원문 보기
Greetings, dear investors. The situation is great for Amazon, Google, Microsoft, while Meta is a little worse, but let's discuss the risks and rewards of investing in these companies today. We discussed this in a video specifically dedicated to Google. You can find it at the link in the description below.
And Google had great profits. Stocks reacted somewhat positively thanks to the huge growth in cloud services revenues. So everything looks good, incredible growth, amazing business, you could say . However, if you dig a little deeper, you can notice discrepancies . Interestingly, you can use Google's AI to look beneath the surface.
And I asked what share of Google's cloud business is Anthropic and OpenAI. 40% of them are from Anthropic. That's $200 billion in liabilities for Google: backlog with other cloud services like Google TPU, and Google's spending to meet the needs of businesses like Anthropic. And if you look at the big picture, Anthropic and OpenAI together account for about half of the two trillion dollars in order book of all the major cloud providers: Amazon, Microsoft, Google and Oracle.
And all of these companies will be investing roughly $200 billion a year, as we'll discuss, and they're betting that customers like OpenAI and Anthropic will recoup those costs. Of course, you grow by 13 billion a year if you spent 50 billion on capital investments , AI research and everything else.
But these businesses are not yet profitable, perhaps never will be due to competition, and they constantly raise money for their own needs , which then goes back to Google. Likewise Nvidia, excellent performance. The stock hasn't performed impressively over the past year, there have been ups and downs, it's a little higher now, but still a decent 15%.
The P/E ratio is only 30 with stunning growth, incredible development, and almost doubling of revenue. However, to continue growing, Nvidia needs to finance customers to buy Nvidia chips at higher prices. No negotiations, nothing. They need to fuel this demand, limitless demand, so that everyone keeps buying.
And here comes the funding cycle, which means that most of this revenue and AI growth is not currently sustainable. This is a huge adventure. Another example: a lease for 50 billion , first 20, then 30— all of this goes into the order book. $750 billion bet on AI, a cyclical tech bubble similar to the one we already saw during the dot-com crash.
Companies need to maintain this growth rate because everything is based on one rate—now it's AI, and they give money to firms that in some way, building data centers or whatever, buy or use their own products. This is called cyclical financing. There is no real profitability in what they do. So this is a purely cyclical nature of the deals, also with SK: here is 350 billion for the purchase of chips for OpenAI.
This is practically buying from yourself and increasing your income. Now let's move on to Microsoft. Huge stock reaction. If you look at the fourth quarter results, they are excellent . 32% growth in the intelligent cloud. But if I look at the balance sheet a little bit, fixed assets—if you check here, they went up from 9 billion, from 109 billion to 313.
If you look at depreciation and stuff, it only went up by 10 billion. If what they're investing in and their data centers don't have a long lifespan, especially with new Vera Rubins coming out every year, if they were to amortize that 313 billion over 5 years, the cost would be 60 billion per year, and they're at 40 billion now .
So that will grow over time, especially as they invest more, and in just one quarter, they invested a new 50 billion. And why are n't they investing in Anthropic? This is for OpenAI: a staggering $625 billion in contracted future cloud revenue. We'll discuss Oracle, which grew rapidly last year on these orders and then rolled back, but okay —cloud orders again: OpenAI, which does nothing with the cloud, spending little money—is Apple, and the market is punishing it for it.
However, Apple has a P/E ratio of 47, which is expensive in itself. However, if you look at the capital intensity, they only spend 14 billion on everything that happens. So, Apple is not betting on AI. Very interesting. Amazon: huge growth, huge costs. Again, the same principle. 20% growth in net profit.
OK. Because of their investment in Anthropic, 53 billion. Perfectly. OK. So, huge growth there. However, there is also a 37% increase here. Operating profit is growing rapidly. But then again , Anthropic is a massive catalyst . And Anthropic is again not profitable. If we look at Amazon's cash flow, it's negative because the company is investing huge amounts of money: $170 billion in the last 12 months alone , and it's going to grow to $220 billion over time because everyone is doing it, and we have to spend more than Mark Zuckerberg, who, again, is the kid that nobody wants to play with , so he has to compete, but he just ca n't.
The market doesn't like it. As a child, his sisters also refused to play with him . What is, is what is. But it will raise spending to 130–145 billion. So far, he has spent only 51 billion. Here we already see that expenses and expenditures have increased by 55%, and income by 28%. So actually operating profit is falling, margin is falling, and here we show, " Okay, this investment is not as profitable unless you have a direct cyclical customer." Meta also plans to spend a lot, and this is where we will see the true profitability of AI, of course, when it is not cyclical.
Mark says that building the infrastructure for superintelligence is key to making the benefits of this technology accessible to everyone, and that's the main point: great for everyone, AI for everyone, but we invest for profitability. Shareholders invest to get rewards, not for everyone. So, there is some ambiguity here, but okay, let's move on to the other side of things - off-balance sheet assets.
Meta invests with BlackRock, 80% of the funds will be invested by BlackRock. Meta will invest 20%, and then they will see what to do. So, in addition to all the investments that we discussed—200 billion per hyperscaler— there are also off-balance sheet investments. Here we have hyperscalers, expected at 700 billion in 2027.
Then there is hidden debt. Hyperscalers currently have approximately $1.65 trillion in off-balance sheet debt in the form of hidden liabilities. How does it work? Specialized companies, capacity purchase and leasing agreements, higher costs. Here's how it happens. Nobody wants to talk about it because it's not a fashionable topic right now.
The costs of this will only become noticeable after 5 years, and until then, it is an eternity for investors or market speculators . In any case, if you look at the companies, it's hundreds of billions in such deals, up to 1.65 trillion. Okay. And if I look at 1.6 trillion, 700 billion a year, things of that nature.
If I look at the revenue over the last 12 months for these companies, look at this. Capital expenditure per revenue or profitability of cloud technologies: Amazon is in the market, Google spends 2 times, and Microsoft spends 3 times its cloud revenues. This is total income. Meta spends 10 times its cloud revenue on capital expenditures, including cyclical funding.
This is something very, very dangerous . And if you look at the leases, that 's $800 billion in contractual procurement obligations for Alphabet, Meta, Microsoft, Amazon; Oracle has a little less of them, but they have n't started yet. All of this is a bet on the success of AI. And the biggest risk, in my opinion — and this might surprise you — is that AI actually works.
If AI works, someone will ask: how to get the same thing for 10% of the cost? When AI reaches the level of general intelligence , all these investors will see a completely different picture. And this could change everything, because technology is changing. China is investing in the development of AI.
No matter what China does, you don't want to be seen in a Chinese car, but these cars are better, and you won't care where your AI comes from as long as it works and you pay less. Now everyone is experimenting, and it's all funded through this kind of " cyclical" funding. So few people ask: is there a real return on this AI?
That's the problem. The conclusion regarding risks and rewards is that none of these businesses are what they used to be. They used to be efficient cash flow generating machines like Apple. Low capital expenditure, high profitability, high margin, they just printed money, engaged in share buybacks, etc.
Now these companies are high- cap investments, betting on two clients: Anthropic and OpenAI, which have unstable business models. That's the adventure with AI. How does this unravel? Nvidia's revenues are recognized now, and expenses are recognized in 5-10 years. This is data from Goldman Sachs . If we look at the annual depreciation of computing power based on the lifespan of the chips, by 2031 we will have almost a trillion dollars in depreciation per year, which is crazy because the return on investment is unknown.
They need hundreds of billions in profit on this trillion in expenses to recoup the funds already invested. Where will they come from? I have no idea, look at Alphabet's stock price. It took off on the AI bubble to justify a market cap of $4 trillion. I always confuse billions with trillions. You need to double your capital.
You need to get to almost 9 trillion. To have a valuation of 9 trillion, they need to generate 400 billion in net profit over 5 years. This is more than current income levels. Who is crazy here? Microsoft is again about the same . Meta, where will we get all these profits from AI? It doesn't work yet, and then someone will make it cheaper.
Another AI hyperscaler, like e-commerce once was. The e- commerce boom happened while everyone was staying at home. Then it stabilized, and now we're in an AI situation. Same with Microsoft. But this Microsoft graph may reveal something else. Those of you who have been investing for less than 10 years will not know this.
But who remembers the early 2010s? Everyone hated Microsoft. The shares showed no movement. After the dot-com crash, stocks didn't make a profit for 15 years. It's very likely that the same thing will happen to hyperscalers, if you're very , very lucky. There simply is n't enough income and cash in the world to justify the investment, and the best-case scenario is already priced in —priced higher than perfection, and that's never happened before in history .
So I'm just saying it's too risky. Of course, as stated in this comment, I said that Apple is overvalued at 280. Now it's at 340. It's a little less now, but okay, in a bubble, a bubble can continue to grow. It was decreasing. Now we have received reports of fictitious revenue growth and the price is rising again.
I won't predict stock prices, they could double in six months. I'm just saying that the finances—if you take a calculator and calculate where Google is going to get 400 billion in profit, above and beyond all the depreciation, the round robin funding, and everything else to justify it—it's just a 1% chance.
This is what is included in the price. So these things are not for me. We will stick to the cost. You can check out what I'm doing on my research platform, and I'll see you in the next video.
Comments 0