Your three turns (layoff -> spending cuts -> revenue decline) describe the demand side precisely. I'd add a fourth that's less visible: knowledge destruction.
When companies eliminate workers, they don't just lose spending power - they lose the accumulated understanding of how things work. Boeing has been rehiring retirees and contractors to recover capabilities its workforce no longer fully holds. TSMC is discovering it now - $40B factory in Arizona, but American engineers need a year of training in Taiwan because the knowledge walked out with the people a generation ago.
And there's an irony in the "80% capability" claim that keeps appearing everywhere. Even if true - where does the remaining 20% live? In the cognitive layer that only humans carry. The one being eliminated. We're trying to fix deindustrialization by pouring gasoline on it: solving the loss of human knowledge by removing more humans. Which means even the optimistic "people will retrain" argument has a problem it hasn't reckoned with: the capacity to retrain is itself being eroded by the same system.
The dead economy isn't just one where nobody can afford to buy. It's one where nobody remembers how to make. And that loss doesn't show up in any quarterly metric - which is exactly why it keeps happening.
Your Camus point lands. The person standing in front of you is not an input to a utility function - and what they know cannot be extracted from them without extracting them.
Absolutely correct. The knowledge loss would be enormous if this goes forward in the direction I fear it might. This is one of the things I’ve been talking about in my field (law): if we use AI to sub in for associate labor, at some point there are no senior experts. Biology is undefeated.
This extends to everything. Even now, it took a huge search to find craftsmen able to repair Notre Dame after the fire—because we just don’t build that way any longer. Now expand that to music, writing, chip-fabrication, and on and on…
Actually, the craftsmen were available because there are conservatories to keep the skills alive, such as Compagnons who tour France for several years to acquite crafts. But ignoring the skills acquired was a major trend even before ai
this is why I'm seeing it as the last industrial revolution, after this the industrialist making a something factory and offering jobs... is unmasked as a way the world can no longer work.
Time for a real structural revolution and post-industrial economics.
There are many innovative people among us. We only need a shot and the ability to compete. It's been hard for anyone to tend with the conglomerates, but now that more people are looking for alternatives, this is small business and individual entrepreneur time to shine. If there is a will, there is a way.
I am currently in the process of creating a community center with the hopes of eventually being the go to grocer, daycare, barber, tailor, secondhand store, etc. Long term goals, but I'm going to work towards it because what else can I do. And if enough of us do this, we can create networks of goods for us and by us and barter them and eliminate money exchange and taxation and whatever else.
I am one person. Even 5 of me can make a difference.
When I read what you write after I write what I wrote, I feel like we’re thinking the same thoughts but you’re reading better books. Thank you for both making me feel like I’m not crazy and educating me to think better.
We’re definitely circling the same set of ideas—but I’d challenge the assertion that I’m reading better books. Any piece that pulls in Peirce is working at a pretty, pretty high level.
This is informative post.. I’d like to add that the financial structure behind the AI boom is a carbon copy of 2008, with private equity using SPVs to hide over $120 billion in debt and pooling it into asset backed securities, essentially mortgage fraud with NVIDIA chips.
The narrative that AI can replace workers is a lie; studies from MIT and Anthropic, plus real-world experience, reveal a massive gap between the pitch and reality. Meanwhile, the political endgame is already on paper. Peter Thiel’s recent manifesto and Palantir's operations reveal a shift toward control through force, where domestic surveillance and foreign warfare are the same product.
The "dead economy" is just the cover.
With 80% of Americans pessimistic and elites preparing for a post consent society, Organizing isn't just a .
Slogan It is the only option we have left. Thanks for writing this.
Yeah, I think the 2008 reference point is more salient than the Dot Com bust that is too often invoked. First, for the economic reasons you lay out. I'm just waiting to see synthetic ABS start spinning out of this shit. When I was first out of law school, I clerked and saw Floyd Abrams—Floyd fuckin' Abrams, who argued the Pentagon Papers case—argue that Moody's couldn't be held liable for its credit ratings because they were FREE SPEECH. I hope he needed that money thanks to a divorce or a sick family member or something, because that's dirty shit.
Anyway, the dot com bust is inapposite because the buildup of internet infra (fiber optic broadband, etc) had intrinsic and long-lasting value. We're still using it. The hundreds of billions we're pouring into data centers are DOA. The chips will either become obsolete or burn out, and we'll have to replace them on a cycle. We won't even get the benefit of the lasting infrastructure from this rutting festival of death.
That's assuming we get a financial crisis*--which would be a godsend. My worry is we don't get one and the nightmare future imagined in this piece comes to be.
*Should such a fortuitous event happen, there must be a massive popular movement to prevent any QE or bailouts of the financial system. It must be allowed to collapse.
Your assessment of the infrastructure build is not quite true. The data centers themselves, the buildings, the power infrastructure, the networking, the cooling, the racks, all have long lasting value, even if different generations of chips get cycled through over time. I worked on the telecom buildout and there was plenty burned through there, too. The Lucent and Nortel 5E switches, the DSL equipment, point to point wireless networks, the different generations of mobile wireless standards...all gone now, even as we’re left with exponentially more bandwidth capacity than we had then.
The physical infrastructure has durable value independent of the chip cycle, and the telecom analogy is well taken. But I'd push back on the scale comparison. The top five hyperscalers are on track to spend somewhere north of $600 billion in capex this year alone, with roughly three-quarters of that directed at AI infrastructure. That dwarfs the telecom buildout's peak years by an order of magnitude. The question isn't whether data centers are worthless; they aren't, but whether the revenue to justify investment at this scale actually materializes. If it doesn't, we're not talking about some investors holding the bag. AI-related investment accounted for 39% of US GDP growth in the first three quarters of 2025. A collapse at this scale is a severe recession at minimum, and possibly a global depression.
Sure, but again, that’s investment. Long term capital going against an uncertain long term outcome. All big investment cycles eventually have their busts. It’s been 19 years since the last big one, so yeah, we’re due. Good to be prepared, but it happens.
I don’t mean to be cavalier, but I don’t exactly have hope that a super wise government would know what to do to prevent it. Suppress the AI boom in the US and then China owns it. Using fiscal policy and the Fed to manage it would just push the bubble elsewhere, even if the people in Washington were way wiser than they actually are.
I actually think AI will be a great economic development and do not agree with the dead economy theory. The population isn’t really growing, so AI automation will be needed to make workers more productive. The history of economics says that people will create new jobs even as old ones go away, and that should be the assumption until proven otherwise.
Btw, everyone knows Block’s announcement was BS. They overhired during the pandemic boom and they used AI as an excuse to cover up their bad management. Elon had already done the same thing at Dorsey’s other company (Twitter). He was a spendthrift.
All fair points, and I appreciate the good faith engagement. Please take my comments as a way to think through these issues as opposed to an argument.
Busts are normal, but scale matters. The telecom bust was roughly $2 trillion in lost market value and contributed to a mild recession. If the AI buildout is running at $600+ billion in annual capex and it unwinds, the exposure is structurally different. "It happens" is true but undersells the systemic risk when AI investment accounts for nearly 40% of GDP growth. The underlying economy is not doing well, and an AI crash could trigger contagion. Normal economics, when bubbles happen. Still messy.
On government: there's a slippage in your framing I want to flag. You start by questioning the idea that government could manage any of this, then slide to "suppress the AI boom" as though regulation and suppression are the same thing. They aren't, or at least needn't be. The question is whether we can have oversight, disclosure, antitrust enforcement, and labor protections without killing the industry. We've managed this in pharmaceuticals, aviation, and nuclear energy, all domains where regulation coexists with American dominance. The idea that any guardrails at all equals suppression is a posture taken by the VC and startup class, and I say that as a startup lawyer who loves my people. But there are priors baked into that world that are invisible from the inside—chief among them that private enterprise self-corrects and government can only make things worse.
Same framing move with China. "We can't regulate because China won't" assumes that any oversight means ceding the entire field. There's little to support that. And it's worth noting that China is actually taking a different approach: leaner, open models that run efficiently and do impressive work locally, rather than the closed hyperscaler model we're pouring $600 billion a year into. If the competition argument is supposed to justify our approach, it helps to notice that the competitor isn't even running the same race. The framing conveniently serves the interests of the companies arguing against oversight.
I think you're right on Block, and that's actually worse. If companies are already using AI as cover for layoffs that have nothing to do with AI capability (and many are papering over the post-shutdown hiring spree with "AI layoffs"), the narrative is doing displacement work independent of the technology. The story becomes self-fulfilling.
The 2008 comparison is far more accurate than the dot‑ com analogy. Dot‑com left behind durable infrastructure ,fiber, broadband, backbone networks ,that we still use today. The AI data center boom leaves behind consumable hardware with a 3 year lifespan, financed over 7 years, losing 70% of its value by the time the debt comes due.
The same players from 2008 are back: investment banks, private credit funds, and securitization desks repackaging GPU leases into structured products. When synthetic ABS starts spinning out of this stuff, we’ll know exactly where we are in the cycle.
What makes this dangerous is concentration. The companies driving this boom represent 30% of SPY and 40% of QQQ. If one sneezes, the entire market gets sick. Retail investors and pension funds will be left holding the bag ,again.
This is the fastest, largest, and most leveraged capital project in U.S. history, and unlike past build outs, it leaves behind no lasting infrastructure. Just depreciating chips and amortized debt.
Then you had all the other negative variables at play.The market is delusional but can stay that way for a long time.
Appreciate the engagement, too. I’m not absolutist about letting the free market run at all times. I just don’t particularly trust the people doing the regulating any more than I trust the ones doing the building. It’s hard to know the future, so regulators making guesses isn’t necessarily any better than investors making guesses. I would hold up the EU as a pretty good example of where the regulators have been guessing wrong and have basically kept their economies out of the tech era. The GDP per capita of the leading countries in Europe are equivalent to Mississippi or West Virginia (at best).
The rest of the economy is weak because we’re still unwinding the excess of the pandemic era boom. Yes we technically avoided recession but many sectors have been dragging since then and high interest rates drag on housing and durable goods. The AI buildout is a high percentage of a low GDP growth number. Also the economy is twice as big as it was during the telecom and housing booms (which should be looked at together as the set up for the Great Recession) so not sure what we’re seeing now is that unusual.
Anyway I do agree it will end in a big bust, but the wider world of private credit and sovereign debt will likely be part of it too.
You make some very valid points. I agree that regulation often results in 'toothless' agencies, and the EU's approach frequently crosses into overkill. Which is why innovation is slow in Europe.Perhaps the answer lies in an independent organization with delegated authority rather than a standard government bureaucracy.
That said, I simply do not trust the current narrative coming out of Silicon Valley. While the AI boom has single-handedly driven this three-year bull market, the rest of the economy remains fragile. Working in the AI trenches every day, I see the reality: the valuations for players like OpenAI and Anthropic are irresponsible. They are the lead actors in a circular financing shell game, and I wouldn't be surprised if one of them is gone by 2028.
When you factor in the 'perma war' in Iran, stubborn inflation, record consumer debt, and our looming sovereign debt crisis, the concentration of wealth in these few AI companies is terrifying. People love to say 'this time is different,' but it never is.
. For this not to implode, everything would have to go perfectly, and life and markets are never perfect. As an insider, I can tell you: AI integration for businesses is not the smooth transition the PR firms claim. I believe in the future of AI, but the foundation isn't nearly as solid as
This is all well reasoned but I think some of the variables you set aside have the potential to override this trajectory completely. In particular, the tenuous status of petrodollar hegemony and US hegemony in general, and the political fallout of a financial implosion that looks increasingly inevitable. Sure, Thiel et al. think they can ride a societal collapse into durable power, but I think those plans run into an angry, broke, and heavily armed American population like a brick wall, eventually. There's no historical precedent for a surveillance state that can hold up in these conditions.
You're right, and I appreciate the generous read. The piece is already 5,000 words and I had to stop somewhere! What I was trying to map is a constellation of pressures, not a prediction of exactly how things play out. The dollar question, a broader financial unraveling, the war in Iran...any of those could accelerate or redirect the trajectory in ways that make the AI displacement problem look like a secondary concern. Or they compound it. I don't know. Nobody does. That's part of the problem with the people who claim to.
If you hang with the executive class, they openly state that their goal os to reduce workforce numbers by 80% (this is a real conversation and a real goal being worked towards by multiple of the richest men in tbe world). However, if you hang with the executive class, you also understand that these people fundamentally don’t understand how their iwn products are built and have no realistic way of assessing whether AI is capable of doing the work. The concerning part is that investors might reward them in the short term for gutting their companies anyway. Thus I think the phrase dead economy or zombie economy is very appropriate. We already live on a world where the top 10% support 50% of economic spending. Some people call it a K-shaped economy, the tech bros call it a permanent underclass, the reality is they’re both talking about the same thing
There’s another concern here that I think’s important to talk about as well. AI companies aren’t making nearly as much money as they’re spending- at some point, investors will demand profit.
AI tools are cheap now, but when investors demand returns I see the prices rising dramatically. All those cost gains for companies who choose to automate will disappear as they pay out the nose for a system that’s too integrated to switch away from.
Wow, this is the best take on AI I have read. It is everything about AI that keeps me up all night every night put into precise language.
Our future is one where the best you can hope for is to be allowed, if your AI dictators permit it, is to stay home all day gambling on crypto and scrolling bullshit on social media. I’ve gotten to the point where I think about that all day and it’s destroying my mental health.
I think these companies are using AI as a scapegoat for PR purposes. The effects of the Covid Lock down for 3 years for some states and countries. Is now just beginning to be felt, compound that with the crashes we've had since 99, that we've never fully allowed to recover from.
Look at the timeline of events, covid lockdown happens, economy collapse worse then ever seen in the history of mankind, 50% of all business never reopened, recovery has been covered by EQ(even tho they now a call it something different) Revenue increase doesn't equal more sales of units sold, it just shows people are paying more per unit.
Think about it, they can go to shareholders or stockholders and say, look sales are down, while renue is up & this can't be sustained so we have to lay people off or fire them. PR wise it's much easier to blame AI after all it's the new boogeyman.
Fact is history shows that when there is a recession or depression those that work white collar jobs get hit the hardest and the fastest. In my Labor Monetary Theory I call these job Fiat Labor, easily scaled but produce little labor value.
Comparing tech changes of the past, is comparing apples to organs or trying to mix oil and water together. Tech back in the day didn't eliminated work, it just changed how that work was done, those that drove wagons teams learned to drive trucks, horses where used well into WW2, it wasn't until the 50s that you saw the total collapse of the horse market.
AI isn't going to automate anything, what is going to do is just change how that work is done. Most of what you listed is already automated, through excel and other such programs. Most reports have been automated for 30 years now. White collar jobs have been in the decline since the 80s, almost collapsed totally in the 90s had a boom in the early 2000s, but have never been near the levels that they were in the 80s and before.
Cloud based AI are going to die again like they did in the early 2000s, because of all the safety protocols that make them worthless for most professional users. But localized AIs are where things are going. Off web, trained to pacific tasks, limited to no safety protocols.
I agree that this is a plausible outcome. I do think, though, that the investment pouring into the hyperscalers is going to have a catastrophic impact on the US (and therefore, the global) economy when the numbers flatline.
Two things can be true at once: that the only thing that justifies the investment we're seeing is the usurpation of the labor market and that it's unlikely that the usurpation happens in any systematic way. We may retreat to running local open models (and that the Chinese have a much better hypothesis of how this plays out). The downstream fallout of the crash will likely be a fucking nightmare, even in mostly-good scenarios.
Isn't the argument that AI won't take "real jobs" but it will eliminate large swathes of what Graeber called "bullsh*t jobs"?
The question then becomes what percentage of modern work is bullsh*t, and what's the sensitivity and specificity of our identification of bullsh*t jobs?
If you're a fan of Douglas Addams you might recall the joke about getting rid of all the telephone sanitation technicians...
I'm going to answer your question with a question, which I know isn't proper, but so be it.
Question is what jobs are left since the 90s to be automated by AI? Coding has really been Automated for almost 20 yrs, customer service since mid 90s, Data collection and everything associated with it has really been automated since the mid 80s. Entertainment industry in large part has been automated since the 90s.
So what's really left?
I use AI because I don't have the access or money to find a human to do what I need done. It's augments the skills I have and is a cheap way to do the skills I lack in or don't have.
But overall it's not going to change societies ills
With the combination of AI and robotics, I can't think of any sector of the economy that wouldn't have major job impact. From bartenders, paralegals, and warfighters, most major job positions will be filled by less humans than AI/Robotics, unless society says differently.
Even combined with robotics, you won't see mass adoption, to expensive. For example look at the auto industry, they are in big trouble. All there robotics are in need of replacement. the original company that sold them their systems went out of business and new system are too expensive to purchase.
There was a article about 8 years ago, were the CEO of Ford said it would take them 60 or 100 years to recoup the cost of new robotics and AI systems. The Jetsons future is in a time far far far far far away.
I consider three scenarios. The first one is that your are right. Adoption of AI/Robitics will be slow either because of technical barriers or social adoption. One consequence of that scenario is that all of the AI investment currently being done will not have paid off in a typical investment timeframe. Classic bubble burst. Tremendous economic pain all around.
Scenario 2 is that the AI/Robotics impact to jobs follow robotics impact to jobs over the last 50 years. From 1970 to 2020, robotics is now doing about 40% of the work at factories. If AI/Robotics have the same impact to knowledge work over the next 25 years (because things continue to happen faster), replacing 40% of mostly entry level knowledge workers, would have a profound impact to the American job landscape as well as the higher educational system. Why go to college if you are not going to get a job?
The third option is that AI helps solves some of the technical challenges experienced when robotics was first introduced, thus shortening the time of the impact, and increasing the scope of the impact.
I don’t know which option is going to ocurr but we need good leadership to help guide the country through this.
I agree that all three of those scenarios are possible. scenario 2 tho is the less likely only when localized AIs really become available to the masses will it become viable.
The trend I'm seeing in AI today is the same one I saw back in the early 2000s with the 1st roll out with AI. It had limited safety protocols worked pretty well, then people complained about it and they implemented heavy protocols and the systems failed. We are seeing the same cycle this time around. Limited safety protocols systems work pretty damn good, inter heavey safety protocols and systems are failing to produce.
Localized AIs with limited safety protocols and specialized training will be the future, unless governments and the gatekeepers get in the way. The thing I worry about is AI being put into things like our cars, appliances, tvs, extra.
Isn't the argument that AI won't take "real jobs" but it will eliminate large swathes of what Graeber called "bullsh*t jobs"?
The question then becomes what percentage of modern work is bullsh*t, and what's the sensitivity and specificity of our identification of bullsh*t jobs?
If you're a fan of Douglas Addams you might recall the joke about getting rid of all the telephone sanitation technicians...
I am about to subscribe to you after seeing the restack of a quote from this article on SLART’s ‘stack. You are a man after my own heart!
You know your Camus and in particular, the Rebel, which has become, over time, my bedside book. He not only had the zeitgeist by the tail, he was prescient on the hydra-headed variables of the scene of our own time - in our case, a technological advance that its creators, swollen with cash, now appear to be in danger of believing their own bullshit with - except at bottom they really don’t, because privately, they know what this will disrupt. But they have no exit plan for what will happen when the shit hits the fan - I suspect they will believe themselves to be so rich that they can afford some sort of siloed communities of themselves, complete with all the bric-a-brac from bad sci-fi movies - pods, cryogenic freezing, and/or anything else that enables them to live till 150 with the appearances of those who aren’t a day over 65.
In short, they are the dandies Camus described. There’s Peter Thiel, a gay man who could care less about other gays and tolerates the politicians he sucks up to who are virulently homophobic. There’s Sam Altman, a Jewish-American who could probably care less about other Jews and sucks up to the raging anti-Semites in the admin. There’s Mark Zuckerberg, another Jew with a Chinese wife who sucks up to politicians and others in the admin who, again, are fang-dripping anti-Semites and racists, and would, as an added fillip, love to see his wife stripped of her citizenship and deported. I could go on and on but the picture is clear.
The point about who gets to live, who gets to decide who gets to live, and who are the human bacilli who are expected to consent to their own evisceration is alive and well in the talk amongst the rottenest in these circles. And your bringing in Camus at that point is brilliant. He broke with Sartre and the French Marxists because he saw through the lies of the Soviet Union. They only wanted to see heaven made horizontal. He saw industrialized slavery. He saw through any system that demanded endless sacrifice in one’s own time for a future that its architects didn’t understand but knew would render themselves redundant one way or another, so to perpetuate their own unearned power and addled prestige, they kept postponing their version of the Rapture. The parallels with AI are too close to be ignored. They’ve created a monster they don’t understand, but for as long as they can cash in on it, humanity be damned.
Well said Lee, I’m glad you got to read this gem through my sharing. As grim as it seems, we must be able to do something. There’s more of us than the few AI global elites.
I appreciate the detailed analysis. One thing I didnt see but kind of expected was more on dead capital markets and how AI has accelerated the machines trading with machines such that the market movements may have already taken human reality out of the picture. Curious about your take on that.
That’s definitely of a piece with what I write about here. The truth is, I had to stop somewhere. The piece is already over 5,000 words and that’s a bit topic. Maybe fit for another piece down the road!
When the “proletariat”, or more accurately, labor, is no longer needed it shall be liquidated. The largest killing of humans in history may be just around the corner.
Our rulers rape kids for fun, how far do they go when their lives are on the line?
If they chicken out, I forsee the joint stock corp failing in the long run due to entry and scaling costs falling so low that small groups can outcompete them. Imagine a small boutique company able to output joint stock levels of work. Add in 3d printing, assuming it continues to improve in quality and material selection, and you might even see people making things like cars and electronics from scratch. It may lead to massive decentralization if we get lucky.
I try not to let my brain go to your first premise. It’s possible, but there are so many off-ramps before then that I have to believe we’ll find one. We may not, but I have to believe we will.
On your latter point, I suspect we’ll see companies comprised of small teams grow into very successful enterprises without requiring massive fundraising in the same way if things play out in a best-case scenario. The best-case scenario is that we let the hyperscalers die and we ultimately run local AIs and build more and better businesses, create weirder things, and get back to creative endeavors.
Yes, precisely Owen, I am immensely grateful for your brilliant essay and the dialogue that it opens here. I am just starting a substack and have been engaged in my own writing and research regarding user based governance of AI. The crucial role of critically minded creative humans, who stay immersed in technology enough to steer it in the favor of human-positive locally supportive systems and bifurcate into avenues of human expansion, is of tremendous urgency.
That is my lane and all the elements that your research sharpens the focus upon is foundational to keeping the clarity and awareness in said creative humans directed in ways that build-NOW-the benefits that we cannot wait for elites or regulators to create for us. https://www.cetechcorp.com/staying-human
If our society can’t figure out how to make “stuff that is annoying to do gets done automatically” into a net benefit we all frankly deserve nuclear hellfire.
TL;DR version: what are you going to do when AI replaces your job and you no longer can afford to eat? And not just you, but your whole community, because everyone’s job has been replaced by AI? Because that is what the tech industry is building towards.
Yes, yes, YES! Thank you for this article! I've been saying for a long time now: who is going to buy all the stuff that companies sell when many/most of us don't have jobs!
I only push back on the notion that the ones building this system haven't thought it through. If you look at the secure places they have and are building for themselves you will see that it has been thought out. We are extra. That is all.
By the title, I was thinking you’d mention a specific, increasing fear of mine, that the economy will split into two, with one closed loop of AI companies working with/for AI companies, draining the real economy of wealth.
Your three turns (layoff -> spending cuts -> revenue decline) describe the demand side precisely. I'd add a fourth that's less visible: knowledge destruction.
When companies eliminate workers, they don't just lose spending power - they lose the accumulated understanding of how things work. Boeing has been rehiring retirees and contractors to recover capabilities its workforce no longer fully holds. TSMC is discovering it now - $40B factory in Arizona, but American engineers need a year of training in Taiwan because the knowledge walked out with the people a generation ago.
And there's an irony in the "80% capability" claim that keeps appearing everywhere. Even if true - where does the remaining 20% live? In the cognitive layer that only humans carry. The one being eliminated. We're trying to fix deindustrialization by pouring gasoline on it: solving the loss of human knowledge by removing more humans. Which means even the optimistic "people will retrain" argument has a problem it hasn't reckoned with: the capacity to retrain is itself being eroded by the same system.
The dead economy isn't just one where nobody can afford to buy. It's one where nobody remembers how to make. And that loss doesn't show up in any quarterly metric - which is exactly why it keeps happening.
Your Camus point lands. The person standing in front of you is not an input to a utility function - and what they know cannot be extracted from them without extracting them.
Absolutely correct. The knowledge loss would be enormous if this goes forward in the direction I fear it might. This is one of the things I’ve been talking about in my field (law): if we use AI to sub in for associate labor, at some point there are no senior experts. Biology is undefeated.
This extends to everything. Even now, it took a huge search to find craftsmen able to repair Notre Dame after the fire—because we just don’t build that way any longer. Now expand that to music, writing, chip-fabrication, and on and on…
The machine stops
Actually, the craftsmen were available because there are conservatories to keep the skills alive, such as Compagnons who tour France for several years to acquite crafts. But ignoring the skills acquired was a major trend even before ai
this is why I'm seeing it as the last industrial revolution, after this the industrialist making a something factory and offering jobs... is unmasked as a way the world can no longer work.
Time for a real structural revolution and post-industrial economics.
And longer reading lists 🤓
https://en.wikipedia.org/wiki/Idiocracy
There are many innovative people among us. We only need a shot and the ability to compete. It's been hard for anyone to tend with the conglomerates, but now that more people are looking for alternatives, this is small business and individual entrepreneur time to shine. If there is a will, there is a way.
I am currently in the process of creating a community center with the hopes of eventually being the go to grocer, daycare, barber, tailor, secondhand store, etc. Long term goals, but I'm going to work towards it because what else can I do. And if enough of us do this, we can create networks of goods for us and by us and barter them and eliminate money exchange and taxation and whatever else.
I am one person. Even 5 of me can make a difference.
When I read what you write after I write what I wrote, I feel like we’re thinking the same thoughts but you’re reading better books. Thank you for both making me feel like I’m not crazy and educating me to think better.
We’re definitely circling the same set of ideas—but I’d challenge the assertion that I’m reading better books. Any piece that pulls in Peirce is working at a pretty, pretty high level.
Really enjoyed your latest.
This is informative post.. I’d like to add that the financial structure behind the AI boom is a carbon copy of 2008, with private equity using SPVs to hide over $120 billion in debt and pooling it into asset backed securities, essentially mortgage fraud with NVIDIA chips.
The narrative that AI can replace workers is a lie; studies from MIT and Anthropic, plus real-world experience, reveal a massive gap between the pitch and reality. Meanwhile, the political endgame is already on paper. Peter Thiel’s recent manifesto and Palantir's operations reveal a shift toward control through force, where domestic surveillance and foreign warfare are the same product.
The "dead economy" is just the cover.
With 80% of Americans pessimistic and elites preparing for a post consent society, Organizing isn't just a .
Slogan It is the only option we have left. Thanks for writing this.
Yeah, I think the 2008 reference point is more salient than the Dot Com bust that is too often invoked. First, for the economic reasons you lay out. I'm just waiting to see synthetic ABS start spinning out of this shit. When I was first out of law school, I clerked and saw Floyd Abrams—Floyd fuckin' Abrams, who argued the Pentagon Papers case—argue that Moody's couldn't be held liable for its credit ratings because they were FREE SPEECH. I hope he needed that money thanks to a divorce or a sick family member or something, because that's dirty shit.
Anyway, the dot com bust is inapposite because the buildup of internet infra (fiber optic broadband, etc) had intrinsic and long-lasting value. We're still using it. The hundreds of billions we're pouring into data centers are DOA. The chips will either become obsolete or burn out, and we'll have to replace them on a cycle. We won't even get the benefit of the lasting infrastructure from this rutting festival of death.
That's assuming we get a financial crisis*--which would be a godsend. My worry is we don't get one and the nightmare future imagined in this piece comes to be.
*Should such a fortuitous event happen, there must be a massive popular movement to prevent any QE or bailouts of the financial system. It must be allowed to collapse.
Your assessment of the infrastructure build is not quite true. The data centers themselves, the buildings, the power infrastructure, the networking, the cooling, the racks, all have long lasting value, even if different generations of chips get cycled through over time. I worked on the telecom buildout and there was plenty burned through there, too. The Lucent and Nortel 5E switches, the DSL equipment, point to point wireless networks, the different generations of mobile wireless standards...all gone now, even as we’re left with exponentially more bandwidth capacity than we had then.
It is how investment works.
That's a fair correction—I overstated it.
The physical infrastructure has durable value independent of the chip cycle, and the telecom analogy is well taken. But I'd push back on the scale comparison. The top five hyperscalers are on track to spend somewhere north of $600 billion in capex this year alone, with roughly three-quarters of that directed at AI infrastructure. That dwarfs the telecom buildout's peak years by an order of magnitude. The question isn't whether data centers are worthless; they aren't, but whether the revenue to justify investment at this scale actually materializes. If it doesn't, we're not talking about some investors holding the bag. AI-related investment accounted for 39% of US GDP growth in the first three quarters of 2025. A collapse at this scale is a severe recession at minimum, and possibly a global depression.
Sure, but again, that’s investment. Long term capital going against an uncertain long term outcome. All big investment cycles eventually have their busts. It’s been 19 years since the last big one, so yeah, we’re due. Good to be prepared, but it happens.
I don’t mean to be cavalier, but I don’t exactly have hope that a super wise government would know what to do to prevent it. Suppress the AI boom in the US and then China owns it. Using fiscal policy and the Fed to manage it would just push the bubble elsewhere, even if the people in Washington were way wiser than they actually are.
I actually think AI will be a great economic development and do not agree with the dead economy theory. The population isn’t really growing, so AI automation will be needed to make workers more productive. The history of economics says that people will create new jobs even as old ones go away, and that should be the assumption until proven otherwise.
Btw, everyone knows Block’s announcement was BS. They overhired during the pandemic boom and they used AI as an excuse to cover up their bad management. Elon had already done the same thing at Dorsey’s other company (Twitter). He was a spendthrift.
All fair points, and I appreciate the good faith engagement. Please take my comments as a way to think through these issues as opposed to an argument.
Busts are normal, but scale matters. The telecom bust was roughly $2 trillion in lost market value and contributed to a mild recession. If the AI buildout is running at $600+ billion in annual capex and it unwinds, the exposure is structurally different. "It happens" is true but undersells the systemic risk when AI investment accounts for nearly 40% of GDP growth. The underlying economy is not doing well, and an AI crash could trigger contagion. Normal economics, when bubbles happen. Still messy.
On government: there's a slippage in your framing I want to flag. You start by questioning the idea that government could manage any of this, then slide to "suppress the AI boom" as though regulation and suppression are the same thing. They aren't, or at least needn't be. The question is whether we can have oversight, disclosure, antitrust enforcement, and labor protections without killing the industry. We've managed this in pharmaceuticals, aviation, and nuclear energy, all domains where regulation coexists with American dominance. The idea that any guardrails at all equals suppression is a posture taken by the VC and startup class, and I say that as a startup lawyer who loves my people. But there are priors baked into that world that are invisible from the inside—chief among them that private enterprise self-corrects and government can only make things worse.
Same framing move with China. "We can't regulate because China won't" assumes that any oversight means ceding the entire field. There's little to support that. And it's worth noting that China is actually taking a different approach: leaner, open models that run efficiently and do impressive work locally, rather than the closed hyperscaler model we're pouring $600 billion a year into. If the competition argument is supposed to justify our approach, it helps to notice that the competitor isn't even running the same race. The framing conveniently serves the interests of the companies arguing against oversight.
I think you're right on Block, and that's actually worse. If companies are already using AI as cover for layoffs that have nothing to do with AI capability (and many are papering over the post-shutdown hiring spree with "AI layoffs"), the narrative is doing displacement work independent of the technology. The story becomes self-fulfilling.
not trying to troll but if i try to argue with claude he speaks like this
The 2008 comparison is far more accurate than the dot‑ com analogy. Dot‑com left behind durable infrastructure ,fiber, broadband, backbone networks ,that we still use today. The AI data center boom leaves behind consumable hardware with a 3 year lifespan, financed over 7 years, losing 70% of its value by the time the debt comes due.
The same players from 2008 are back: investment banks, private credit funds, and securitization desks repackaging GPU leases into structured products. When synthetic ABS starts spinning out of this stuff, we’ll know exactly where we are in the cycle.
What makes this dangerous is concentration. The companies driving this boom represent 30% of SPY and 40% of QQQ. If one sneezes, the entire market gets sick. Retail investors and pension funds will be left holding the bag ,again.
This is the fastest, largest, and most leveraged capital project in U.S. history, and unlike past build outs, it leaves behind no lasting infrastructure. Just depreciating chips and amortized debt.
Then you had all the other negative variables at play.The market is delusional but can stay that way for a long time.
Appreciate the engagement, too. I’m not absolutist about letting the free market run at all times. I just don’t particularly trust the people doing the regulating any more than I trust the ones doing the building. It’s hard to know the future, so regulators making guesses isn’t necessarily any better than investors making guesses. I would hold up the EU as a pretty good example of where the regulators have been guessing wrong and have basically kept their economies out of the tech era. The GDP per capita of the leading countries in Europe are equivalent to Mississippi or West Virginia (at best).
The rest of the economy is weak because we’re still unwinding the excess of the pandemic era boom. Yes we technically avoided recession but many sectors have been dragging since then and high interest rates drag on housing and durable goods. The AI buildout is a high percentage of a low GDP growth number. Also the economy is twice as big as it was during the telecom and housing booms (which should be looked at together as the set up for the Great Recession) so not sure what we’re seeing now is that unusual.
Anyway I do agree it will end in a big bust, but the wider world of private credit and sovereign debt will likely be part of it too.
You make some very valid points. I agree that regulation often results in 'toothless' agencies, and the EU's approach frequently crosses into overkill. Which is why innovation is slow in Europe.Perhaps the answer lies in an independent organization with delegated authority rather than a standard government bureaucracy.
That said, I simply do not trust the current narrative coming out of Silicon Valley. While the AI boom has single-handedly driven this three-year bull market, the rest of the economy remains fragile. Working in the AI trenches every day, I see the reality: the valuations for players like OpenAI and Anthropic are irresponsible. They are the lead actors in a circular financing shell game, and I wouldn't be surprised if one of them is gone by 2028.
When you factor in the 'perma war' in Iran, stubborn inflation, record consumer debt, and our looming sovereign debt crisis, the concentration of wealth in these few AI companies is terrifying. People love to say 'this time is different,' but it never is.
. For this not to implode, everything would have to go perfectly, and life and markets are never perfect. As an insider, I can tell you: AI integration for businesses is not the smooth transition the PR firms claim. I believe in the future of AI, but the foundation isn't nearly as solid as
the valuations suggest.
Markets can run a long time on hype and delusion.
This is all well reasoned but I think some of the variables you set aside have the potential to override this trajectory completely. In particular, the tenuous status of petrodollar hegemony and US hegemony in general, and the political fallout of a financial implosion that looks increasingly inevitable. Sure, Thiel et al. think they can ride a societal collapse into durable power, but I think those plans run into an angry, broke, and heavily armed American population like a brick wall, eventually. There's no historical precedent for a surveillance state that can hold up in these conditions.
You're right, and I appreciate the generous read. The piece is already 5,000 words and I had to stop somewhere! What I was trying to map is a constellation of pressures, not a prediction of exactly how things play out. The dollar question, a broader financial unraveling, the war in Iran...any of those could accelerate or redirect the trajectory in ways that make the AI displacement problem look like a secondary concern. Or they compound it. I don't know. Nobody does. That's part of the problem with the people who claim to.
If you hang with the executive class, they openly state that their goal os to reduce workforce numbers by 80% (this is a real conversation and a real goal being worked towards by multiple of the richest men in tbe world). However, if you hang with the executive class, you also understand that these people fundamentally don’t understand how their iwn products are built and have no realistic way of assessing whether AI is capable of doing the work. The concerning part is that investors might reward them in the short term for gutting their companies anyway. Thus I think the phrase dead economy or zombie economy is very appropriate. We already live on a world where the top 10% support 50% of economic spending. Some people call it a K-shaped economy, the tech bros call it a permanent underclass, the reality is they’re both talking about the same thing
There’s another concern here that I think’s important to talk about as well. AI companies aren’t making nearly as much money as they’re spending- at some point, investors will demand profit.
AI tools are cheap now, but when investors demand returns I see the prices rising dramatically. All those cost gains for companies who choose to automate will disappear as they pay out the nose for a system that’s too integrated to switch away from.
For sure. I believe the technical term for it is a rug pull. Or predatory capture. Those tokens will get pretty, pretty pricey at some point soon.
Learn to run local models now.
Wow, this is the best take on AI I have read. It is everything about AI that keeps me up all night every night put into precise language.
Our future is one where the best you can hope for is to be allowed, if your AI dictators permit it, is to stay home all day gambling on crypto and scrolling bullshit on social media. I’ve gotten to the point where I think about that all day and it’s destroying my mental health.
I think these companies are using AI as a scapegoat for PR purposes. The effects of the Covid Lock down for 3 years for some states and countries. Is now just beginning to be felt, compound that with the crashes we've had since 99, that we've never fully allowed to recover from.
Look at the timeline of events, covid lockdown happens, economy collapse worse then ever seen in the history of mankind, 50% of all business never reopened, recovery has been covered by EQ(even tho they now a call it something different) Revenue increase doesn't equal more sales of units sold, it just shows people are paying more per unit.
Think about it, they can go to shareholders or stockholders and say, look sales are down, while renue is up & this can't be sustained so we have to lay people off or fire them. PR wise it's much easier to blame AI after all it's the new boogeyman.
Fact is history shows that when there is a recession or depression those that work white collar jobs get hit the hardest and the fastest. In my Labor Monetary Theory I call these job Fiat Labor, easily scaled but produce little labor value.
Comparing tech changes of the past, is comparing apples to organs or trying to mix oil and water together. Tech back in the day didn't eliminated work, it just changed how that work was done, those that drove wagons teams learned to drive trucks, horses where used well into WW2, it wasn't until the 50s that you saw the total collapse of the horse market.
AI isn't going to automate anything, what is going to do is just change how that work is done. Most of what you listed is already automated, through excel and other such programs. Most reports have been automated for 30 years now. White collar jobs have been in the decline since the 80s, almost collapsed totally in the 90s had a boom in the early 2000s, but have never been near the levels that they were in the 80s and before.
Cloud based AI are going to die again like they did in the early 2000s, because of all the safety protocols that make them worthless for most professional users. But localized AIs are where things are going. Off web, trained to pacific tasks, limited to no safety protocols.
I agree that this is a plausible outcome. I do think, though, that the investment pouring into the hyperscalers is going to have a catastrophic impact on the US (and therefore, the global) economy when the numbers flatline.
Two things can be true at once: that the only thing that justifies the investment we're seeing is the usurpation of the labor market and that it's unlikely that the usurpation happens in any systematic way. We may retreat to running local open models (and that the Chinese have a much better hypothesis of how this plays out). The downstream fallout of the crash will likely be a fucking nightmare, even in mostly-good scenarios.
Hyperscale AI data computing is already obsolete. Stand alone, (even air gapped) localized AI is the only sane solution.
Isn't the argument that AI won't take "real jobs" but it will eliminate large swathes of what Graeber called "bullsh*t jobs"?
The question then becomes what percentage of modern work is bullsh*t, and what's the sensitivity and specificity of our identification of bullsh*t jobs?
If you're a fan of Douglas Addams you might recall the joke about getting rid of all the telephone sanitation technicians...
I'm going to answer your question with a question, which I know isn't proper, but so be it.
Question is what jobs are left since the 90s to be automated by AI? Coding has really been Automated for almost 20 yrs, customer service since mid 90s, Data collection and everything associated with it has really been automated since the mid 80s. Entertainment industry in large part has been automated since the 90s.
So what's really left?
I use AI because I don't have the access or money to find a human to do what I need done. It's augments the skills I have and is a cheap way to do the skills I lack in or don't have.
But overall it's not going to change societies ills
With the combination of AI and robotics, I can't think of any sector of the economy that wouldn't have major job impact. From bartenders, paralegals, and warfighters, most major job positions will be filled by less humans than AI/Robotics, unless society says differently.
Even combined with robotics, you won't see mass adoption, to expensive. For example look at the auto industry, they are in big trouble. All there robotics are in need of replacement. the original company that sold them their systems went out of business and new system are too expensive to purchase.
There was a article about 8 years ago, were the CEO of Ford said it would take them 60 or 100 years to recoup the cost of new robotics and AI systems. The Jetsons future is in a time far far far far far away.
I consider three scenarios. The first one is that your are right. Adoption of AI/Robitics will be slow either because of technical barriers or social adoption. One consequence of that scenario is that all of the AI investment currently being done will not have paid off in a typical investment timeframe. Classic bubble burst. Tremendous economic pain all around.
Scenario 2 is that the AI/Robotics impact to jobs follow robotics impact to jobs over the last 50 years. From 1970 to 2020, robotics is now doing about 40% of the work at factories. If AI/Robotics have the same impact to knowledge work over the next 25 years (because things continue to happen faster), replacing 40% of mostly entry level knowledge workers, would have a profound impact to the American job landscape as well as the higher educational system. Why go to college if you are not going to get a job?
The third option is that AI helps solves some of the technical challenges experienced when robotics was first introduced, thus shortening the time of the impact, and increasing the scope of the impact.
I don’t know which option is going to ocurr but we need good leadership to help guide the country through this.
I agree that all three of those scenarios are possible. scenario 2 tho is the less likely only when localized AIs really become available to the masses will it become viable.
The trend I'm seeing in AI today is the same one I saw back in the early 2000s with the 1st roll out with AI. It had limited safety protocols worked pretty well, then people complained about it and they implemented heavy protocols and the systems failed. We are seeing the same cycle this time around. Limited safety protocols systems work pretty damn good, inter heavey safety protocols and systems are failing to produce.
Localized AIs with limited safety protocols and specialized training will be the future, unless governments and the gatekeepers get in the way. The thing I worry about is AI being put into things like our cars, appliances, tvs, extra.
Isn't the argument that AI won't take "real jobs" but it will eliminate large swathes of what Graeber called "bullsh*t jobs"?
The question then becomes what percentage of modern work is bullsh*t, and what's the sensitivity and specificity of our identification of bullsh*t jobs?
If you're a fan of Douglas Addams you might recall the joke about getting rid of all the telephone sanitation technicians...
I am about to subscribe to you after seeing the restack of a quote from this article on SLART’s ‘stack. You are a man after my own heart!
You know your Camus and in particular, the Rebel, which has become, over time, my bedside book. He not only had the zeitgeist by the tail, he was prescient on the hydra-headed variables of the scene of our own time - in our case, a technological advance that its creators, swollen with cash, now appear to be in danger of believing their own bullshit with - except at bottom they really don’t, because privately, they know what this will disrupt. But they have no exit plan for what will happen when the shit hits the fan - I suspect they will believe themselves to be so rich that they can afford some sort of siloed communities of themselves, complete with all the bric-a-brac from bad sci-fi movies - pods, cryogenic freezing, and/or anything else that enables them to live till 150 with the appearances of those who aren’t a day over 65.
In short, they are the dandies Camus described. There’s Peter Thiel, a gay man who could care less about other gays and tolerates the politicians he sucks up to who are virulently homophobic. There’s Sam Altman, a Jewish-American who could probably care less about other Jews and sucks up to the raging anti-Semites in the admin. There’s Mark Zuckerberg, another Jew with a Chinese wife who sucks up to politicians and others in the admin who, again, are fang-dripping anti-Semites and racists, and would, as an added fillip, love to see his wife stripped of her citizenship and deported. I could go on and on but the picture is clear.
The point about who gets to live, who gets to decide who gets to live, and who are the human bacilli who are expected to consent to their own evisceration is alive and well in the talk amongst the rottenest in these circles. And your bringing in Camus at that point is brilliant. He broke with Sartre and the French Marxists because he saw through the lies of the Soviet Union. They only wanted to see heaven made horizontal. He saw industrialized slavery. He saw through any system that demanded endless sacrifice in one’s own time for a future that its architects didn’t understand but knew would render themselves redundant one way or another, so to perpetuate their own unearned power and addled prestige, they kept postponing their version of the Rapture. The parallels with AI are too close to be ignored. They’ve created a monster they don’t understand, but for as long as they can cash in on it, humanity be damned.
Well said Lee, I’m glad you got to read this gem through my sharing. As grim as it seems, we must be able to do something. There’s more of us than the few AI global elites.
I appreciate the detailed analysis. One thing I didnt see but kind of expected was more on dead capital markets and how AI has accelerated the machines trading with machines such that the market movements may have already taken human reality out of the picture. Curious about your take on that.
That’s definitely of a piece with what I write about here. The truth is, I had to stop somewhere. The piece is already over 5,000 words and that’s a bit topic. Maybe fit for another piece down the road!
Hope you choose to go down that road. I've been curious how much that is a factor.
When the “proletariat”, or more accurately, labor, is no longer needed it shall be liquidated. The largest killing of humans in history may be just around the corner.
Our rulers rape kids for fun, how far do they go when their lives are on the line?
If they chicken out, I forsee the joint stock corp failing in the long run due to entry and scaling costs falling so low that small groups can outcompete them. Imagine a small boutique company able to output joint stock levels of work. Add in 3d printing, assuming it continues to improve in quality and material selection, and you might even see people making things like cars and electronics from scratch. It may lead to massive decentralization if we get lucky.
I try not to let my brain go to your first premise. It’s possible, but there are so many off-ramps before then that I have to believe we’ll find one. We may not, but I have to believe we will.
On your latter point, I suspect we’ll see companies comprised of small teams grow into very successful enterprises without requiring massive fundraising in the same way if things play out in a best-case scenario. The best-case scenario is that we let the hyperscalers die and we ultimately run local AIs and build more and better businesses, create weirder things, and get back to creative endeavors.
Yes, precisely Owen, I am immensely grateful for your brilliant essay and the dialogue that it opens here. I am just starting a substack and have been engaged in my own writing and research regarding user based governance of AI. The crucial role of critically minded creative humans, who stay immersed in technology enough to steer it in the favor of human-positive locally supportive systems and bifurcate into avenues of human expansion, is of tremendous urgency.
That is my lane and all the elements that your research sharpens the focus upon is foundational to keeping the clarity and awareness in said creative humans directed in ways that build-NOW-the benefits that we cannot wait for elites or regulators to create for us. https://www.cetechcorp.com/staying-human
By the numbers.....
https://talknet.substack.com/p/221-billion-suddenly-dead-by-2030?utm_source=share&utm_medium=android&r=7rxxzu
If our society can’t figure out how to make “stuff that is annoying to do gets done automatically” into a net benefit we all frankly deserve nuclear hellfire.
I may have some bad news for you, Miles 🤣
TL;DR version: what are you going to do when AI replaces your job and you no longer can afford to eat? And not just you, but your whole community, because everyone’s job has been replaced by AI? Because that is what the tech industry is building towards.
Yes, yes, YES! Thank you for this article! I've been saying for a long time now: who is going to buy all the stuff that companies sell when many/most of us don't have jobs!
I only push back on the notion that the ones building this system haven't thought it through. If you look at the secure places they have and are building for themselves you will see that it has been thought out. We are extra. That is all.
Incredible piece, so many tidbits of insight.
By the title, I was thinking you’d mention a specific, increasing fear of mine, that the economy will split into two, with one closed loop of AI companies working with/for AI companies, draining the real economy of wealth.
Yeah, I think that is a potential downstream consequence of this.
You’ve inspired me to write about it, maybe it resonates (I borrowed your title :) ) https://romanbelaire.substack.com/p/dead-market-theory?r=74s39d
Awesome. Will check it out.