In the past few days, something rather incredible happened: Amazon and Google, two “cash cows,” started to gear up to directly issue debt. When I read this headline, “Google to join exclusive 100 club with century-long bond,” I couldn’t hold back a smirk on my face while thinking, “This is it!”
What’s the matter with a century bond? While cash cows like Apple issuing debt can be justified when, cough cough, it is a good way to pay less taxes and continue buying back shares to inflate the stock price at the expense of undermining the resilience of the business in case of an economic downturn, issuing a century bond instead is a pretty strong signal that a “growth” company is turning into “value.” In a nutshell, the days of spectacular growth are over, as was always the case for every single company that issued century bonds before:
- Ford Motor Co.: Raised $2.5 billion through a 100-year “green bond” in 2021.
- Électricité de France (EDF): The French energy utility issued 100-year bonds in 2014 and 2021.
- Motorola: Sold 100-year debt in 1997, which was the last major tech issuance of this kind until 2026.
- IBM: Issued century bonds in 1996 during its period of tech dominance.
- Walt Disney Company: Issued a notable 100-year bond in 1993 maturing in 2093.
- Coca-Cola: Turned to century-long debt in the early 1990s.
- FedEx: Another major American corporation that issued 100-year debt during the 1990s.
- Norfolk Southern: Issued 100-year bonds in the 1990s.
- J.C. Penney: Issued a $500 million century bond in 1997 (however, the company later filed for bankruptcy in 2020).
Hold on a second, isn’t what I just said clashing completely with the whole AI bandwagon narrative that the whole sector will generate trillions of dollars in revenues in the future and unleash spectacular growth? Yes, it does.
Let me share a secret with you: when a fixed-income investor decides to purchase a bond or not, their goal is to earn an income and be repaid in full. Consequently, whoever is buying these bonds issued by tech companies isn’t betting on AI at all, as the misleading narrative wants you to believe, but is relying on the current business to remain profitable and able to service the debt obligations. If a fixed-income investor is interested in betting on the future business of a company, they buy convertible bonds, not senior bonds. Hilariously, Motorola was the last tech company that issued a century bond in the buildup of the DotCom bubble, and, not surprisingly, in 2011, it almost filed for bankruptcy. Why not surprisingly? Because asset-light companies like tech businesses aren’t fit to bear long-duration debt, since all it takes for them to go bust is a startup in a garage coming up with a new technology that disrupts their business model, exactly like Google did with AOL or Yahoo. Now, I am sure you would argue that Amazon, Google, or Oracle are issuing debt to build the AI infrastructure that will serve as collateral for it. Let me ask a question here: how reliable is the profitability of an infrastructure whose components face fast depreciation and a quick path to obsolescence? Easy answer: not very reliable since, unless that infrastructure is massively profitable in a short period of time, it won’t be able to repay itself and allow the company to smoothly service its debt obligations. This is the exact mistake investors made a few months ago when they jumped headfirst to buy Oracle bonds, only to then quickly regret it, as I explained in “AI IS BECOMING ORACLE’S LITTLE BIGHORN.”
The whole AI charade is now more than 3 years old, and revenues aren’t materializing in the spectacular fashion that every single tech leader preached for years. The excuse? There is not enough compute to scale. Why is that somehow a bottleneck that only applies to US companies is a question no one dares to ask when those are the only ones racing to build as many datacenters as possible, while all the rest of the world is watching, as I discussed in “THE DATA CENTERS FRENZY WILL BE REMEMBERED AS THE LARGEST WASTE OF CAPITAL IN HISTORY.”
Alrighty then, let’s keep building up datacenters, because that’s all you need to unlock the biggest revenue pot in the history of mankind. But since this will be a race where the winner takes it all (“THE AI END GAME: WINNER TAKES IT ALL AND EVERYONE ELSE LOSES MASSIVELY“), these hyperscalers need to be the first ones to strike gold. As a result, since their cash flow will not be enough anymore to keep pace, they have to fill the gap by issuing debt, exactly like Oracle did and Amazon and Google are doing now.
Wait a second, weren’t lenders already supporting the datacenters build-up, lending money to SPVs and allowing hyperscalers to keep most of the debt off the books? Yes, that’s exactly what has been happening till Blue Owl folded: “Oracle stock dips 5% as Blue Owl Capital pulls out of funding $10 billion data center.”
Why did Blue Owl, arguably the biggest supporter of the debt-funded datacenters build-up, fold? Because the datacenter’s business is not only not profitable, but as I mentioned earlier, the collateral provided against the debt loses value very quickly. The best example of this whole dynamic is CoreWeave. Since I stated that “COREWEAVE IS THE WEWORK OF AI” at the beginning of January, something important happened, albeit the market broadly ignored it: CoreWeave has been thrown a $2 billion lifeline by Nvidia (“Nvidia Invests $2 Billion More in CoreWeave, Offers New Chip“).
Why such a move? Here is CoreWeave’s loan repayment schedule. Let’s not forget loans that are guaranteed mostly by H100 GPUs, whose commercial value is now much inferior, and the company is hardly generating any cash to make repayments.
Furthermore, even if CoreWeave already cut its CapEx plans from $20 – 23 billion to $12-14 billion for 2026, there is no chance on earth it can keep that plan in place when lenders are not backing up the company anymore. The reason why CoreWeave had to fall back on its supplier, customer, and investor, Nvidia, for a cash infusion. The end of this story is already written: CoreWeave will either file for bankruptcy or be acquired for pennies on the dollar by a much larger company; it’s only a matter of when, not if.
Here is where the big nonsense lies: if a datacenter company’s business isn’t viable, how would that change if directly taken over by a larger company? Of course it won’t, which is the reason why Oracle is now in trouble. Would that change if the operations were directly and fully handled by the hyperscalers? Of course not, because the math doesn’t change, and what the math is screaming is that the revenues to be expected from AI will be a fraction of the projections presented so far.
I want to conclude by representing this chart from 5 months ago:
Because the whole AI bubble has been inflated thanks to the largest circular financing scheme ever built, the ponzi scheme, oops sorry, I meant to say bubble, will remain in place till all participants are able to suck in more cash than the one that is being incinerated. Mag7 now issuing debt outright will allow this scheme to continue a little longer, but once even the lenders willing to hand over cash to them are exhausted, then there won’t be any way to keep this whole house of cards standing.
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