The bill nobody wanted to see
Big Tech’s AI spending spree is starting to look less like a flex and more like a tab that keeps getting handed back to the table. A Nikkei Asia study, as cited in the piece, says Meta, Oracle, Alphabet, Microsoft, and Amazon may collectively have about $1.65 trillion in obligations that don’t show up plainly on their balance sheets.
That’s the part that makes investors squint. Add in the roughly $1.35 trillion they already report, and the total picture starts creeping toward $3 trillion. In other words: the AI party is expensive, and a lot of the cost is hiding in the coat room.
Why the accounting nerds are suddenly the main characters
The liabilities are tied to things like data center leases and GPU contracts — the kind of commitments that can live in footnotes until the facilities are live or the hardware is delivered. That may be fine on paper, but it means the true size of the bet is easier to miss than a missed call from your landlord.
A few details are especially spicy:
- Meta’s hidden liabilities were pegged at roughly $420 billion, nearly triple its recorded debt.
- Oracle’s off-balance-sheet obligations have reportedly ballooned to $273.3 billion, as it builds around the Stargate project with OpenAI.
- S&P already cut Oracle to BBB- on July 9, just one notch above junk, citing AI infrastructure credit risk.
The real investor question: can the AI math actually work?
Ed Zitron’s big gripe is simple: the promised compute revenue would need to be enormous to justify the buildout. If the payoff doesn’t show up fast enough, the risk isn’t just slower growth — it’s financing stress, weaker free cash flow, and a lot more “we’ll explain later” in earnings calls.
And the worry doesn’t stop at the public market. The article says much of this financing runs through private credit and special purpose vehicles, which is where retirement and insurance money can get dragged into the story without signing up for the drama.
Big picture: AI infrastructure may still be a massive long-term opportunity, but this report is a reminder that even the coolest tech trend can turn into a balance-sheet hangover if the returns don’t arrive on time.