Transcription
The AI bubble may have just ended without ever popping. Unless you've been living under a rock, you are aware that stocks like Micron have been going parabolic recently because of the massive demand for memory that is driving prices of memory skyhigh and it seems like there's no end in sight. Stocks like SanDisk are doing the exact same thing. While stocks like Microsoft are falling off of a cliff and stocks like Amazon are seemingly going nowhere.
This has led many people to the consensus opinion that AI is a giant bubble. All of these massive companies are spending unsustainable amounts of money on the AI race and that there is no way this could ever possibly result in profits. And therefore, the consensus opinion is the narrative that seemingly everybody believes is that this will end in a terrible popping of this huge massive bubble.
However, Bloomberg just released a report that shows revenue may have actually just passed the point where this whole thing is now sustainable. In other words, not a bubble. They show that global AI sales reached $25 billion in the first quarter of this year, which exceeds the industry's estimated $21 billion in depreciation costs. And even though the margins are thin, this specific milestone suggests that AI companies may be beginning to cover the cost of their capital spending.
This is a chart that goes back to quarter 1 of 2023. And the top line on this chart shows capex depreciation. In other words, the amount of money that these large companies are losing. Whereas the orange line underneath represents the amount of revenue these companies are collecting in return for what they are spending. And you can see this entire time they've been negative. However, recently they have crossed a tipping point where quarterly revenue exceeds capex depreciation.
This is an extremely important finding to pay attention to as a signal because like I said, it seems like the consensus narrative out there is that this whole entire time all the spending has been completely unsustainable. There's no way it could ever get to the point where it was sustainable. But it seems like we've already gotten there.
The figures in this chart were based on a data set that tracked spending on AI across more than 1,000 companies. So looked at company filings, executive statements, press reporting, cloud provider disclosures, and because some of these companies are private, it is difficult to get exact numbers. However, even if there are inaccuracies in the data because of estimations, it does seem to be at least directionally correct. That is because all these numbers continue to improve over time.
Now, for those of you who have been paying attention to this space, you're probably thinking, well, how is depreciation being counted here? And this analysis does assume a six-year depreciation life. Many people have argued that this is wildly optimistic to the point of being just completely inaccurate. And that is because of the rapid pace of chip innovation which renders old hardware less valuable within much shorter than 6 years. But the reality is that just because there is a new chip that is better, it does not render the old, worse, slower chips useless or completely without value. That is why Nvidia's 4-year-old H100 chip has still kept most of its value. Hourly rental price remains at nearly 80% of its launch level. You can see over the past couple of years, it has drifted down somewhat, but it doesn't hit zero. And even Matt Garmin from Amazon Web Services said that Amazon has still not retired their 6-year-old Nvidia A100 servers due to continuing demand. And that is because in a very similar way to chips advancing, becoming more powerful and more efficient, we also see a similar trend with the models. Open source and Chinese models like Deepseek mean that older chips can maintain their usefulness maybe much longer than many people expected was possible.
The reality of bubbles is that a true bubble rarely emerges out of pure free market conditions. In fact, basically all of the biggest true bubbles in history are fueled by money printing. In fact, going all the way back to what many consider the original bubble, which was the tulip bubble, where tulips, the flower, rose in price 26 times in January of the year 1637. And then when that bubble popped, they fell to 1/20th of their peak price just a week later. When you look into it, you find that it was not the product of irrational speculation. It was monetary and financial distortions. This is because the Dutch Republic at that time was Europe's commercial hub and Amsterdam had recently launched their own central bank. They began debasing coinage created abundant liquidity which fueled broader asset price inflation as well as fueling the tulip bubble. And you can find very similar stories coinciding with all of the major bubbles throughout history. An obvious recent example is the massive money printing that the Federal Reserve and the US government did during 2020 and 2021 and the massive bubbles that that fueled from things like NFTs to crypto to stocks as well as even consumer prices.
However, what you don't usually find is a bubble that emerges out of pure free market conditions. And the reason why is because the pricing mechanism, economic calculation, and profit motives are always there in free markets. Those only get distorted and washed away when you have massive credit expansion and money printing where all investors are trying to rapidly and feverishly find a place to dump their money to get a good return in order to stop themselves from being debased but also to take part in the rapid speculation that everybody is seemingly making a lot of money from. And we don't have that right now.
We have companies like Amazon spending unheard amounts of money on capex. Microsoft spending unprecedented amounts of money as well. Meta to the point where just the four large hyperscalers alone are expected to spend a combined $5.3 trillion of capex spending by 2030. These are not companies that have just been handed massive amounts of newly printed money by the US government or by the central bank. These are companies that see a real opportunity to make real profits and they're going all in. And it looks like the data is already starting to show that that bet is very likely to be a profitable one.
I want to be clear, nothing is certain in markets. And the consensus is usually right. That's why it's the consensus. But as Ray Dalio likes to say, in order to make money investing, you have to bet against the consensus and you have to be right. And considering that just the Magnificent 7 alone have underperformed the S&P 500 for almost a year and a half now, largely due to the consensus that their spending is nowhere near anything that could turn a profit for any long period of time. It's possible that that opinion may start to turn around soon, which may be an opportunity. But don't take my word for it. As always, do your own research and make sure you keep the number one thing the number one thing, which is risk management. Make sure that no matter what you choose to do with investing, you are covering your downside, limiting your risk, making sure that no matter what happens, the most you can lose is a little. Thank you so much for watching. Have a great day.