저자: 미상 · 외국인
요약 보기
엔비디아 등 빅테크의 AI 데이터센터 투자가 Cisco식 과잉 인프라 투자 버블일 수 있다는 마이클 배리의 주장을 소개. 감가상각 기간 연장(3~4년→6년 이상)으로 이익이 부풀려졌을 가능성을 지적. AI 투자 조정 시 빅테크·고베타 자산 전반 매도와 유동성 축소로 비트코인 등 크립토에도 영향 가능성을 언급. 다만 배리의 주장을 시장 전환 신호로 보지 않고 리스크 프레임워크로 활용하며, 레버리지 자제와 장기 보유분·투기분 분리를 권고.
다듬은 글
We are not Enron. That's exactly what Nvidia just said. But whenever a company has to compare itself to one of the biggest accounting scandals in history , you know there's at least enough nervousness in the system to make it worth taking a closer look . But before we do that, this video is powered by Best Wallet , a multi-in wallet that I use to hold my long-term crypto .
More on that later in the video. So, all the world's major corporations, with very few exceptions, have just gone all in on artificial intelligence. They are investing billions, sometimes even trillions of dollars over time, in data centers, chips, and of course infrastructure. The big winner of this wave is, of course, Nvidia.
Incidentally, Nvidia is no longer primarily a gaming graphics card manufacturer. They have become the heartbeat of modern data centers. If you look at their sales figures, the gaming segment is now almost a side issue. The focus is now clearly on data centers and, of course, AI. And because BigTech's weighting in major indices is so massive, Nvidia, along with a few other hyperscalers, has become not just a stock, but a systemically important asset for the global economy.
If this AI bet goes wrong, it's not just a problem for a few nerds in Silicon Valley; it also affects ETF investors, pension funds, sovereign wealth funds, and ultimately us as crypto investors. Because, according to various analyses, a large part of the current growth in the USA is driven by AI-driven investments.
I'm sure this isn't the first time you've done this. But if this engine sputters, we're not just talking about a few red candles at Nvidia, but about the risk of a real growth spurt, perhaps even a multi-year recession. The intriguing question, therefore, is this: Were they really right in saying they were not Enron, or is the whole thing perhaps more like a second Cisco, i.e., a massive infrastructure bet that will eventually collapse if demand falls short of expectations?
And this is exactly where Michael Berry comes in, the guy from Big Shot, the man who shorted the housing bubble back then, now writes about these very topics on his own blog, and Nvidia felt compelled to respond directly to him . But what is a core argument? He's basically saying that we remember the dot-com bubble incorrectly, because if you ask people today what it was about back then, you mostly hear names like pets.com or some .com startups with no revenue.
So, purely fantasy stories with unrealistic ratings. Defenders of today's AI bet say that back then there were no real winners, but today there are . Google, Microsoft, Meta, Nvidia, they all earn billions. So everything should actually be fine. But at this point, Barry says: "Wait a minute, because the dot-com bubble was n't just a meme festival without any revenue; the truly dangerous part was a massive infrastructure overbuild by profitable companies.
The best example of this is Cisco. At the time, one of the most valuable companies in the world. They sold routers, switches, network technology—various fundamental building blocks of the internet. The narrative was that the internet is the future, so we have to build a massive amount of infrastructure now.
The result was that far more was built than could be used in the short and medium term. Capacity, lines, hardware—everything was maxed out. The implementation looked great in the short term , but it was a supply-side bubble, a so-called 'supply- side bubble,' as Barry so aptly calls it. And he says that's exactly what's happening again right now, only this time with AI data centers, GPUs, and cloud infrastructure.
The big hyperscale companies—Google, Microsoft, and Amazon—are throwing themselves headlong into AI, building gigantic datasets, and pushing more and more new GPU generations from Nvidia into their warehouses. But the direct cash flow from AI..." Compared to capital expenditures (Capex), AI is still in its infancy.
The market isn't hyping today, but rather the prospect of an extremely profitable AI future. Now comes level two of Barry's critique, which concerns accounting. It's a seemingly dry topic, but one that can be massive: depreciation. GPUs are the backbone of Nvidia's business model. These are the chips that train all these large models and later run them in operation, i.e., in inference.
Barry examined how the major players treat these GPUs in their books. Previously, such data center sets were typically depreciated over three to four years. Now, some large companies have extended the useful life to six years or more . But what does that mean in practice? If you depreciate the same chip over six years instead of three, you spread the costs over a longer period.
This makes the profits in the early years appear higher. On paper, everything now looks more profitable than it actually is. This is problematic, of course, because this only works if the components actually last that long in reality. are economically usable, as prescribed or as depreciated. But if technological progress brings you a new chip generation every year that is twice as efficient, then the old parts often become economically worthless much faster, even if they are still physically within the inventory.
Now things get interesting again. Either you yourselves communicate quite openly that you want to further increase the pace of development. But you now want to bring new top generations to market every year with much more performance and better efficiency per watt. On the one hand, you say these GPUs easily last 6 years or more, they are totally stable in value.
On the other hand, the CEO says, at least in essence: "Our chips are getting better so fast that it is no longer worth working with old chips for many customers." And this is exactly where Burry's criticism comes in, because if the actual economic useful life is shorter than the depreciation plans suggest, the profits are inflated and the balance sheet values of the assets are therefore far too optimistic.
This either does n't turn it into Enron, i.e., not pure accounting fraud. Invented revenues, but quite frankly, it could lead to a massive valuation overvaluation , similar to what happened with Cisco. And now, of course, the question that interests us all most: What does all this have to do with Bitcoin and crypto?
First, Nvidia, BigTech, and AI packages are a huge part of what's driving current stock and index valuations. Many ETFs, pension funds, and Force, including in Europe, are disproportionately invested in these companies. So, if it turns out that the return on these gigantic AI investments is lower than expected, or that profits have been inflated through aggressive write-downs, it could lead to a major revaluation .
In other words: falling BigTech share prices, risk reduction in funds, less appetite for anything high-beta , i.e., crypto. The so-called second-round effects could also be brutal, because if large portfolios now reduce risk, the decision is often not whether to sell Nvidia or Solana, but rather a blanket reduction of risk.
Tech, growth stocks, and crypto—everything gets thrown into the mix. dumped and sold off. The second important point, however, is this: AI Capex is a liquidity issue. As long as AI is celebrated as the new big growth narrative and the stock markets reward this narrative, financing conditions for risk assets tend to be better.
Capital flows into Force, Force, buying, and tech and crypto. Valuations support each other. But if this narrative breaks, for example, because word gets out that data centers aren't generating reasonable returns or because depreciation issues surface, then this wave of liquidity can quickly turn. And the third point is that the AI cycle influences sentiment towards future technologies in general.
If AI is suddenly perceived as an overhyped New Economy 2.0 , this can also have a short-term impact on crypto. In the long term, one could argue that if classic tech cycles are disenchanted again, a scarce non-lucrative asset like Bitcoin actually gains in attractiveness. But the path to that can be volatile and painful.
But one more very important point: Barry is n't saying that AI It's fake. He says there could be another infrastructure overbought situation, with far too much construction happening far too early, on a far too large scale, combined with accounting tricks that neatly mask the short-term effects . And if he's right, if this is true, the risk isn't just that a few textbooks will correct it.
The risk is that a large part of current GDP growth and market valuations are based on something less stable than it appears. So, for us crypto investors, this means the following: The fact that someone like Michael Berry has a good argument right now does n't mean the market will turn around tomorrow.
Michael Berry has predicted ten recessions in the last five years, it seems. If one of them happens, everyone celebrates again. If most don't, everyone quickly forgets. Especially when shorting the stock market, you can be right about something, but still end up pretty poor. Particularly if you bet too early and too aggressively on falling prices.
The second point is, for me, such analyses are clearly not a game-changer. It's not a short signal, but simply a risk framework that I keep in mind . If AI companies like Kapacks and Nvidia continue to go haywire, you'll want to examine the situation more closely , determining which part of your portfolio you're consciously willing to take more risk in and where you generally prefer to remain more defensive, for example, with a long-term Bitcoin hotspot stack without leverage, instead of going all-in on the next AI speculation.
Thirdly, for crypto, in my view, this means: leveraged trades and moonback strategies should always be considered within the context of this macro perspective. If you 're currently in a phase where AI, Big Tech, and growth are being hyped to the max and valuations are still based on perfect success, you should perhaps think twice about how much additional risk you want to add in the altcoin sector .
And that's precisely where structure comes into play. If you're serious about investing in Bitcoin and crypto , you need a clear separation between your long-term holdings ( hodling), your speculative holdings (trades , narrative plays, etc.), and your other holdings (trades, narrative plays, etc.).
AI coins and L1 betting. I use Best Wallet, among other things, as the base for my long-term portfolio. Bitcoin, Ethereum, Solana, and other core holdings are stored there. So, if I think the market offers good opportunities , I buy more on the exchange and then transfer it back to Best Wallet to keep the long-term portion clearly separate from the speculative portion.
As always, you can find the free wallet link below in the description. I don't believe Nvidia is Enron. I'm sure it 's not a complete scam, but the parallel to Cisco—a story based on real technology that, at times, overextends itself through massive overbuilding and excessive optimism—is certainly worth considering.
But whether this AI gamble ultimately turns out to be a new industrial revolution or an expensive overreaction with a long trough, nobody can say for sure today. What you can control, however, is how much your portfolio depends on this gamble paying off perfectly. Bitcoin, crypto, AI—everything hinges on global M2.
Liquidity, risk appetite, and of course, these balance sheet valuations all come together. The better you understand how these factors intertwine, the less of a pawn you are and the more of an active player you become. If this video helped you, please leave a like and let me know in the comments whether you see Nvidia and AI as a new super bubble or a legitimate long cycle with interim cash flow.
And if you want more of these Marrotech crypto insights, subscribe to the channel and turn on notifications.
Comments 0