General Custer’s mistake, which he ultimately paid for with his own life, was to lead an attack against a force of Native Americans far superior in number to his own cavalry regiment, which ultimately surrounded and defeated him. Oracle is currently surrounded, not by Native Americans, of course, but by debt. An incredible amount of debt it has burdened itself with, aiming to become a major player in the AI space and to profit from the trillion-dollar revenues the AI space is projected to generate, even if these, after several years already, remain a mirage. AI will generate more revenues in the future for sure; however, the longer these fail to materialize to the extent many imagined, the more realistic investors’ expectations become. This is why investors are starting to worry about Oracle’s debt burden and whether the company will be able to repay it, at least.
Please take a look at the year-over-year free cash flow growth Oracle itself is reporting, numbers that should scare the hell out of all those people and financial institutions exposed to Oracle.
Hold on, shouldn’t this situation improve once Oracle starts to monetize the $523 billion of RPOs from 2026? Here is the problem: its major customer, OpenAI, which accounts for ~60% of the orders, doesn’t and will not realistically have the money to pay Oracle, which is expecting to receive a total of $60 billion in payments from it already in 2026. I explained why OpenAI is a disaster waiting to happen a few months ago in “OPENAI: LITTLE CASH, BUT PLENTY OF PROMISES, IN AN EFFORT TO AVOID BANKRUPTCY” and even if SoftBank keeps its promise to deliver $30 billion of fresh cash to OpenAI (“SOFTBANK’S KAMIKAZE BET ON OPENAI“), that will still not be enough to pay all the bills coming due in 2026.
Can Oracle’s other customers, like META and Microsoft, compensate for the shortfall caused by OpenAI? Realistically, they cannot, unless they shift their business model to corporate charity.
It is unknown how many billion dollars of GPUs Oracle committed to buy from Nvidia. The last available figure back in May claimed a $40 billion commitment, but surely that amount is much larger today because of OpenAI orders (“Oracle to buy $40 billion of Nvidia chips for OpenAI’s US data center“). Nvidia itself showcased over $500 billion of future orders, but the chances that cash will continue pouring into the coffers of Nvidia are becoming more of a factor of how willing debt investors will be to allow it. Considering where Oracle’s Credit Default Swaps are trading, debt investors’ willingness to keep supporting Oracle’s quest to become a major player in the AI space is decreasing at an alarming rate.
At this point, it is understandable to hear Oracle’s management modify its strategy, aiming for a more sustainable approach to serve AI until it starts generating the biblical amount of revenues many people are still dreaming about. This is what the management just said:
“Customers may bring their own chips to be installed in our data centers, and suppliers may lease their chips rather than sell them. Both of these options enable Oracle to synchronize our payments with our receipts and borrow substantially less than most people are modeling. We expect and are committed to maintaining our investment-grade debt rating.”
Hold on a second, according to what Oracle just said, doesn’t this imply that there are companies out there that bought Nvidia GPUs but do not have the capacity to install and operate them? Furthermore, is Oracle aiming to buy and lease Nvidia GPUs back to Nvidia rather than buy and pay for them in full on its own in a similar fashion to what some neocloud companies have already been doing? [“Nvidia signs $1.5bn deal to lease its GPUs back from Lambda“]
If the answer to both questions is “yes”, that doesn’t bode very well for Nvidia’s business, does it?
A sign that there might be plenty of Nvidia GPUs sitting idle in warehouses rather than being installed and operational is the number of data centers still under development.
I know what you are thinking right now: how will Nvidia realistically continue to sell millions of GPUs if its customers are so far behind in installing and profiting from the ones they already purchased? The good old circular financing trick is the answer, even if in the scheme, Nvidia is becoming more and more a lender of last resort, because equity investors are maxed out and debt investors are becoming cautious, rather than Nvidia remaining the biggest beneficiary of the cash inflows.
This is why it becomes critical for Nvidia to formalize the LOI signed with OpenAI (“OpenAI and NVIDIA Announce Strategic Partnership to Deploy 10 Gigawatts of NVIDIA Systems“), where the former will inject capital and lease GPUs to the latter, which will then likely turn around and deliver them to Oracle. This is exactly what I predicted in “THE LAST INNING OF NVIDIA’S GREAT REVENUE FABRICATION SCHEME JUST STARTED“:
- Nvidia commits $100bn of funding to OpenAI
- OpenAI will use that $100bn commitment from Nvidia as a guarantee to raise the $300bn it committed to pay to Oracle to purchase compute
- Oracle will use OpenAI’s commitment, now backed by Nvidia, to raise debt to buy Nvidia GPUs and scale up its infrastructure to serve OpenAI’s demand
- OpenAI will then pay Oracle with the money funded by Nvidia
- Oracle, at this point, will use that money to pay for the debt it funded to purchase Nvidia GPUs in the first place
To conclude, I believe that Oracle shares being down more than 10% post-earnings release in after-hours trading, after they lost all the gains from the previous quarter when investors rushed to buy its shares, mouth-watering over the $350 billion jump in RPOs announced at that time, is a statement on the growing skepticism about the sustainability and ultimately profitability of the whole data center buildout, especially in the US. Skepticism that I strongly believe will become panic when investors realize what I keep warning about: “THE DATA CENTERS FRENZY WILL BE REMEMBERED AS THE LARGEST WASTE OF CAPITAL IN HISTORY“
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