We are living through a surreal moment in the world, in fact, several very surreal moments for various reasons, but one of them, perhaps the one with the greatest economic impact in the short term, is the investment bubble in artificial intelligence, which is actually investment in large language models, which are one type of artificial intelligence, because there are several others that are profitable, that have healthy investments, that have good returns. But this one in particular is very strange. To give you an idea, almost 3% of the US economy's GDP is being invested in this. They are being tied up solely in the construction of data centers to run large language models for some companies. This is the largest concentration of resources in the development of a technology that we have on record in history. The entire Apollo mission, nearly 13 years of program development from the first rockets to six lunar walks that have never been repeated, did not cost in total, over all those years, what is being invested in data centers in 2026 alone. Again , only in 2026. So it's time to see what the last technology bubble was , who burst it, who profited greatly from it, what financial maneuvers were involved, and how much this past rhymes with what we are seeing today, and how China can come out on top of this, believe it or not. Ah, but before we get to the topic, here's a tip from Revolute to help you plan your trips using your money wisely. In the Revolute app, you can select stays around the world with discounts of up to 60% and bonus points. You can earn up to 15 points per dollar by booking your stay there. Then you can use the points from your stay to convert them into miles. The Revolute app already offers more than 40 airlines where you can use your points. Most of them are international, but there are also some national ones. And Revolute is one of the only accounts in Brazil that allows you to transfer points to international airlines. If you start using Revolute today, you'll have time to buy your airline tickets using only the benefits of the card. And with Revolute cards you earn points on debit or credit purchases, both domestic and international. With the Ultra card, for example, which is the one I use, I earn three points for every dollar spent on credit. All my expenses immediately turn into benefits. And if you're not into frequent flyer miles, you can exchange them for everyday benefits , such as gift vouchers. You use your points and redeem vouchers to use at these partner establishments. It has transportation, accommodation, shops and more shops, food, a market, and many other amenities. Just go back to my previous videos and you'll see how this list keeps growing. And the best part is, by opening your Revolute account through my link, you're guaranteed up to R$200 back. Simply use your card and you'll receive cashback on all eligible purchases during your first 30 days, directly into your account, starting with your first purchase, whether in Brazil or abroad. back. It has been more than 60 years since the stock market, or rather, the SP500 index of the largest companies on the United States stock exchanges, has had such concentrated capitalization in so few companies as Nvidia, Apple, Microsoft, Alphabet, Meta, Amazon, Broadcom, and others, with, as I write this, almost 40% of the value of all these SP500 stocks concentrated in the 10 largest companies. But the last time the market was this concentrated, in 1965, the top 10 companies were very diverse, ranging from food and chemicals to computer and car manufacturers. Now, if it's not the technology companies that are developing large-scale programming languages ​​and artificial intelligence models, it's the companies that sell components to them or that have invested in them, except for the occasional Aranco oil company that pops up on the list, but even then, it might be supplying gas and oil to those who are burning it in data centers. This is because the value invested in the stock market is currently at an all-time high, with a large influx of money from individuals in the United States who are buying stakes in companies. as a form of investment. It's a huge and unprecedented gamble on a future where this type of technology will dominate society and generate unprecedented profits. Because if that's not the case, then it's going to be complicated. But this bet is based on some serious problems. It's time to take a look at the bubble of the 2000s, at what one of the great recent examples of fraud, In Ron, did, and use that lens to think about how what we're experiencing now seems much more complicated when we lift the rug to see what's underneath. First and foremost, Enrone is an excellent case study of the scale a financial scam can reach, where highly qualified individuals, believing themselves superior and smarter than all their competitors, can seize an opportunity to make a large sum of money, creating a trap that leaves the next victim to pay the price. One particular lesson is how they use loopholes, schemes, lack of transparency, and the goodwill of the press and investors to hide billions of dollars in debt on projects with no return. And I recommend Smartest Guys in the Room again, which provides a very good explanation of what follows . Inrou had a business that was actually profitable, which was the production and distribution of natural gas during a period of high demand in the United States . And then she leveraged the fame and success of this growth to move into new areas such as energy trading and even telecommunications. Something I explain in more detail in the video about infrastructure privatization. Now, it's more worthwhile for us to take a look at the creative financial practices that the company has devised and refined to grow its size and increase the value of its shares. For example, it was the first company to count the presumed profit of a contract as revenue before receiving the money or knowing the actual cost of operating it. How do I declare a signed agreement to build a streaming service with Blockbuster? Yes, a pre-Netflix Netflix with a profit of $110 million , imagine the money they could make with that, and then maintain that positive value on the balance sheet, even when the contract was abandoned. And that actually resulted in losses; all they had to do was be optimistic, project fantastical scenarios, and they managed to break record after record in profits, inflating the stock price as a result. To give you an idea of ​​the accounting creativity, when they decided to invest in telecommunications, when they entered the dot-com bubble that was booming between 1995 and the 2000s, at full steam, they discovered fiber optic switching. The companies and startups that were competing to install the fiber optic network that would power the internet they were designing—which did indeed happen, but took decades to truly materialize—discovered that buying fiber optics was one thing. Now, installing this as infrastructure took much longer. So, companies were competing to buy the available fiber optics to create a strategic stockpile and deprive competitors of the resource, but then they were left with this fortune in fiber optics sitting idle for months or years in a warehouse, in a shed, while the installation faced the real world. And then they discovered at RON that if they sold the fiber they had stored to another company and bought it back, the sale could count as an immediate gain, while the expense of buying back the fiber could be treated as an investment. And the investment can be spread over up to 20 years in the case of fiber optics, which is its depreciation period . After all, it remains valid for that entire period. And if everyone wants fiber, it remains very valuable even after it's been used. Then it wouldn't lose as much value. So look how the business turns out when you take that mindset into account. Not that this is real. I'm selling you R$10 million worth of fiber optics. Then you sell me 10 million fiber optic cables back. I record 10 million in revenue from the sale and deduct half a million from that, which is the depreciation I spend spread over the first year of my investment. In other words, I'm recording a profit of 95 million, which is that difference. And you too. And nobody had to pay anybody. No one needed to move anything from the inventory; nothing was actually exchanged. But these two companies managed to convince shareholders that we're already making a profit even with all this fiber sitting idle. So, in this crazy fiber buying frenzy, buying and stockpiling equipment and all that fuss, Cisco, which was the leading manufacturer of communication equipment at the time, saw its market value inflate until it became the most valuable company in the world for a few months, until the bursting of the bubble brought down the value of its shares, which continued to be worth at 1/4 to slightly less than they were at their peak for 25 years until the recent demand generated by the current growth of AI brought it back to that market value. And most importantly, what made much of what Inron became possible was the creation of special entities, separate companies, that absorbed the cost and risk to keep Inron's balance sheet clean. Some of them even became major fronts for huge schemes. For example, if an Inrro power plant were to produce for many years, they would have to wait 20 years to be able to say they made X profit. They could, in fact, sell that plant to a company created for that purpose, a spinoff, for the full value of all those years and count that as immediate profit. So, this company only served this tax maneuver, to declare a presumed profit from the plant for decades. Or they could, for example, buy a portion of the unused fiber optic cable from their inventory and perform that fiber replacement. Or this company could be created to buy the toxic assets, the part that was actually causing losses, for a much higher price than it was really worth. So, by cleaning up the balance sheet and further inflating the company's profit, and then completing this circular scheme, they could buy back that company, also counting the purchase as a profit. And where did the money for all these maneuvers come from ? from bank loans and from the sale and appreciation of INRON's shares . Therefore, for banks and the stock market, they had to show a constant and growing profit, which kept the company's reputation high and also kept a line of credit open so they could lend more and more and borrow to pay off loans. Then they started using separate companies, these companies created for that purpose, to do securitization, which is to group the debt and sell that debt as a security to shareholders who will keep the profit from the interest they would be paying to the bank. They could also create companies with creditors who bought something from Enron, such as declaring that they bought a certain volume of natural gas or oil, paying upfront, and then after a while selling it back in installments or leasing it back to Enron with interest. They were essentially taking out a loan, but on paper it counted as a foreign investment or a sale, further inflating the stock price. The challenge was to keep investors amazed enough that they wouldn't want to look closely at what was happening behind the scenes while this debt was growing, growing, growing, hidden away. And to cope with ever- increasing debt, more shareholders had to come in. The loan guarantees, the financing of these shell companies, and the investment partnerships used INROM's shares as collateral. So, for example, it would create a company that bought a loss-making plant from Inrone, using Inrtava shares to generate cash, thus freeing up the expense on the balance sheet and causing the shares to rise, which covered that expense. They financed the debt with the bank again, using shares that were now worth even more as collateral. And the money from a loan or a securitization package sale was used to pay the interest on other loans they had already taken out. As long as the stocks continued to rise, their appreciation funded this financial pyramid scheme that just kept growing. And the problem would be who was left on the day that the stocks stopped growing as they did . The driving force behind all of this and the motivation of those who participated was precisely this influx of investment money, both from the enthusiastic and misled public, fueled by false projections, and from the people who bought shares in the process and profited from it. Contracts, promotions, employee goals—everything within RON was geared toward ensuring that the stock would continue to rise in value, preferably by more than 10 to 15% per year. Hence the need to evoke profit from any agreement, even if that profit did not actually exist, and to keep jumping from market to market, first in natural gas, then in electricity, and then even in internet, fiber optics, and telecommunications, to give the impression that there was room for growth. For more than a decade, Enron grew, reporting record profits and increasing in value, while internally it accumulated debt upon debt. But since these debts were in the name of shell companies, they did not appear on the balance sheet. And what the process did was precisely to inflate Inron's balance sheet and profit, pleasing investors and causing the stock to rise. At the same time, the high-ranking officials who knew what was happening and were making it happen were receiving absurd bonuses for meeting growth targets. They won, won, and won. They would get stock options, bonuses, and all sorts of other things, and then jump ship before the bomb exploded. Those who also made a lot of money were the auditors who should have noticed the irregularities, but were instead making too much money off AON while it was growing. The bomb could have exploded one day, and it did, but those who participated made a lot of money in the process. The day the debts became too large, and the stocks actually stopped growing, and the castle collapsed because there was no longer any way to finance the debt, the loss fell on those who bought the shares and lost all their value. Those who took out loans were left with interest and had insurance on those loans. Those who worked within Inrone setting up this entire scheme had already pocketed the multimillion-dollar bonuses. In other words, it fulfilled the function of generating profit for those who were able to profit. Those profits did n't go to those who believed in the stocks. Now that we understand this process of loan upon loan, generating a tangled web of obscure companies and processes that remove debt from the balance sheet, let's take a look at the current situation regarding the artificial intelligence boom. Until yesterday, until 3 years ago, Alphabet, which owns Google, Meta, which owns Facebook, WhatsApp, Instagram, Amazon, Oracle, and Microsoft were the companies that had a lot of money to invest, a lot of money to spare, because they profited much more than they cost to operate. They never needed to take out a bank loan until very recently . So, initially, when the competition to invest in building data centers begins, these are companies that have unprecedented conditions, which other companies don't have, to finance the insane investment in infrastructure, in chips and data centers, where these chips will be installed to count on a future return. That's what they always did without any problem. But the hype surrounding language models has become so great that those who aren't investing more than their competitors seem to have missed the boat, and their stocks are falling. Then everyone has to invest much more than the others, who are already investing more, who need to invest more. And having to spend to maintain status, the expenditure on chips and the construction of data centers reached such a high and large point that even this unprecedented balance sheet of these companies cannot sustain it, and they began to need both external investment and loans. And that's where things start to get a little complicated. Starting with foreign investment, the shares of these companies continue to rise, and the reasons seem very promising. To capture the company that has grown the most from all of this, the most valuable company in the world right now. Who's selling the PAS (Personal Protective Equipment) for this gold rush, who's selling the jeans to these miners, who 's selling the telecommunications equipment to the internet companies? Currently, Nvidia is the company that makes the leading high-end chips for processing large language models in data centers. In its August report to investors , for August 2026, Nvidia gave the great news that its profit doubled compared to the previous year and, even more, is expected to almost double again in 2027. That's insane, it's the largest company in the world of that size, worth 5 trillion, saying that from 2025 to 2026 it doubled its revenue and is expected to almost double again by 2027. Look at the size of that market. This makes it one of, if not the most profitable company in the United States. with a tiny detail, a very small detail that they put in the supplementary footnote of your financial report. About 70% of that revenue came from five unnamed clients. Now, if we look at the five largest data center builders, who happen to be buying billions and billions in chips from Nvidia—companies like Alfa, Bet, Amazon, Meta, Microsoft, and Oracle— about 70% of the revenue of these five companies comes from two clients, OpenAI and Antropic. In other words, everyone's there, Nvidia is hanging on to five companies, and those five companies are hanging on by duct tape. in these two companies, Open AI and Antropic. And who is investing in OpenAI and Antropic with the money they are spending on Nvidia, Alphabet, Meta, Microsoft, and other things? Those investing in them include Nvidia, Alphabet, Microsoft, Meta, Oracle, and Amazon. What appears slightly circular until you look closer, then you'll see it's completely circular. The biggest investor in OpenAI in 2023 was Microsoft. And OpenAI became Microsoft's biggest customer, spending all the money that was invested in it with Microsoft. Amazon and Google invested in OpenAI's competitor , Antropic, which used that money for Amazon's cloud computing and Google's cloud computing and chips. Nvidia then took advantage of this stock appreciation to invest in promising startups and large companies, such as OpenAI, CorewaWave, and XAI, which, when they don't buy Nvidia chips directly, spend money with companies that do buy Nvidia chips. Then Microsoft and Nvidia also decided to invest in Antropic, which had been sidelined, and it started using Microsoft servers with Nvidia chips. Then Amazon also invested in OpenAI, which also started using the Amazon cloud. Then Oracle, which arrived late, also invested in OpenAI, which then used that money to utilize Oracle's cloud, and so on. The money keeps changing hands, everyone's announcing absurd investments and absurd demand. Stocks around the world are rising, and those who invested in companies are seeing their money multiply, while the bubble is left for those who come later. And do you know who else gets involved in that? Who's showing up? Special companies created to receive money and make investments outside of the companies' financial balance sheets. Meta, for example, has created individual organizations for each data center it builds, linking the debt and loans to these companies and not to Meta itself. And she 's not the only one, everyone is doing the same to maintain a very positive balance, despite the expenses. And then you can see things like a company that was founded with shares from Nvidia and Musk's XAI, which buys chips from Nvidia, assumes that debt, and leases those chips to XAI. So it's not that Musk's company incurred the debt; the debt was incurred by a third party who uses Musk's company's money to pay off that debt and lends those chips to Musk's company. There is also a pool of investors with Nvidia creating debt securitization packages. In this way, they are able to support startups that couldn't afford to buy chips on their own, or that can't pay for the chips they need because they aren't financially self-sufficient, but will be eventually. This creates even more demand for Nvidia chips, which everyone is buying in a mad rush, but forgets about fiber optics. In practice, this is removing that debt from Nvidia's balance sheet . In other words, all this spending that I mentioned at the beginning of the video, which I 'll come back to shortly, is only a portion of the actual spending that these companies are incurring. And what are the guarantees for these investments and loans? The companies' stocks and the chips, because after all, chips are in demand and they're worth a lot. Chips made by these same companies, bought by these same companies or by others that appear in the process working for them, such as Corew, which is a company that received chips and investment from Nvidia and contracts from Microsoft, OpenAI, Antropic, and Meta to build data centers for them, which generated, you can imagine, profit, right? Who needs profit? Core has generated a small debt of $50 billion in loans, which has left investors who put money into their stock a little worried. Not because it's illegal to create these companies, as was the case with RON, which hid companies it had created for that purpose. Now everything is done openly and in a much more auditable way, but that's because it can be financially unsustainable, even though everyone knows what's going on, like what happened with Cisco when the internet bubble burst in 2000, mainly because everyone is stuck in the same mess now, buying from each other, generating demand for each other. And the situation only gets more confusing with the loans they've started taking out from banks, companies, and governments, since they need even more money and even investment isn't enough. So they are lending money all over the world. Big Tech companies, those that are building data centers, have always been very profitable; they've always been the companies with the highest margins and cash reserves until very recently. So now they're entering the global market to compete for loans with a capacity no one else possesses, enabling them to lend money for data center construction everywhere, much faster and for far greater amounts than the rest of the world can afford to lend. These companies are lending billions and billions of dollars, Canadian dollars, Swiss francs, dollars, euros, pounds, and whatever else comes along. And so, not only are they taking on unprecedented debt, trapping several countries in the process, but they are also making money available for investment and loans around the world scarcer and more expensive for all other companies. To give you an idea of ​​the scale of what's happening, this could be affecting the value of countries' treasury bonds. They may reach a point where they are competing with governments for loans, which is a whole different scale. And taking the banks along for the ride, as is the case, for example, with Japan's SoftBank, which has taken on an unprecedented historical debt to continue investing in OpenAI, on which they have already spent more than 60 billion dollars. So, they're burning through money to keep the company afloat until they can pass this hot potato on to the next person, and who knows what comes next. And so, at this rate of spending, hypersales, which are these companies that are building data centers, are on track to spend around 10 trillion dollars by 2032, which corresponds to more than 3% of the United States' GDP, directed to just one industry, which is the largest expenditure, the most concentrated expenditure in the country's history. In the past, they have experienced cycles of technological explosions that absorbed investments far exceeding actual demand. That's what happened in the biggest craze of all, which was the construction of railroads at the end of the 19th century, where many people took on debt and put far more money into building far more railroads than they actually needed, which ended up consuming 2.2% of the GDP. The dot-com bubble , which occurred long before and much faster than the actual growth in internet demand in the 2000s, saw approximately 1% of GDP being invested in telecommunications. Current spending in 2026 will already reach almost 2% of GDP. There is such a concentration of money that in the first half of 2026, spending on data centers was the only type of private construction that grew in the United States. An additional 9 billion dollars was spent. Meanwhile, spending on any other form of private construction—houses, buildings, industries, shops, warehouses, shopping malls, you name it— fell by almost 50 billion in the same period. They've sucked up so much money and so much manpower that data centers are the only private construction expanding in the United States, growing at the expense of everything else, taking the place of everything else. Even the price of copper is going up, all over the world. But now, in addition to having considerably larger and more concentrated investment than past bubbles, this growth is also absurdly less sustainable, much more depreciable than previous ones. One of the arguments in favor of a bubble, which is, "Oh, it might be a bubble, but in the end, everything they're doing could be important, it could be used," is that even for past economic bubbles, the investment cycles left behind a lot of infrastructure that continues to be used and perform well after the bubble bursts. Railways are still important more than a century and a half later. The fiber optic network from the 2000s is still in use and has underpinned much of the internet growth that followed. However, both of these cases involve infrastructure construction, where many things can be overlooked. Any type of train can travel on a railway, any type of data can travel through a fiber optic cable. The bulk of current spending on artificial intelligence, with large language models, is the purchase of highly specialized chips, developed with an architecture that is still being refined and is becoming outdated in about a year and a half, from one generation to the next, which is used specifically within data centers for current language models, coming from Nvidia. And the generations of chips can be incompatible with each other. In other words, a newer chip won't work in outdated data centers that are several years old. In many cases, motherboards from the previous generation cannot support chips from the next generation. So, if they want to keep this race going using cutting-edge chips, they need to replace everything together. And just a reminder that Revolute sponsors the Audi Formula 1 team and is already the most downloaded finance app in Europe. If you travel there, Revolut is present in several airports, with over 80 million customers worldwide. You can trust me and take advantage of the fact that if you open your Revolute account through my link, you're guaranteed up to R$200 back directly into your account. And going back, in a great coincidence that you'll see is familiar, the companies that are buying all these chips have started to extend the depreciation time of the chips from two or three years to five or six. They began to consider that these chips continue to be valuable for a longer period, so that this entire expense can now be spread across the financial report over several more years as an investment, while the chips continue to count as assets, even if they are in storage. Even if that's the case now, unlike with fiber optics before, and the demand for IR (Intelligent Reliability) from data centers keeps the chips in demand for longer, worth more, with a very heated used market, they may continue to be worth more for 10 years. This is still far from being the durability of a fiber optic cable or a train track, let alone their universality. And with these data centers also comes a demand for electricity, water, and other regional resources, which are now being spent with investment money so heavily that it raises the price of everything and captures electricity from factories and homes that need to close or relocate. So these are expenses that, in addition to everything else, alter the entire region where they occur. And the mess with all this chaos is so big that it overshadows the real indebtedness of these companies, even more so with intermediary companies helping to clean up the financial balance sheet. The Wall Street Journal calculated that by mid-2026, behind the $600 billion in recorded balance sheet spending, Alphabet, Amazon, Meta, and Microsoft had incurred debt and signed commitments totaling more than one and a half trillion dollars that are not appearing on their balance sheets . Another 2.5 trillion dollars. This means that the actual debt is almost five times greater than what is recorded and declared. While Nvidia's CEO is projecting that companies will have to spend between 3 and 4 trillion dollars more on infrastructure, buying their chips, if they want to keep Nvidia's stock growing, or rather, if they want to remain competitive in artificial intelligence. All this just to buy chips that can't be plugged into data centers right now. They're not being plugged in because it takes time to build the specialized data centers and because there simply isn't enough electricity in the United States for what they're building right now. It's quite reminiscent of that fiber optic cable that was stockpiled by companies in '99. According to Goldman Sachs, these companies that are building data centers at this rate with this spending would need to generate annual AI revenue of at least $300 billion to recoup the investment that has been recorded on their balance sheets. Ideally, they would have to move more than 1 trillion dollars a year to have a safe profit margin. But the current revenue of all these companies isn't even enough, being very generous and ignoring the fact that much of it is circular. $200 billion . And the bulk of this spending is concentrated and being done by OpenAI and Antropic. And even if that revenue grows, expenses will grow along with it, because for these companies to grow, they have to subsidize what they provide. So, if they grow, the more they spend and owe, the faster they grow . And to make matters worse, this spending is all tied to companies lending to or investing in other companies, as we've already discussed here. What really happens is that those who want to isolate risks when entering this market end up diluting their money across the entire market, getting tied to everyone, and then everyone spends together, everyone bets together, and if one goes bankrupt, everyone has the potential to go bankrupt together. Ultimately, at the heart of it all, everyone is depending on Antropic and OpenAI to generate profit. And what about their health, what about their profit? It depends on what you consider profit or revenue. If you believe what they declare as annualized revenue, which could be somewhere between $40 billion for OpenAI and $60 billion , $65 billion in the case of Antropic, that's their profit. Then you ask: "What is an analyzed recipe?" It involves extrapolating revenue from a representative period to estimate how much the company will make over a year if current demand remains the same. What period is this? Is this the busiest day of the year times 365? Is this the best week times 52? Is 12 times a good month? It's impossible to know because they don't disclose this information. Most likely, they are choosing the best date, the best possible account, and extrapolating from there. What is the real profit? So, how much are these companies actually making? Ah, that's a huge profit if you ignore how much the company spends, according to the CEO of Antropic. In other words, they simply ignore the fact that they must spend dozens of times more than they receive, because their current business model is unsustainable and their biggest clients are the startups in which they and their investors have invested. And of course, keep buying the stocks when they go public, and as long as the stock price goes up, everything's fine, everyone's making money. And to make this entire debt viable, which only really starts to have to be paid off in 2027—so you already know when the chaos begins—these language models would need to revolutionize the world, in practice. They would need to make jobs, many jobs obsolete enough, to become worth trillions of dollars, which doesn't seem to be happening because their demand remains extremely concentrated in the very companies in which they invested. And there's another elephant in the room, which is the chance that this type of business model , with large data centers running in massive environments, may not even exist in the future. Because right behind the language models of these companies, spending this obscene amount of money in the United States, are the Chinese models, which are much cheaper to develop, much cheaper to run, and open-source models that people can download and run on their computers, and are becoming increasingly popular, more popular than the US models. This shows that instead of using a chatbot like Chat EPT, which depends on a giant data center and a subscription plan that, even subsidized, costs you much less than it actually costs OpenII, people may prefer to run low-cost local models in their companies or homes. To the point that even Silicon Valley companies in the United States are adopting open-source Chinese models and solving their problems without having to subscribe to Antropic or Perai or whoever their intermediaries are. And it is in the long term, now speaking of China, that the factories are implementing and refining artificial intelligence in robotics, production, and logistics, which is in fact the future business or industrial demand for artificial intelligence, in addition to large language models, as the United States government report itself points out , highlighting this enormous weakness and vulnerability of theirs. And there's something else, besides everything else, as I explain in the video about nuclear energy, China continues to grow its electrical energy infrastructure and already has surplus infrastructure to run artificial intelligence and data centers, which in the United States is the major bottleneck in all of this computing. And because of the United States, due to the chip restrictions they themselves created and imposed on China, and because the Chinese later began developing programming language models, they continued developing openly cheaper models that use far fewer chips to run, which ended up generating a cheaper and more sustainable strategy in the long term, while the United States embarked on this mad race, getting involved in the government and betting on spending to win and winning to spend. So, in this loop of investing in what's growing and growing because people are investing, in the name of doing something bigger and better, things are swelling much more than they did in the 2000s until now, with some differences that have the potential to make this bubble even bigger in the United States. Because, in addition to an absurdly greater potential for economic collapse with all this concentration in these companies—unlike Inron, for example, in 2000, which flopped when the dot-com bubble burst and began to be questioned by investors at the time— now companies like OpenAI and Antropic still have some strategies to distract that can make this bubble grow a lot while people are looking at other things. They can, for example, continue to leave their models loose, connected to the internet and with all the computing power they need to be able to hack into other companies' systems, and then say that artificial intelligence is gaining consciousness, escaping, and that it is too dangerous. Therefore, they need regulations to prevent the use of Chinese models in the country. They might also come up with an excuse for the fact that the models haven't improved significantly for years now, despite absurdly higher costs, which is the excuse that it's better to slow down, because it's becoming too dangerous. Therefore, you don't have to say that you've reached a ceiling and have no more money or information available to train a better model. And most importantly, they can continue professing fear, employing this apocalyptic marketing strategy, claiming they are creating a God in the machine, a superhuman intelligence that will change the world and leave everyone rich or poor or unemployed or with whatever . And we have to hope that's the case, because if what awaits us at the end of the construction of these data centers is simply a slightly better and much more expensive programming language model, China has been ahead for much longer without incurring that expense. Hey, but look on the bright side, at least if this happens, the price of the electronics that are currently being sought after for the construction of data centers will return to normal levels. Cell phones should go back to costing less, as should computer chips, memory, and other things. It's just everything else that shouldn't continue as normal. If you'd like more conversations like this, subscribe to the channel here so you receive regular videos and non-fiction, our everyday science podcast with experts, always from human to human, as we navigate this difficult period. Until next time.