There is not a single day anymore without an announcement of some sort from the AI bandwagon. Qualcomm just announced it will start selling its own AI chips (“Qualcomm Challenges Nvidia And AMD With Data Center AI Chips“), AMD announced a new $1 billion agreement with the DOE (“US Department of Energy forms $1 billion supercomputer and AI partnership with AMD“), OpenAI delivered a strong message to the US government about the lack of electricity supply (“OpenAI says U.S. needs more power to stay ahead of China in AI“), and even Saudi Arabia now wants to be part of the AI race (“Saudi Arabia, Rich With Oil, Wants to Be Known as the A.I. Exporter“). All of this news I just mentioned represents only the most prominent stories that hit the wire in the past 24 hours. Every single day, including weekends, the AI bandwagon’s PR machine runs at full steam. Why do you think this is happening? Simple answer: brainwashing of the masses.
Nvidia and OpenAI have been leading the charge since the very beginning, preaching nonstop about the need for trillions of dollars in investments to win the “AI arms race.” However, when they are confronted with the reality that all these investments remain unprofitable after many years, their silence is deafening.
Oracle is right now the best example of everything that is wrong with the current AI narrative. Not only is the company barely making any money from buying and renting even the older generations of GPUs, but it is also taking on an unbearable amount of leverage to serve OpenAI’s future orders when the AI darling doesn’t and will never realistically have all the money to pay for them.
This is how JP Morgan described the current situation after taking a closer look at the reality of the numbers: “Oracle’s stock jumped by 25% after being promised $60 billion a year from OpenAI, an amount of money OpenAI doesn’t earn yet, to provide cloud computing facilities that Oracle hasn’t built yet, and which will require 4.5 GW of power (the equivalent of 2.25 Hoover Dams or four nuclear plants), as well as increased borrowing by Oracle whose debt to equity ratio is already 500% compared to 50% for Amazon, 30% for Microsoft, and even less at Meta and Google. In other words, the tech capital cycle may be about to change.” These words have been, of course, ignored by exuberant investors who continued to bid higher all stocks belonging to the AI bandwagon.
Now let’s zoom out to take a look at the broader landscape. As you can see in this chart, the number of data centers operational in the US right now is greater than all other countries combined and ten times higher than Germany, the second country by number of data centers.
Yet, nobody sees anything wrong with this. Let me ask a question: if there is a global AI arms race ongoing, then why aren’t all other countries participating? The answer is simple, yet the truth of it is hidden from the public because knowledge of it would crash the AI bandwagon stocks bubble in the blink of an eye. In order to build AI models and applications, there are three critical components: hardware, software, and data. The first one is the amount of computing power needed to run training and inference, the second one is the blueprint upon which training and inference are performed, and the last one is the total amount of information available to train the various types of models. There is only a finite number of data points with enough quality generated daily by people and companies operating in the real world. Their growth is rather slow and cannot be forced to be faster to the point that warnings about exhaustion of high-quality data to train AI models already appeared long ago (but they were, of course, ignored): “Elon Musk says all human data for AI training exhausted“. How did AI companies try to work around this problem? By creating synthetic data, with the obvious consequence that AI models started to be more and more prone to “hallucinations,” and as a matter of fact, the quality of all outputs plateaued many months ago. The failure of the ChatGPT 5 launch is the best example of what I am talking about. Then there is the other issue of copyright infringements by AI companies in the process of scraping data from the internet, but that’s a story for another day.
The importance of software in the AI supply chain is perhaps the point that escapes common understanding the most. Despite DeepSeek proving with its open-source model the importance of it already back in January (“China’s DeepSeek says its hit AI model cost just $294,000 to train“), the news has been, not surprisingly, erased by Western mainstream media headlines very quickly. Why? Because the approach that DeepSeek and other Chinese companies have to AI development will inevitably bust the whole narrative about the need for millions of GPUs, thousands of data centers, and a biblical amount of electricity to operate them. When AI infrastructure spending became such a critical component of US GDP growth in recent years (“Spending on AI data centers is so massive that it’s taken a bigger chunk of GDP growth than shopping—and it could crash the American economy“), it should not surprise anyone that the US government has no interest in reining it in.
Wrapping all this up, it is clear that the AI bandwagon led by Nvidia and OpenAI is trying to “brute force” its way to leadership, overcompensating for glaring lacks in the quality of their software. Is this sustainable? If, after more than $1 trillion spent on AI infrastructure over many years and still meager revenues generated, people should have realized by now that the answer is “NO.” However, the show must go on till the very end because the AI stocks bubble is obviously more and more a matter of “national security” for the US.
After the whole public realized the great circular financing scheme I have been documenting for years among the companies belonging to the AI bandwagon—maybe because it’s too obvious and too big to hide now, and MSM couldn’t avoid talking about it—prominent journalists of the WSJ are finally starting to look under the hood and question the whole nature of the revenues generated so far in the recent article: “Microsoft Needs to Open Up More About Its OpenAI Dealings.” Slowly, the whole structure of lies, deceptions, and accounting fraud is starting to fall apart. Something that was already there since the very beginning, but MSM preferred to ignore it, not wanting to spook their prized advertising customers:
- 22nd May 2024: WHAT IF NVIDIA SIMPLY TAGGED ALONG WITH A MICROSOFT AZURE SCAM PLAYBOOK?
- 10th October 2024: THE SMOKING GUN THAT PROVES HOW OPENAI IS MICROSOFT’S REVENUES LAUNDROMAT
Yes, not only have companies part of the AI bandwagon been round-tripping cash among themselves—cash mostly siphoned out from hyperscalers and private investors—but they have also been literally fabricating cash out of thin air in the form of computing credits. This is something that, once again, I warned about long ago in “HOW TO FABRICATE REVENUES FOR DUMMIES (A GUIDE).”
If you want to have a sense of the amount of computing credits printed out of thin air and used to inflate revenues without cash transactions behind them, the best way is to dig into the SEC filings of companies like Microsoft or Nvidia and look for obscure financial statement items like “unearned revenues” or “commitments for future computing purchases.” As you can see from this chart I bring here as an example, the amount of “unearned revenues” reported by Microsoft DOUBLED since their first investment in OpenAI in 2019.
Anyone familiar with basic math and accounting can quickly realize how the growth in revenues reported by Microsoft’s Azure business since 2019, especially in the more recent years, is mostly attributable to the spending of these credits by companies like OpenAI or other AI start-ups Microsoft invested in.
I personally find it ridiculous that it took so long for the likes of the WSJ or the FT to start questioning the basis of the whole narrative they contributed to spreading for years, with articles like “Why Fears of a Trillion-Dollar AI Bubble Are Growing” being more of an example of how these media are now trying to cover their backs rather than genuinely doing something about putting a stop to the whole nonsense, since they still welcome everything being fed to them by the AI bandwagon’s PR machine.
As I said many times, AI will be a useful tool in the future, but similar to how the Internet developed and spread until becoming part of our daily lives, this will take time and, most importantly, business sustainability. With $1 trillion already spent and more than $2 trillion already earmarked for the years ahead (for which there is only a fraction of the funding available as I proved in “OPENAI: LITTLE CASH, BUT PLENTY OF PROMISES, IN AN EFFORT TO AVOID BANKRUPTCY“), the chances of real revenues starting to materialize all of a sudden and making the whole US model sustainable are realistically ZERO. What instead has a high chance of happening is the creation of an immense wasteland of data centers that will be so desperate to attract customers that they will cut costs for computing to the bone, making it accessible also to “garage companies” that don’t have the billions of dollars to pay for it today, but that can instead create the real revolutionary AI tools that will ultimately change our lives in the future like Meta or Google did in the aftermath of the DotCom bubble—once again, another repeat of the Gartner hype cycle.
Of course, the wasteland of data centers will also mean a biblical amount of capital incineration and losses for the broader society, with the US being the one on the spot to account for most of it. Sadly, though, until the music stops, everyone wants to enjoy the party till the very end. A party that, in this case, is already being hosted in a building on fire.
JustDario on X | JustDario on Instagram | JustDario on YouTube
SYNNAX PROMO
Get exclusive intelligence on stocks you can’t find anywhere else – insights that could give you a serious edge. Claim your free Synnax trial month now: https://synnax.app/PROMO/JUSTDARIO