Lately, I haven’t written that often like I used to about AI and all the wicked narratives that propelled the stock price of many companies in the space into the stratosphere. The reason is very simple: after more than fifty articles over 2 years (all archived here), until today, I haven’t found anything “new” I could bring to your attention, my dear reader. Most of what happened in the past few months has been the ultimate realization of so many warnings and predictions I brought up when few were listening, and I am not a big fan of just writing commentaries. Today, however, I finally have new discoveries I am excited to bring to your attention.
If Nvidia wants to keep the circular financing/revenue round-tripping going for a little longer, it will be forced to raise debt.
In the past weeks, all hyperscalers except Nvidia reported earnings for Q4-25, and one thing became obvious: if they want to persist in pursuing their nonsensical CAPEX plans, the cash that their legacy businesses generate isn’t enough anymore.
What’s happening here? Pretty simple: after more than 3 years since the launch of ChatGPT in November 2022, AI tools aren’t generating the biblical amount of revenues everyone dreamed of. Of course, excuses have been plenty. The most popular was the lack of enough computing power, even though one year ago, DeepSeek showed the whole world that people who knew what they were doing did not need a ton of computing power and hundreds of thousands of GPUs, as I described at that time in “THE REAL ERA OF AI BEGINS, THE ONE OF THE AI CHARLATANS ENDS“.
One year later, hyperscalers stick to their CAPEX plans of a magnitude never seen in history before, persisting in what is a more and more obvious incredible waste of capital resources, as I, of course, warned long ago in “THE DATA CENTERS FRENZY WILL BE REMEMBERED AS THE LARGEST WASTE OF CAPITAL IN HISTORY“:
- Google: $175B-$185B in 2026 vs $119B estimate
- Meta: $115B-$135B in 2026 vs $110B estimate
- Amazon: $200B in 2026 vs $146B estimate
- Microsoft: Run rate (based on its latest quarter) at $120B
However, as you can see in this chart, only Microsoft still generates enough free cash flow to cover all its CAPEX commitments.
Not surprisingly, while META already issued significant debt that, so far, it was allowed to keep off its books thanks to an accounting trick now being questioned even by its auditor (“Meta Auditor EY Raised Red Flag on Data-Center Accounting“), Amazon and Google quickly filed a shelf to issue debt securities shortly after their earnings, with the latter moving rather fast to tap the bond market to fill the gap between its projected free cash flow and the additional cash needed to maintain its CAPEX plans.
Nvidia is going to report earnings for Q4-25 today after close, and these are the investments the company announced and partially finalized so far in the past few months:
- OpenAI: $30 Billion
- Intel: $7.9 Billion
- Anthropic: $10 Billion
- Synopsys: $2 Billion
- xAI: $2 Billion
- Nokia: $1 Billion
- CoreWeave: $2 Billion
The total? Roughly $50 Billion in commitments.
Here is where the math starts shaking. In Q3-25, Nvidia reported roughly $22.5 billion in free cash flow and a total of roughly $60 billion in cash and short-term investments that can be quickly turned into cash. Assuming similar figures in Q4-25, these latest commitments will already deplete 60% of its cash and equivalents resources. Why will Nvidia then need to issue debt on its own? As you can see in this chart I posted for the first time months ago, and anticipated in my research years ago, the whole circular financing scheme can allow the size of the whole AI Bandwagon to continue growing at the current rates only if third-party equity and debt investors continue to provide cash into it, since revenues from real customers aren’t materializing in the magnitude expected.
Yes, most of the Nvidia commitments I listed before will be used by those companies to purchase GPUs from Nvidia, guaranteeing stickiness in short-term revenues. However, Nvidia will have to pay for the costs to make those GPUs, for the cost to continue repurchasing its shares, and overall operational costs, ultimately eroding its cash pile at a faster pace. As a consequence, at some point this scheme can carry on growing only if Nvidia itself starts raising debt, like its fellow hyperscalers have already started to do, even if Oracle showed the whole investment community how dangerous that strategy is if enough revenues do not quickly materialize. Oracle, after several months of misleading investors about its growth prospects, is now facing the first class action lawsuit for securities fraud after the sharp drop in value of its stock and bonds (“Oracle Corporation (ORCL) Investors: April 6, 2026, Deadline in Securities Fraud Class Action Lawsuit Filed by Kessler Topaz Meltzer & Check, LLP“). Mark my words: this will be the first of many class action lawsuits in the future.
What do you expect the impact on Nvidia stock and credibility will be after investors realize that the company isn’t an endless high-margin money printer like all investors have been brainwashed to believe since the whole AI charade started years ago?
CoreWeave results on Thursday will force investors to strongly question the whole AI narrative
Unlike hyperscalers, CoreWeave never had a legacy business to fund its ambitions with. On the contrary, as I described in many articles over the years, it was a perfectly financially engineered attempt to ride the AI wave as much as possible and as long as possible for the benefit of its original investor: Magnetar Capital. Hilariously, the same hedge fund that profited greatly before the 2008 GFC by manufacturing all sorts of radioactive CDOs first and then betting aggressively against them.
While Magnetar might not be shorting CoreWeave or other AI stocks right now, surely they have been trying to exit their trade as fast as possible since they have been allowed to after CoreWeave’s IPO, with barely a day going by without them and other insiders selling shares since then. Notice how, since CoreWeave’s IPO unlock back in August, there has been a total of ZERO insiders who bought shares in the company, even if the price is now trading almost 50% below its all-time high.
What is CoreWeave’s big problem? The company not only raised too much debt to service, but now lenders are unwilling to keep lending the company the money it needs to maintain its CAPEX plans.
The first red flag already occurred months ago when the company had to fall back on issuing Convertible Debt rather than Senior “Junk” debt: “CoreWeave Debt Deal Scooped Up By Investors. The Stock Is Dropping Again.” Not surprisingly, the stock dropped after the announcement because, unlike senior bonds, convertible bonds implicitly carry the risk of diluting the existing shareholders.
The second red flag then occurred when CoreWeave filed an 8-K on the 2nd of January 2026, hoping everyone was too distracted to notice. But unfortunately for them, I did.
On that day, the company disclosed that the terms of a data center loan facility with two of its lenders were renegotiated, not only significantly loosening the covenants the company had to meet in order not to trigger default, but even allowing the company to pay its lenders with equity in lieu of cash in case it did not have enough to repay that loan.
The third red flag occurred a few weeks later when CoreWeave announced it expanded its “partnership” with Nvidia, which, as per the new agreement, bought $2 billion more of CoreWeave shares in a private placement. We already knew that CoreWeave was struggling months ago to continue developing its data centers, and Nvidia—which is a client, supplier, and shareholder of the company that anchored its IPO, otherwise it would have busted—already stepped up its support to CoreWeave as I highlighted back in November in: “NVIDIA EARNINGS: MORE QUESTIONS THAN ANSWERS ON THE STATE OF THE AI BUBBLE.”
The fourth red flag just occurred a few days ago when a headline took many by surprise: “Blue Owl shopped debt for a CoreWeave data center. Lenders weren’t sold.” This headline made public what, for those who were paying attention (sadly not many), was clear for months: lenders were not willing to bankroll CoreWeave anymore.
Let’s put all this together now, shall we?
As you can see in the table below from the latest CoreWeave 10-Q, the company faced $4.5 billion in debt principal repayments by the end of 2026.
While the company still reported $1.8 billion in cash and $1 billion in unusual “restricted cash” in its latest 10-Q, that was arguably just a fraction of the cash the company needed to complete its 2025 CAPEX plans. After the management reduced 2025 CAPEX plans from $20-23 billion to $12-14 billion, as of Q3-25, only $6 billion were effectively spent on property, plants, and equipment. Here is the big question: has CoreWeave used the combined $4.25 billion it raised in Q4-25 to complete its 2025 CAPEX plans, putting itself at risk of immediate default in the next months unless a white knight bails them out, or is the company about to announce it will fall short of its CAPEX plans for 2025 and announce a significant reduction of its CAPEX ambitions for 2026, limited to the cash inflow it will get from its only real paying customers so far: Microsoft and Nvidia? Any answer to this question will inevitably impact the valuation of CoreWeave very badly.
At this point, back to the question in the title of this article, there is a very high chance that the answer will be “yes,” and whoever is ignoring the significant amount of red flags here, failing to prepare in advance, might risk paying a very high price for that in the next 48 hours.
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