TLDR News Global
July 30, 2026
AI Companies' Hidden Debt Crisis: $1.65 Trillion Off-Book Liabilities
TL;DR
AI companies have hidden $1.65 trillion in off-book debt through long-term purchase agreements and lease arrangements, which represents 20% of all US corporate debt issued in 2026 and poses systemic risks to the financial sector.
Key Takeaways
- 1US tech companies borrowed $300 billion in the first 7 months of 2026, with JP Morgan projecting $500 billion total—20% of all US debt issuance, exceeding the 14% peak during the dot-com bubble.
- 2Five major AI companies (Alphabet, Microsoft, Amazon, Meta, Oracle) are hiding $1.65 trillion in off-book debt versus $1.35 trillion in disclosed debt through financial engineering.
- 3Nvidia has $119 billion in binding non-cancellable future purchase obligations with manufacturers like TSMC for chip production capacity.
- 4Oracle has $260 billion in future lease commitments that will eventually appear on its books, leading to its credit rating downgrade earlier this month.
- 5Meta increased bond yields from 7.1% nine months ago to 7.5% for its Texas data center financing, and Oracle's credit default swap costs jumped to 2.12%, meaning $212,000 annually to insure $10 million in bonds.
- 6Long-term purchase agreements and lease arrangements allow companies to commit to major future spending without reporting it as debt on balance sheets in the current period.
- 7Credit backstops promised by AI companies to intermediary firms create undisclosed liabilities that could trigger contagion risks if one major company defaults.
Notable Quotes
“These big AI companies are increasingly relying on slightly dodgy off-book techniques to finance their investments.”
“AI companies committing to buying some very expensive high-end chips in the future—you now owe someone money in the future, but you don't have to chalk it down as a loan on your balance sheet.”
“Oracle has an eyewatering $260 billion of future lease commitments that will eventually land on its books, which is a lot of hidden debt for a company with $67 billion in annual revenue.”
Chapters
1. The Scale of AI Debt Issuance
US tech companies issued $300 billion in debt in the first 7 months of 2026, with projections reaching $500 billion by year-end—20% of all US debt issuance, far exceeding the 14% peak during the dot-com bubble.
2. Hidden Off-Book Debt Discovered
Investigations by Nick and Bloomberg revealed that five major AI companies (Alphabet, Microsoft, Amazon, Meta, Oracle) plus Nvidia are hiding $1.65 trillion in off-book debt, more than double their disclosed $1.35 trillion in liabilities.
3. Long-Term Purchase Agreements
Companies like Nvidia commit to buying expensive chips in the future without reporting it as loans; Nvidia alone has $119 billion in binding non-cancellable purchase obligations with TSMC and other manufacturing partners.
4. Lease Agreement Strategy
Companies finance infrastructure by having intermediaries build data centers and entering long-term lease agreements, reporting only quarterly payments rather than total liabilities; Oracle has $260 billion in future lease commitments—a major factor in its recent credit downgrade.
5. Credit Backstops and Hidden Liability
AI companies provide credit backstops guaranteeing to cover intermediary companies if they run out of money, creating undisclosed liabilities that don't appear on balance sheets.
6. Rising Borrowing Costs and Credit Risk
Meta's borrowing costs rose from 7.1% nine months ago to 7.5% for new data center financing in Texas; Oracle's credit default swap costs jumped to 2.12%, costing $212,000 annually to insure $10 million in bonds.
7. Systemic Contagion Risks
While companies like Alphabet and Meta with diversified revenue are less exposed, companies like Oracle are highly leveraged and dependent on AI revenues; interconnected circular financing creates contagion risks where one company's failure could trigger defaults across the sector.
Key People & Entities
- Alphabet
- AI-adjacent tech giant with diversified revenue streams and lower leverage exposure
- Microsoft
- AI-adjacent tech giant with $1.65 trillion in hidden off-book debt among the five major companies
- Amazon
- AI-adjacent tech giant with hidden off-book debt among the five major companies
- Meta
- AI company financing Texas data center project with 7.5% bond yields; has proven diversified revenue streams
- Oracle
- AI company with $260 billion in future lease commitments and $67 billion annual revenue; recently had credit rating downgraded; credit default swap costs at 2.12%
- Nvidia
- AI chip manufacturer with $119 billion in binding purchase obligations to TSMC and manufacturing partners
- TSMC
- Semiconductor manufacturer providing production capacity for Nvidia's bespoke chip production
- JP Morgan
- Bank projecting AI companies will issue additional $200 billion in debt before end of 2026
Glossary
- Long-term purchase agreements
- Commitments to buy something in the future that create debt-like liabilities but don't have to be reported as loans on balance sheets.
- Lease agreements
- Rental arrangements where companies pay fixed amounts over time; US accounting law only requires reporting each payment as cash outflow rather than total committed spending.
- Credit backstops
- Promises by major companies to cover intermediaries' shortfalls if they run out of money or cannot borrow more, creating contingent liabilities.
- Credit default swaps
- Insurance products priced to reflect the cost of protecting bond holders against a company's default; rising costs indicate increased perceived risk.
- Off-book debt
- Liabilities that companies have committed to but don't have to report on their balance sheets under current accounting rules.
- Contagion risk
- The risk that one company's failure could trigger defaults and failures across interconnected financial networks.