instead, almost every single one has upgraded them as a result of NVIDIA’s continued investments, despite these investments being a sign that these businesses can’t last. 

The few hedge funds and private equity firms I speak to that have any kind of mental clarity are facing pressure from investors misinformed by analysts and the media. Hundreds of billions of dollars — at least $178.5 billion in America alone in 2025 — have been sunk into data center construction based on flawed information, astronomically more flawed than the assumptions that led to the dot com bubble bursting, as I covered in my premium piece from a few months ago. This is like if they built out all that dark fiber for what would amount to a few hundred internet users in 10 years.

These people see NVIDIA’s continued revenue growth as a sign that “demand is unstoppable,” yet that “demand” is entirely contingent on how long investors are willing to ignore reality, much like the rest of the AI industry can only continue as long as everybody keeps up the kayfabe of its supposed inevitability.

It’s time to stop, and force these failsons to stand on their own two feet. 

Stop Coming Up With Rationalizations For The AI Industry

I’m growing tired of the amount of people I read saying that “AI is real, but the economics are irrational,” as if these facts are entirely divorced from one another. 

A GB200 NVL72 rack will be just as expensive to run in 2030 as it is today, and an incomplete data center will still take just as many hundreds of millions or billions to finish in the future too. There are, I believe, at least $200 billion worth of data centers that will never make even a quarter of their costs back before collapsing, and that’s assuming that they ever turn on or their customers exist by the time they do so. 

AI is only “real” because everybody is willing to ignore its blatantly-obvious problems. The only reason that every app has an AI feature or every AI company can still sell a $20, $100, or $200-a-month subscription is because venture capital has yet to walk away from an industry that relies on eternal subsidies. AI data centers only continue to have revenue as long as venture capital and hyperscalers support Anthropic and OpenAI, and their revenues only continue to grow as a result of an endless, society-wide media campaign built on misinformation and API revenues driven by unsustainable venture-backed startups and businesses run by people excited to blow millions of dollars for no reason. 

AI is only as “real” as the excuses that get made for it, and the amount of money those who subsidize it are willing to lose. Venture capitalist subsidies are the only reason that companies like Perplexity or Lovable are alive, which in turn means that a large chunk of both Anthropic and OpenAI’s API revenue is only made possible through those subsidies. 

Demand for data centers is, by extension, only as large as these subsidies can sustain.

Much of this is substantiated by the myth of executive intelligence. Most assume that Sundar Pichai, Satya Nadella and Andy Jassy wouldn’t be as stupid as to burn a trillion or more dollars on data centers for an unprofitable product with demand that only exists because of their own subsidies…except that’s exactly what’s happening. These men have no other hypergrowth ideas, and are more willing to annihilate their cashflows and dominate the Earth with half-finished data centers than to admit that their core businesses will eventually decline.

And because these hyperscalers were so aggressive with their buildouts, the Business Idiots conflated that hunger with some sort of proof of massive demand for AI. 

Yet even NVIDIA’s own earnings show that demand is incredibly-centralized, with 54% of its Q1 FY27 revenue ($44 billion out of $81.6 billion) coming from three customers, up from two customers accounting for 30% ($13.2 billion) in Q1 FY26. I assume one or more of these are hyperscalers, which means that NVIDIA’s continued growth hinges heavily on the idea that big tech will continue to dump trillions of dollars into its GPUs in perpetuity.

I’m repeating myself, but this is not what a healthy industry looks like. If AI data center demand were evenly-distributed and sustainable, NVIDIA’s revenue wouldn’t depend mostly on three customers. Similarly, the entire media wouldn’t be loudly ignoring a short seller report that suggests that 20% of its FY26 revenue came from illegal sales to China.

We Are Going To Win

As I’ve said, AI is only as real as its subsidies. ChatGPT is only free to hundreds of millions of people because OpenAI is able to raise hundreds of billions of dollars, much like Anthropic is only able to subsidize its subscribers anywhere from $8 to $13.50 for every dollar of revenue because of endless venture welfare. 

The underlying economics suggest that no subscription-based AI service — let alone a free one — makes any kind of financial sense, and the only reason that everybody has had such unrestrained access is because the media and the markets approved it, and the people with the money are deluded and disconnected from the process of value creation on almost every imaginable level.

Any statements around “Anthropic actually being profitable on inference” are products of fantasy and magical thinking, distilled copium for people that would rather delude themselves into believing that none of this ever made sense. Again, the assumption is that “companies would never just burn a lot of money,” but that too is catering to the greater myth of executive competence, something that nobody who spends any amount of time around managers or executives would ever believe. 

GitHub Copilot let people burn thousands of dollars on a $39-a-month subscription as a means of expressing growth. I absolutely, 100% believe that both OpenAI and Anthropic are doing the same, and that neither of them has some magical way of making inference cheap enough to justify letting people burn thousands of dollars on a $100-a-month or $200-a-month subscription. To give them the benefit of the doubt is to empower them to continue to raise money by conning their investors and the general public, and to continue perpetuating an era of software that runs contrary to what makes technology good.

Their goal is simple: to ram as much of this through to as many people as possible to get them to spend as much money as possible…until they work out a way to make OpenAI or Anthropic or these endless data centers into something approaching a real business. One of the greatest mistakes we can make in our lives is to assume that the rich and powerful have any idea what the future holds, or that they have any grand plan or strategy. 

It’s very likely that Dario Amodei and Sam Altman’s plan is to keep burning money until somebody who works for them comes up with a way not to, and in the interim their plan is to get as many users as they can to keep raising money. 

Similarly, Microsoft, Google, Meta and Amazon’s plan is to keep building data centers in the hope that they’ll have a reason to use them by the time they’re built. There is no other plan. They do not have a secret invention coming. They do not have AGI in a box in their office. They do not have anything, and the reason they’re spending so much money and shoving AI into everything you use is because they have no fucking clue what to do.

This is why Dario Amodei makes wild claims about AI replacing 50% of all white collar workers or Microsoft AI CEO Mustafa Suleyman claims all white collar labor will be automated in 18 months — because the actual products themselves aren’t impressive enough to win you over or justify the hundreds of billions of dollars being sunk into AI. They say these things to make you think that they have a scary and powerful technology behind the scenes that does not exist. And yes, that includes Mythos.

The forceful, harassment-grade incursion of AI services into our daily lives is not a sign of its power, but a gesture of the lack of confidence and fear in the hearts of its progenitors. Good shit sells by telling you why it’s good — dodgy shit sells by tricking and scaring you and taking advantage of Business Idiots who think that using an LLM to type emails and spending 12 hours a day on Twitter constitutes “work.” 

I believe the vast majority of these data centers go unused and/or unfinished, and that most AI startups will die once the venture capital subsidies dry up. I believe that neither OpenAI nor Anthropic have a future, and that their revenues are only made possible through venture subsidies for startups using their models and the experimental revenue of Business Idiots that don’t really know why they’re “doing AI” in the first place. 

AI demand remains a result of a societal psychosis and a weakness in those who are meant to scrutinize the untrustworthy.

Its unraveling will be framed as impossible to see coming because nobody in power had bothered to look. 

It’s easy to feel hopeless. We’re at a point where the greed and the shamelessness and the stupidity is at a fever pitch. 

We’ve reached a time the mask has started to slip, and the C-suite imbecile class is unabashed about its loathing of people, as was the case when the CEO of UK bank Standard Chartered CEO, Bill Winters, talked about how those at risk of losing their jobs to AI are “lower-value human capital,” at the same event where he said the company would likely shed nearly 8,000 roles in the coming years due to AI. 

Winters would later apologize for his choice of words — although, to be clear, he was being absolutely honest when he made those remarks. That is what he believes.  

Everything feels rough because the AI industry is equal parts desperate and over-confident. AI executives believe that they can cram enough promises of money into the system that the system would rather cannibalize itself than admit that it made a mistake. Sundar Pichai, Andy Jassy, Larry Ellison, Elon Musk, Mark Zuckerberg and Satya Nadella will gladly annihilate hundreds of billions of dollars to avoid the inevitable, but once they do, it’ll be gruesome. 

At the same time, the things that they need to happen — actual profitability, actual returns on investment, actual tangible proof that this is a real thing rather than something they all have to actively conspire to keep alive — aren’t happening at all.

Each week, we hear about new AI megaprojects that will dominate our countryside with blinding lights, endless noise and fume-belching gas turbines at such a scale that it feels impossible it could ever stop. The system is absolutely going to try and exhaust itself to keep it going. The government bought $9 billion of Blackwell GPUs, which, to be clear, isn’t a “Too Big To Fail“ moment so much as it’s a way to keep NVIDIA’s plates spinning for another quarter.

In truth, the amount of money that NVIDIA needs to keep this going is so extreme that it is now a test of how long the debt markets and the hyperscalers can keep sustaining the hype. A trillion dollars in annual revenue is necessary by the end of 2028, which would require over 30GW of data center capacity to be built by then at a time when only 5GW at most appears to be under construction. 

Nevertheless, even the sweatiest, least-trustworthy boosters have begun to sneak in statements about “we’re probably not in a bubble,“ or “yeah it’s a bubble, but it’s a good bubble.” Jeff Bezos, when asked about the AI bubble, said that you “shouldn’t worry about it,” which…is not really sufficient, is it Jeff? 

None of this is to say that the mood is good! The vibes are disastrous. Everybody is exhausted. Those who love AI vibrate with a strange soullessness, constantly talking about the incredible power of AI without ever showing what it did or, perhaps, what all that supposed saved time got them. It sucks to work at basically any hyperscaler right now. 

Basically every person in every job has had somebody intimate they’re going to lose their job to a computer every time they open the newspaper or use a website, and every app has some sort of desperate, vulgar pop-up about a feature that will generate some bullshit, obfuscating the features you actually want to use in favor of those that might lose the company money, because the company has to prove to the people that invested in the company that they’re “futuristic.” Alternatively, their CEO has either mild or severe AI psychosis to the point that they have decided to violate your user experience. AI is a non-consensual technology at its heart.

But they are losing. They all know it. They are acting desperate. 

It seems that there are nearly as many announcements of new large data center developments as there are cancellations of said data center projects. While hyperscalers can dismiss that as a simple reallocation of capital, and nothing to worry about, it’s harder to ignore the growing backlash against these facilities from locals — and the success that locals have achieved in blocking (mostly temporary, but some permanently) any future developments. 

And it gets worse. Anthropic had to conspire with Elon fucking Musk to conjure up a single profitable quarter to swindle the media and its investors one last time. In response, OpenAI either leaked or had leaked immediately following that it had a negative 120% margin and ChatGPT growth had stalled. Anthropic is either the single-most successful grifter of all time or speed-running a con where it fudges together numbers to raise endless amounts of money to keep its billion-dollar burn going. 

These are not the actions of honest, sustainable companies that will exist in the future.

I believe that we are on course for a truly horrible crash, the likes of which may rewrite the venture capital industry and mortally wound one or more hyperscalers, as well as fundamentally divide society on so many levels into those that fell for this and those that did not. This will, in the short term, be absolutely fucking horrible for our markets and our wider economy as a result of the time-bomb of private credit and private equity. In the long term, I see it as a “They Live” moment for many millions of secret imbeciles and cretins in our midst, and I don’t think it’ll be easy to wash the stench off for those that really pledged themselves to the graveyard smash here.

We will win, long term. What they are doing is not working. The future will not be without pain, nor will it be easy, or pleasant, or something I relish in. But in the long term I think this is a moment where the greater Business Idiot incursion faces a reckoning with a reality it believed it could change through sheer force of will.

These people don’t know how to build things that work anymore, and thus the only thing they can do is spend money and fire people. They believe in nothing other than growth, and one cannot exist on belief and hype alone, at least not forever.

And I can’t wait to watch what happens when it collapses.

I’ll close this piece with the regular CTA — please, subscribe to my premium newsletter ($7 a month, $70 a year, you’re gonna love How OpenAI Kills Oracle and The Hater’s Guide To Private Credit — but with a little explanation as to why I do the things I do.

I write this newsletter to hopefully do three things:

• First, to tell you that the Business Idiot class wants you to doubt yourself, because whether you recognize it or not, they’re engaged in acts of information warfare against you.
• Second, to remind you that facts are facts, and numbers are numbers, and that no amount of puffery or obfuscation can change pure mathematical reality. The AI bubble is exactly that, a bubble, and like all bubbles, it will eventually pop.
• Third, to remind you what it is we’re fighting for. Because every newsletter I write isn’t simply about highlighting mathematical stupidity, or corruption, or dishonesty.

I do it because I believe, fundamentally, that these people — Altman, Amodei, Nadella, and the many, many other villains that I’ve mentioned in these pages — are bad people, and their values are the antithesis of my values. I care about people, and humanity, and truth, and they do not. 

I deeply love technology, and feel it made me the person I am today. It allows me to do wonderful things, connect with wonderful people, and discover endless troves of incredible information. The computer is marvelous. The computer has done many wonderful things for me, despite what all the Business Idiots say.

I see LLMs as a violation of everything that great computing stands for. The AI industry encourages its users to both accept and present low-quality work and demands that they constantly defend the industry from those who would demand better from it. It is inefficient, power-intensive, environmentally destructive, and inherently sold based on things that it might do, providing far more value to scam artists and con men than it does to its end users. 

This is a mask-off moment for both the ruling class and those captured by capital, and an opportunity to look around you and see who is most-easily fooled.

No industry of value needs to mislead you or make you feel bad for not adopting their technology. No trustworthy individual will ever see the need to humiliate or attack somebody for being insufficiently excited about a product. No CEO that talks of a theoretical future as a means of selling you software in the present should be trusted. No technology that makes mistakes with regularity should be defended.

And no industry that demands everything from us — our land, our energy, our water, our jobs, our art, our writing, our attention and every dollar we have — should ever be treated with anything but revulsion.