I lost count of how many articles, posts, and analyses I have written over too many years documenting what was REALLY happening in the AI space, as opposed to what was happening according to the fantasies and stories broadcast non-stop by all actors with various degrees of vested interest in spreading the gospel. As I wrote last week in “ON WALL STREET, NOBODY CARES ABOUT CRUDE OIL; THEY ALL CARE ABOUT WHEN THE AI BUBBLE IS GOING TO BURST”, this is what people completely overlooked for so many years:
“The AI narrative won Wall Street not because the economics made sense, but because it offered a story strong enough to justify unlimited capex, unlimited leverage, and profits that only exist in presentation decks. That story survives as long as markets can pretend the funding stack is stable and the growth curve is inevitable.”
On Sunday evening, the WSJ published this exclusive that, once upon a time, would have been welcomed with plenty of joy by investors: “Nvidia in Talks With OpenAI to Guarantee $250 Billion Financing for Data Center”. This was the comment I immediately posted on X on the matter:
Feel free to look at the comments on that post, many of which were from the large crowd of Nvidia fanatics that haunted me for years. A behavior that, unfortunately for them, only strengthened my motivation to carry on while others threw in the towel as the Nvidia stock price went higher.
Why did Nvidia shares close on Monday 5% lower? Not because I was right (that has been the case for a very long time), but because investors are now seeing what I see, and the reality is quite scary once you take a closer look at it. However, there is still a secret well kept by Nvidia and all companies highly vested in keeping the whole AI house of cards standing, something that, of course, I have warned about so many times already: The vast majority of GPUs that Nvidia “sold” in the past years aren’t operational in datacenters, but are collecting dust in warehouses of hyperscalers, neoclouds, ODMs, and Nvidia itself.
The math is very simple, yet brutal: just compare the amount of datacenter capacity built since 2022 with the number of GPUs Nvidia claimed it “sold” over the same period of time, as I did. From an accounting perspective, this is why many of these GPUs and related tech purchased by hyperscalers over the years aren’t yet being depreciated. That equipment will only start “losing value”, from an accounting perspective, when it is used. Hence, there is no need to mark it to market (because it isn’t meant to be resold), and depreciation (hence costs) is not showing up in the financial statements.
Imagine what would have happened to Nvidia shares if its sales of GPUs had slowed down as they should have. This is why Jensen Huang pushed the narrative and the circular financing scheme I exposed to its core over 2 YEARS AGO in “MELLANOX, THE CORNERSTONE OF NVIDIA-MICROSOFT REVENUES ROUND-TRIPPING SCHEME” to the extreme. A few months before that, I wrote a guide, “HOW TO FABRICATE REVENUES FOR DUMMIES”, that offered a blueprint to all those companies that wished to follow Nvidia and Microsoft’s footsteps. Because of fear of missing out, combined with a lack of real revenues showing up despite the biblical amount of investment, circular financing and revenue round-tripping was the only way to keep the illusion going while convincing investors and lenders to support the ever-growing need for REAL CASH to support ballooning capex expenditures. Is this fraudulent behavior? Is this a Ponzi scheme? Or is this just a case of collective stupidity that poisoned the minds of some of the (in theory) brightest business people in the world? You know what I think about this.
After years of partying and out-of-control euphoria, regulators not only did nothing to rein in excessive speculation that translated into companies priced at asinine multiples, but also supported the euphoria by authorizing an indecent number of double-levered, triple-levered, and all sorts of highly speculative schemes and ETFs. Wall Street banks could not manufacture them fast enough to feed the insatiable demand of both retail and institutional investors, even if they knew very well that all those investments were doomed to be wiped out at some point. What happened recently in Korea, with hundreds of thousands of investors losing everything (South Korea’s Stock Market Deleveraging Storm: 1.2 Million Accounts Face Margin Calls, 360,000 Retail Investors Wiped Out) and the regulator only acting when the damage was already done (South Korea to ban new listings of single-stock leveraged ETFs), is a preview of what’s going to happen in the rest of the world, especially in the US stock market, where any single metric, from over-concentration to the Buffet Indicator to margin investing and options speculation, has been screaming danger for a long time now. But the music of the party was too loud and covered any warning siren and warning voice in the background.
Now investors and lenders are asking all these companies a few simple, but legit, questions:
- Why aren’t you making money after all this capex spending?
- Why should I finance even more capex spending when you are failing to deliver on your promises?
- All these companies that are booking your services and promising tens of billions, if not hundreds of billions, of dollars in future revenues: do they have, or will they realistically have, the money to pay?
Last week we saw what happens to companies that fail to properly address these concerns: Alphabet tumbled 7% after reporting Q2 earnings, and Intel (whose shares initially jumped 8% right after Q2 earnings thanks to dumb hedge fund algorithms that act only on the always carefully crafted numbers presented by companies to highlight achievements and hide risks) closed the trading day down 8% on Friday after management disclosed they might need to raise new debt to finance growth. This week Amazon, META, and SK Hynix will report Q2 earnings, and I assure you the market reaction will be ugly if they won’t be able to convincingly answer those three very simple and legit questions.
The AI boom didn’t run on “inevitable” economics; it ran on a narrative strong enough to sustain unlimited capex, leverage, and accounting optics while real cash returns lagged. If a meaningful share of the GPUs booked as “sold” are still idle in warehouses, then today’s revenue and margins are not a clean signal of end-demand. They’re a signal of how long the funding stack can be kept stable.
That is why the market’s questions are turning from hype to solvency: who is actually generating profits from this spend, who can keep financing the next wave, and which promised future contracts are backed by customers that can truly pay? When those answers disappoint, the unwind won’t look like a gentle rotation. It will look like a party ending all at once, as depreciation, write-downs, and tighter credit collide with forecasts that were always more story than substance. The headache of this hangover is surely going to be massive.
Dear readers and followers, after many requests over the years for more direct access to me and premium content, I launched a Premium Subscription on Slice Finance. Slice Finance not only lets me share premium content, commentary, and trade ideas for a market-professional audience, but also provides many tools you can use to directly interact with me easily: live chats, group chats, and weekly live streams. I’ve set the subscription fee at a fair, accessible price of 39$ a month with no fixed time commitment. If you’re interested, you can subscribe to my Slice portal by clicking this link: https://bit.ly/JustDario