"Something Wicked This Way Comes" — Why The AI Bubble Isn't What You Think

The biggest bet in the history of capitalism is being placed right now. $700 billion dollar this year alone. 6.7 trillion by the end of the decade. All spent on AI. It is the most expensive thing the human race has ever built. But so far, most people only use it if it's free. And the really scary part, if you're thinking about getting in on any of the upcoming AI related IPOs, is that the industry is taking on massive amounts of debt while only making a tiny fraction of what it's spending. And all the while, valuations just keep going up. It does not take a genius to know the current math does not work. But it also does not take a genius to know that AI could still have a lot of room to grow. And that conundrum is exactly why everybody's asking the same question. Is AI a bubble? But what we're going to be talking about today is that that is the wrong question. The right question is given that bubbles like this have formed many times before. What secrets does history hold that could tell us what landmines lay ahead of us and what the right way to invest is in a moment like this? What you're about to see is that history does indeed hold the key to understanding what's happening in AI right now. There's a very distinct economic pattern that plays out with revolutionary new technologies. And every time a moment like this occurs, some people get rich and others get wiped out. My goal with this video is to show you that pattern and help you end up on the right side of that equation. We're going to do it in three parts. First, I'll break down the pattern, but I also want to show you how there is a trap that's been set just for you. And in the end, I'll show you what the historical winners have all had in common. So, be sure to watch to the end. All right, welcome to part one, the historical pattern. four times in the last 230 years. A world-changing technology that has gone on to be utterly transformational nonetheless wiped out the first wave of investors who funded the whole thing. The canals in the UK that created much needed interior waterways, the British railways that connected the entire country, the American railways that did the same thing for the US but on a much larger scale. and the fiber optic cable buildout that made the modern internet possible. Each one of these technologies ended up making good on their promises in the long run. But for the first wave of investors, it was a catastrophe. It starts in the 1790s. A single canal in the UK had just cut the price of coal in Manchester in half, [music] and Britain lost its mind. Parliament approved 44 new canal companies in just 5 years. Then all of a sudden, the bubble burst. Share prices collapsed [music] and only a handful of those canals would ever go on to make their investors a dime. But the canals themselves stayed in the ground and they carried the coal and iron that powered the entire industrial revolution. The investors lost, but the new economy they accidentally financed [music] roared on without them. 50 years later, it happened again, this time bigger. They called it railway mania. By 1850, British families had poured close to half of all of the investment in the entire economy into just railway stocks, much of it bought on credit. One man, George Hudson, controlled a third of all of the track in Britain and was woripped as a financial genius. Until that is, it came out he was paying old investors with new investors money. It was a Ponzi scheme, which I hope sounds familiar. When it all unraveled, railway shares were cut in half and a generation of savings was wiped out. But the tracks remained. 6,000 miles of them and they ran the British economy for the next 100 years. [music] Then America caught the fever twice. The panic of 1873 bankrupted 89 railroads and set off a depression that dragged on for [music] years. Two decades later, it happened all over again. By 1894, a quarter of every railroad in the United States was in the hands of a bankrupt company. The people who funded that track obviously got ruined. The track itself, though, it never moved an inch and it continued to carry the goods that built the richest economy the world had ever seen. And then in our own lifetimes, fiber optic cable for the [music] internet. During the dotcom boom, telecom companies buried so much fiber optic cable during the internet euphoria that at the bottom of the crash, roughly 90% of it just sat dark. [music] It was paid for. It had all been installed in the ground through a very painful, laborious, and expensive process. But it was all turned off. Not because the internet was a failure. Obviously, it would go on to become one of the most important technologies in human history. It failed at the time because the infrastructure cost was so expensive that the revenue just didn't come fast enough to overcome the debt burden. So companies like WorldCom, Global Crossing, and a string of others went bankrupt waiting for that revenue to come in. WorldCom was the largest corporate collapse in American history up to that point. Hundreds of billions in investor money was just gone. But again, the cable stayed in the ground. It remained an incredibly viable asset and over the next 20 years that dark fiber would finally get switched on and it became the physical backbone of the modern internet. Netflix, YouTube, the entire cloud all runs on it still to this day. The companies that scooped it up for pennies after the crash, companies like Google, used that abandoned fiber to build some of the most valuable businesses on Earth, only made possible by those early investors that ended up losing everything. But putting the infrastructure in place and that gap between the infrastructure and when the revenue comes in is one of the most important things to pay attention to with artificial intelligence. That's the hard reality we all have to contend with right now. The buildout of revolutionary new technologies almost always requires a massive amount of infrastructure. And that infrastructure is insanely expensive. But some technologies are so obviously powerful and so obviously where we're headed as a society that success starts to feel guaranteed. And so despite the incredible risk that there will end up being a timing mismatch between the capital outlay and the debt that has to be taken on to develop the technology and when the money actually starts flowing in, investors still jump in with both feet, often times using [music] debt. Society benefits, but the investors and their creditors get hammered. But the pattern doesn't stop there. It loops around to the next wave of investors who come in and scoop up the distressed assets for much much cheaper and then reap the rewards of that first wave of investors who actually [music] paid for the infrastructure. Okay, that's the full takehome. The dark fiber might have been a bad investment for the first wave, but it was the competitive advantage that allowed the second wave to win. The initial investment was necessary. It had to happen, but it paid out to the benefit of humanity. and the second wave of investors. Now, given that pattern, it's up to all of us to try and figure out where we are in the infrastructure buildout and technological adoption curve that signals that revenue is going to come in as it relates to AI. If we're too early, we're going to get clobbered. But if we're too late, we die broke on the sidelines. And most importantly, if we misunderstand what's different about AI than everything that came before it, [music] we're going to learn the hard way that AI is less like a bubble and more like a time bomb for a very specific reason. Now, to understand what I mean, we have to analyze the economic and structural realities of AI that make it even more prone to the historical pattern that we just walked through than anything that came before it. So, welcome to part two. One of these things is not like the others. The strand of fiber optic cable buried under your city in the 1990s carries more than a thousand times the data today than it did the day that it first went into the ground. The glass and the cable itself is passive. So, nobody has to dig it up to improve performance of the internet. You just need to upgrade the cheap electronics at the ends of the cable if you want better performance. ultimately will reach the limit of the glass which only transmits a signal at about 67% the speed of light but it's already carried us for almost 30 years. You find a similar dynamic with the railroads. While railroad tracks do need to be replaced occasionally, every major component lasts years to centuries and a well-maintained track will run for roughly 30 to 50 years before being completely replaced. With railroads and fiber, cost and durability run in the same direction. The most expensive things, the route, the tunnels, the bridges, the buried cable, they all last the longest. The things that wear out or become obsolete more quickly, the ballast, ties, rail surfaces, modems, routers, etc., they're the cheapest parts, and in most cases, they still last a pretty long time. AI, however, inverts that financial relationship completely. Now, do not get me wrong. There are plenty of things that last within a data center buildout. The physical data center itself, the plumbing for the cooling system, the server racks, all of that. But the single most expensive component, the GPUs, is the fastest to die. We're talking 3 years to obsolescence, give or take, on the most expensive thing in the entire infrastructure chain. No prior technological infrastructure buildout has ever had its most expensive asset also be its most short-lived. And that's an anomaly that is likely to have massive repercussions for not only the first wave of investors in AI, but every investor thereafter. Historically, infrastructure has been dumb. Roads, track, cable, it's all designed to transport something. Now that something might be smart, but the road, the cable, the track itself is not. Instead of the road or the track needing to improve to keep up with the competition, the cars and the modems are the things to get smarter. That distributes the cost of the innovation contained to a smaller, more manageable piece of the technology pipeline. With AI, the infrastructure itself is a huge part of the technology. AI infrastructure is the brain. And while the brain itself isn't the intelligence in and of itself, the two cannot yet be separated. A key part of progressing AI technology is progressing the GPUs themselves. And as of today, they are almost comically expensive. Because a dumb foundation goes out of date only very slowly. You create far less economic drag on progress and revenues can more easily catch up with the initial investments into the infrastructure. The upfront cost may kill the first wave of investors with a distressing degree of predictability, but the second wave can reliably leverage everything that came before. That's why the second wave of investors have historically won. The asset the first wave paid for was still sitting there decades later, as good as new, waiting for somebody smart enough to build something profitable on top of it. But what happens to the investment math when a very substantial segment of the infrastructure itself must constantly be replaced? Right now, roughly a third of all the money pouring into AI is being spent on transient infrastructure, meaning that cost is never going to go away. It's a permanent tax on innovation. Nvidia ships a brand new generation of chips about every 12 months. Each one making the last look slow. Therefore, every AI product is built on a foundation of quicksand. To stay competitive, companies must eternally replace the most expensive part of their buildout. The only light at the end of the investment tunnel as far as AI is concerned is the oncoming train of the next generation of superior chips. We'll get right back to the show in a second, but first let's talk about how your brain actually works. Your brain can run on two fuel sources, glucose and ketones. Most people only ever give it glucose. Think of it this way. You would not put regular unled in [music] a Ferrari. Your brain is like that. It is the most complex machine on the planet and most people are fueling it up with the cheap stuff. And the Department of Defense believed in this enough to spend $6 million developing the technology behind today's sponsor, [music] Ketone IQ. Ketone IQ delivers pure ketones directly to your brain. They cross the bloodb brain barrier and fuel your neurons at the cellular level. There's no caffeine, no sugar, no crash, no stimulants at all. For me, I use it when I really need to be sharp. So, if I'm about to do an interview or something that I'm writing that really takes a lot out of me, and it carries me straight through. [music] Ketone IQ contains no artificial sweeteners or fillers and comes in delicious flavors like peach, green apple, and raspberry lemonade. If you're ready to fuel your brain properly, visit ketone.com/impact [music] for 30% off your subscription order or visit your local Target to get your first shot for free. Let's talk about a number that should concern [music] you. Data breaches have increased 211% [music] in a single year. And every breach means more of your personal data, your address, your phone number, your social security [music] number. It ends up on data broker sites where anyone can buy it. The question isn't whether your data is out there. It is the question is whether you're doing anything about it. That's where incogn. [music] They track down your data across hundreds of sites and remove it automatically. You authorize them once, that's it. They handle everything else. Plus, with custom removals, you send [music] them any link where your information shows up and their team is going to take it down. The threat is growing. Incogn is the response that grows with it. Go [music] to incogn.com/impact and use code impact for 60% [music] off an annual plan. Try it risk-free for 30 days. Now, let's get back to the show. And to make matters worse, in an effort to raise more capital, today's biggest AI companies may be lying about how long the chips will last. If you look at the accounting of the core AI companies today, they're going to tell you that their chips will last 5 to 6 years. But if you ask savvy investors like Michael Bur, the guy famous for calling the 2008 crash, he's going to tell you the real number is closer to two or three. In fact, he's gone on the record and accused the biggest AI companies on Earth of stretching that timeline on purpose to try and hide what he calls a 176 billion dollars in losses. The companies of course deny it [music] and in fairness, nobody really knows the exact number yet, but you don't have to settle that fight to see the danger. Everyone agrees the chips lose value fast. The only argument is over just how fast. And the people investors are trusting to determine that [music] number have a huge incentive to stretch things out while they can. That's the difference between this AI bubble and every bubble before it. Bubbles are all about confidence. When it pops, it will because people will have lost confidence. It doesn't require AI to fail or not be the incredible technology that I think all of us think it's going to be for the bubble to burst. There just needs to be a timing mismatch between how long people can remain confident in the future promise and when enough revenue pours in for those companies to remain solvent. And that timeline gets real short when people are using debt. And if the bubble does pop, the real question is who's going to be left holding the bag? Let's look at that. Welcome to part three. Us retail investors are called dumb money for a reason. It's the fall of 2008. $16 [music] trillion in American wealth evaporates in an instant. Millions of families go from euphoric homeowners with an appreciating asset to dead broke and homeless almost overnight. And the worst part, and never forget this, the disaster was engineered to make bankers rich while hiding risk and passing it on to homeowners and retail investors. And despite that truth, only one person [music] went to jail. The financial system is rigged against the novice. And there is a financial trick that has been used multiple times in living memory to make fortunes for a few off of man-made economic disasters. It goes like this. You take a risky asset and you find a way to hide the risk. It could be by slicing up and repackaging subprime mortgage back securities, giving them a new name and a top tier credit rating. That's what they actually did. Or it could be selling private debt with mismatched loan terms and payment schedules to unsuspecting retail investors and pension funds. That's the current suspected crisis [music] happening in the private debt markets. Or it could be what we're talking about today. taking a legitimately thrilling new technology with insane infrastructure costs that never quite go away and you do a little fancy bookkeeping and voila, one extended depreciation schedule later and the stock price seems far more reasonable. Then you IPO and use excited retail investors as the dumb money exit liquidity. If the bubble bursts, the retail investors left holding the bag and the government may even step in and prop up the companies because the sector is important for national security. Remember, you heard it here first. Let's call this trick the risk waterfall. Comes in many flavors, but the outcome is always the same. In 2008, that trick was called mortgage back securities. The banks building and selling those products were at the very same time they were selling them, actively betting against them. They knew what the risks were and they were counting on the fact that the public did not. The ratings agencies stamped subprime garbage as AAA, which is the safest grade that exists. And when it all detonated, and [music] erased $16 trillion of wealth, the government stepped in, printed money, saved the banks, [music] and hammered the already suffering average American with inflation. The risk was manufactured at the top, repackaged, and pushed downhill onto the people least able to understand it or defend against it. And that playbook has never stopped being used. It's being used right now today. I made an entire video on the version running right now in private credit, which you can watch right there. And it's all too possible that something similar is happening again with AI. If the AI companies really are stretching their depreciation schedules to disguise how fast their chips decay in value, and if this sudden stampede of trillion dollar AI companies racing to go public is even in part [music] about handing the risk to retail investors before people inevitably lose confidence in AI's currently insane valuations based on revenues, then to borrow from McBTH. something wicked this way comes. Let's walk through my rough swag of how this could play out. Start with the debt. The hyperscalers are borrowing hundreds of billions of dollars to buy chips [music] that have to be replaced in roughly 3 years. The banks making those loans obviously don't want to be caught holding paper that's backed by hardware collateral that is rapidly declining in value with each passing day. So they run the 2008 playbook. They slice the debt up. They repackage it. And they sell it off into pension funds to insurers, private credit, and increasingly into retirement accounts. The Financial Times reported on May 3rd of 2026 that large banks including JP Morgan Chase, Morgan Stanley, and SMBC are looking to offload the risk tied to hundreds of billions in AI data center debt because they're hitting their own financing limits. Their goal is to distribute portions of the risk to a wider range of investors through stake sales and risk transfers. The Financial Times own headline was that banks are trying to avoid choking on the debt. Back in December, Bloomberg also reported that Morgan Stanley was specifically considering offloading data center exposure via a significant risk transfer. These are known as SRTs. An SRT is a synthetic securitization where the bank keeps the loan on its books but buys protection on the first loss trunch. So for regulatory purposes, the risk is transferred and the buyer collects 9 to 14% roughly yields and the offbalance sheet version is also already live. In October of 2025, Morgan Stanley arranged over 27 billion in debt and about $2.5 billion in equity for a specialurpose vehicle tied to Meta's Hyperion data center site in Louisiana. The buyers are private credit funds, hedge funds, insurers, and pension funds. Now analysts expect insurers, asset managers, and pension funds with heavy private credit allocations to be the buyers of these SRT structures and discounted loan tunches. So that's channel one. That's how they waterfall that risk down. But they've got another option, channel two. It's the stock. For years, these companies have stayed private, which meant only venture capitalists, insiders, and the already rich could own a slice. But the biggest wave of IPOs in history is now forming. [music] SpaceX has filed, Anthropic has filed, Open AI is preparing to file as well. And together, those listings could push close to $3 trillion of fresh stock into the public market. And an IPO is mechanically an exit for the people that came before. The moment the early money finally gets to sell, sell to whom, you ask? you the public to retail. Analyst Jill Lura of DAD Davidson was quoted by Al Jazera as saying the companies are in quote a race to go public before capital runs out. That sounds about right to me. Now let's put those two channels together of how risk is shuttled downstream. The debt gets sold into your pension as a part of a risk transfer. The stock gets sold into your brokerage account as a risk-on opportunity in the hottest part of the market. And fair enough, it is. If the revenue comes in fast enough, you are loving life. But if it doesn't, you get clobbered. So ask the four questions you should ask of anything in finance. Who created the risk that makes this investment have a return? The hyperscalers in Wall Street would be the answer here. Who packaged it? The banks who do not exactly have a great record? Who's selling it? The banks and the insiders who are cashing out through the IPO. And who ends up holding it? Look at your 401k. Roughly a third of the entire S&P 500 right now is the same handful of AI companies. If you or your portfolio manager are chasing returns in the US stock market, you are very likely to be heavy into AI. Now that may be perfect, but it may also be calamitous. It will all depend on timing. Timing of the dramatic revenue increases that are needed to make AI profitable. And the needed increases are dramatic. And it's about the timing of investor confidence in the sector. And it's the timing of the depreciation rate of the GPUs themselves and what that number ends up being in reality. Now, I can't tell you the timing of any of that. Nobody can. It's going to have to play out. But the technological revolutions that have come before AI have ended up paying off only in the long run. Before that, entire crops of investors [music] have gotten wiped out along the way. My belief is that AI really is the single most important technology ever invented. [music] My intuition, for what it's worth, is that it will be much more transformative than the industrial revolution. And if I'm right, there are going to be so many people that get wealthy off the back of this. But getting the timing wrong is the same as being wrong. If your time horizon is short or you're using debt, the future almost always surprises everybody. And the markets have a way of following the path of maximum pain. Don't lose sight of that. But what I can tell you is that the waterfall risk structure is the same structure that left normal people dealing with the engineered catastrophe of 2008. Now, I'm not predicting the AI is going to be another implosion, but in my own life, I am acting extremely paranoid. Now, my goal here is to give you a fresh perspective on what's happening beneath the surface of this truly thrilling technology. So what does that mean for you? What should you do with this knowledge? Every new revolutionary technology demands an infrastructure buildout. There's nothing weird about that. The canals, the railways, the fiber, but a dangerous combination of enthusiasm and reckless investing tends to wipe out early investors. They get out over their skis with debt and fail to understand how long it can take for even the most incredible transformational technologies to actually generate enough revenue to outpace the debt required for the tech buildouts. If you guys aren't seeing the push back that people are having to AI, if you're not seeing how companies are spending more than they expected and getting less in return, this is all going to take time. [music] And the odds that it takes more time than anybody who's invested wants borders on 100%. So step one is to realize the tech can both be the right thing and have timing that may take way longer than expected to become profitable. So don't overextend yourself by making bets with debt. Second, stay humble. I know the winners seem obvious right now. It seems like you can't miss with the big names, but they're almost never as obvious in this early stage as we expect them to be. So don't bet on any one single horse. Nobody in 1850 could have told you which companies would survive and which would go bust. The same is true with the internet. And the same is true with AI. Bet on the sector, not an individual company. Third, play the long game. The people who get destroyed when a crash comes are those with concentrated bets. Bets placed [music] using debt and bets that require a fast payoff. If you need your money back in a year or two, you could get wiped out by a simple recession. But if you don't need a return for 20 years, [music] history says you're going to win. Even the Great Depression had people back to positive yields within 20 years. You have to be able to hold through the storm to avoid locking in your losses. And last, as always, diversify yourself across economic forces. AI being in a precarious state right now is not the only thing to pay attention to. The US is drowning in debt. We are politically dysfunctional and the world is geopolitically super fragile right now. Assume the future is uncertain and diversify accordingly. One last thing, remember that the system is rigged to an uncomfortable degree against the novice. Unless investing is your full-time job, assume they're out to get you, but there's always a way to win. They cannot stop you from playing the game well. But it is a competition, so play wisely and defensively. All right, if you want to see me explore topics like this in real time, be sure to hit that subscribe button and join me Monday, Wednesday, and Friday at 7:00 a.m. Pacific time. [music] I hope to see you there. Till next time, my friends, be legendary. Take care. Peace. If you like this conversation, check out this episode to learn more. In 2008, a team of neuroscientists in Berlin put people in an fMRI machine and asked them to press a button, left hand or right hand, whatever they wanted. The researchers then watched their brains and what