저자: 미상 · 외국인
요약 보기
AI 설비투자 버블 경고. 마이클 버리가 Substack에서 2027년 만기 숏 포지션 레버리지 확대 및 버블 붕괴 시점 앞당김을 언급. 2026년 AI 자본지출 약 5500억 달러, 2027년 1조 달러 전망, 빅테크 부채 급증이 위험요인. 반면 Anthropic·OpenAI ARR 급성장 등 실수요 근거 제시. 결론은 기술 자체보다 투자 사이클 조정 가능성에 방점. 주택 가격 전망 플랫폼 홍보 포함.
다듬은 글
Michael Beugry, the man who predicted the 2008 housing market crash, is now issuing a disturbing warning about the artificial intelligence bubble. And now he is bringing forward his prediction about when this bubble will burst. A prominent investor who made money on "short" positions says he is more confident than ever that AI will pay off in 2027.
And he believes the AI bubble will burst sooner than he initially thought. “I’m essentially changing my timeframe to earlier,” Bugry wrote in his Substack investment newsletter on Monday. So I want to have more leverage in my “short” positions, some of which are now scheduled to expire in September 2027.
This means that Bugry is probably confident that the AI bubble will burst by then. And, my friends, if the AI bubble really bursts, it could be a disaster for the stock market and the US economy. Because right now, these tech companies are spending more on AI development than ever before in the entire capital spending cycle in US history.
Because, according to the Wall Street Journal, the development of AI is becoming the largest economic bet in US history. At the same time, it is predicted that spending on AI data centers will amount to 3.6% of GDP in the future. This is bigger than the telecommunications and fiber-optic bubble of the late 1990s .
This is more than the investment in interstate highways in the US in the 50s and 60s . This is much bigger than electrification. It is also larger than the investment in railroads in the late 1800s. This means that the US economy is becoming inextricably linked to these large hyperscale technology companies like Google, Meta, Oracle, Microsoft, as well as cutting-edge labs like OpenAI and Anthropic, which are spending all this money on GPUs and building data centers.
In fact, Goldman Sachs just reported that global investment in AI is projected to exceed $1 trillion by 2026, with the majority of that investment coming from the US. The US spends more on AI investments than the rest of the world combined, and this investment comes from about six or seven large public US companies.
And the danger to the economy lies in the cyclical nature of how AI investments fuel the stock market and how stock market valuations affect consumer spending. Let me break it all down for you right now so you can understand what can happen. The massive development of capital investment in AI, which in 2026 in the US is about 550 billion and is expected to grow even more by 2027, is inflating stock market valuations for many different companies, especially technology ones.
These record high stock market valuations are causing a heightened wealth effect in the economy and stimulating consumer spending growth even in the face of inflation. This higher consumer spending is fed back into companies, further increasing their revenue, improving profits, and causing their stock valuations to rise even more.
So if you knock the brakes on AI capital spending, and if some of these hyperscaler companies start to struggle, especially those that use debt to finance their data centers, it could have a domino effect, causing stock market valuations to fall and consumer spending to decline. Of course, Michael Bury has been talking about this for a long time.
He is pessimistic for almost the entire year of 2026. In May, he said that the situation on the stock market was reminiscent of the last months of the 1999–2000 “bubble” . However, since then, over the past four months, the Nasdaq index has continued to rise , and last week it closed at a record high.
So far, stocks have continued to rise in price, leading some to question whether this is a " bubble" at all. So some people are now saying that maybe AI is the best thing that has happened to the economy in a long time, and maybe it will actually contribute to further growth in corporate revenues and, as a result, GDP growth.
In fact, the revenues of the two largest AI companies, Anthropic and OpenAI , have grown rapidly over the past year. As you can see from the Ticker Trends study on Substack, Anthropic's annual recurring revenue has grown to approximately 76 billion in the last 6 months alone. According to these estimates, you can see that it has increased tenfold in the last year .
Currently, Anthropic is still a private company. We don't know for sure what these numbers are. They are expected to go public by the end of 2026, and according to Polymarket, the probability of this happening is 81%. Anthropic, the company developing the Claude model, is going public by the end of this year.
If they go public, we'll know those revenue statistics for sure, but right now it's around 76 billion ARR, which proves real revenue generation from building these data centers. This is not a case where money or value is absent. Additionally, Ticker Trends research shows that OpenAI's annual recurring revenue has reached $50 billion.
OpenAI runs ChatGPT, and those are just crazy numbers, friends. If Anthropic and OpenAI truly have around $70 billion in ARR, they would be perhaps the fastest growing companies in history in terms of revenue scaling . And again, this is important because it demonstrates that building data centers and training models brings some economic benefit.
It is clear that people and businesses pay Anthropic and OpenAI for their modeling services. The question is how durable and sustainable is this in the future? After all, with major technological breakthroughs, it often happens that over time a deflationary component appears , and these things become cheaper.
And if the value of AI tokens declines and GPU chips become cheaper to produce over time , it could cause a bubble to burst given the amount of debt currently financing AI development. For example, according to Vanguard, the debt of hyperscalers—Alphabet , Amazon, Meta, Microsoft, and Oracle —will grow to over $150 billion by 2026.
You can see that just a few years ago these companies were borrowing almost nothing . This is more than a tenfold increase in the last two years. And now their borrowings account for almost 11% of all issued US investment debt. And these capital expenditures for the development of AI data centers are projected for years to come.
They say that in 2026 we had 777 billion. They believe that in 2027 we will reach a trillion. They predict that we will remain at the trillion level from 2028 to 2030. And that's really the main point of contention, is n't it? Will these capital spending plans continue as projected? Because if not, if they go down even a little, the stock market and the economy could face a real problem .
As we see the price of AI tokens—the technology that runs in the background when you chat with ChatGPT or Claude—the tokens that process the information and your responses, their prices are plummeting. They are reaching new record lows. According to a CNBC report using Silicon data, the value of the tokens is halving.
This potentially means less revenue for Anthropic and OpenAI, which could mean less financial feasibility of building data centers, which could lead to the cancellation of these projects. And I think this is all very exciting, friends, because AI, infrastructure development, and data centers—this is a very controversial topic, is n't it?
And people often have extreme views on this. Some believe that the development of AI and data centers is the worst thing that could happen. And that's really exciting, my friends. Bernie Sanders, along with Greg Cazar, a congressman from Texas, just introduced a bill to ban artificial superintelligence and temporarily halt the development of advanced AI.
This act would permanently ban the development and deployment of superintelligent AI, and would also oblige the US to seek international agreements to prevent its creation in any other country in the world. The idea is to ban superintelligence, so that no individual or organization can develop or deploy AI systems that surpass human intelligence or are capable of overthrowing human governments.
The penalty for this is up to 20 years in prison for those who violate these rules. So, that's just one side of the coin, right? Some people just want to stop it in its tracks and maybe even make it a crime. On the other hand, there are many people who believe that AI is not that cool, and in fact it may just be a "bubble." Many of you think so, because I conducted a survey and asked: do you think AI is a "bubble"?
3,000 of you have already voted in this poll in just 9 hours. 72% of you think that yes, it is a "bubble." Only 18% think not, and 10% are not sure. I think another interesting thing that suggests a bubble of capital investment in AI is the number of data centers that already exist in the US. A statistic that simply shocked me: how many data centers are already operating in the US and how it compares to the rest of the world.
Today, there are already 4,700 data centers in the United States. And, of course, a huge number are still planned to be built. Let me remind you, friends, a data center is that large building that is built to house all the graphics processors and servers inside. It's the GPUs and servers that cost the most money.
Approximately 85-90% of the cost of a data center is the graphics chips and servers inside. The exterior design of the building itself is not that expensive. In second place is the United Kingdom—568. In third place is Germany—533. In fourth place is France—393. In fifth place is China— 376. In sixth place is India—300.
The crazy thing is that the US has more data centers than the next 17 countries combined. Absolutely incredible . So, you know, you don't even need to be a number expert or have deep knowledge of AI to understand that such graphs look and smell like "bubbles." And when you see spending reaching a certain percentage of GDP, an all-time high, higher than when the railroads were built, it looks like a "bubble." However, when this "bubble" bursts, it will not necessarily mean that AI is irrelevant or useless.
This would simply mean that there is a change in the capital spending cycle in the US and globally. And these changes in capital spending cycles will potentially cause stocks to fall . This may lead to lower consumer spending, but that doesn't mean AI isn't valuable or real. And to understand this, just look at what happened to the railroads in the late 1800s in the United States.
This chart from the Federal Reserve Bank of St. Louis shows the number of miles of railroads built in the United States from about 1830 to 1950 . I mean, what an incredible data set this is. And what you can see is the ups and downs of railroad construction, peaking in 1872, and then the collapse of construction volumes.
And, in fact, this crash from 1872 to the late 1870s caused a depression in the United States. That collapse in railroad construction caused the Panic of 1873, which escalated into a financial crisis. And this was actually called the Great Depression in the US until another Great Depression in 1929 and the early 1930s.
You had bank failures, skyrocketing unemployment, railroad companies closing, and it seemed like the world was coming to an end. It wasn't until 1879 that railroad construction picked up again, actually peaking in 1882 when nearly 12,000 miles of track were laid. Then came another downturn, leading to the depression of 1882–1885.
Business activity fell by 33 %. Then the railroads resumed again, reaching a peak in 1887, when over 13,000 miles of track were laid. Then another correction, which led to two recessions in the late 1880s and ultimately culminated in the Panic of 1893, which was one of the worst recessions the United States had experienced before the Great Depression.
The unemployment rate could reach 18%. And to be clear: I'm not saying we'll see a repeat of this with the AI capital spending bubble. I don't necessarily think unemployment will hit 18% or that we'll see a depression when capital spending on AI inevitably slows. However, I think what this proves is that large capital spending cycles and industrial bubbles typically experience a boom, then a contraction, then a boom again, and I wouldn't be surprised if that happens with AI and data centers.
And if at some point, maybe in 2027 or 2028, construction slows down, the stock market crashes, unemployment rises, maybe a two-year recession begins, but then a few years later, construction booms again. And I think that this is a pretty appropriate model for considering such a situation. There is no black or white, no zero or one , AI is not the worst thing in the world and not the best.
This is a productive technology that brings real business revenue to the economy and will likely provide greater efficiency for the American people and workers in the future. But current investment levels are the highest ever and are likely not sustainable. And when they fall, there could certainly be some correction.
And when a correction occurs , it will likely affect the housing market. To find out where home prices are heading in your area and whether now is a good time to buy a home, go to www.reventure.app. This is real estate software I developed to help you understand the dynamics of your local housing market.
Of course, what happens in the economy has a big impact on home prices, and we at Reventure track all of this data for you. We generate a 12-month price forecast so you know where values are headed in your area before you buy, so you can make better deals and avoid overpaying. Go to www.reventure.app now and upgrade to our monthly plan, for just $39 per month, to access all of this data and prepare for your home buying or selling decision.
Comments 0