Five tech giants are hiding $1.65tn in AI debt, using the trick that toppled Enron

The market is still valuing the AI buildout as if most of the burden is optional capex, when the more important issue is that these companies have effectively created long-dated fixed obligations that behave like debt in a downturn. That matters because the first derivative is not insolvency; it is higher perceived leverage, which raises the equity discount rate and makes each incremental dollar of AI revenue less valuable if monetization lags. The pressure point is not all five names equally: the market can tolerate this at Microsoft and Alphabet because of balance-sheet flexibility, but Oracle and, to a lesser extent, Amazon are more exposed to a regime where funding costs rise before the returns show up.

The second-order effect is on competitive behavior. Once one hyperscaler commits to a larger footprint, rivals are forced to match to avoid product and latency disadvantage, so the capex race becomes self-reinforcing and margins across the cloud stack can compress even if top-line growth remains healthy. That shifts bargaining power toward infrastructure suppliers and utilities in the near term, but it also increases the probability of a future demand reset if enterprise customers balk at pass-through pricing or if AI workloads do not monetize quickly enough. In that scenario, the losers are the most levered builders, not the largest cash generators.

The timing matters: over the next 1-3 months, the trade is about guidance, lease disclosures, and any incremental debt issuance or duration extension, not the absolute size of the buildout. Over 6-18 months, the real catalyst is whether AI revenue starts to justify the embedded obligations; if not, these names can de-rate on free-cash-flow quality rather than headline growth. The contrarian point is that the balance-sheet optics may be overblown for MSFT/GOOGL/META, but underappreciated for ORCL, where the market may still be pricing growth as if financing is nearly frictionless.