Skip to content
NewsResearch

Nikkei study: Big Tech AI infrastructure debt grew 8x to $1.65 trillion, now exceeds on-balance-sheet borrowings

· by Pondero Newsdesk

The short version

A Nikkei Asia study published July 21 found that Alphabet, Microsoft, Amazon, Meta, and Oracle carry $1.65 trillion in off-balance-sheet AI infrastructure obligations, up 8x since 2022 and above the $1.35 trillion they disclose as formal debt.

Nikkei study: Big Tech AI infrastructure debt grew 8x to $1.65 trillion, now exceeds on-balance-sheet borrowings

A Nikkei Asia study published July 21 put a concrete number on the accounting gap inside Big Tech's AI buildout: $1.65 trillion in off-balance-sheet obligations at five US technology companies, a total that now exceeds the $1.35 trillion they formally disclose as debt.

What

The five companies covered are Alphabet, Microsoft, Amazon, Meta, and Oracle. Their hidden obligations come from two sources: data center leases and long-term GPU supply contracts. Both require payment regardless of demand but do not appear in standard debt disclosures. Per Nikkei's reporting, the off-balance-sheet total grew roughly eightfold since 2022 as AI investment accelerated.

Meta holds the single largest exposure. Its off-balance-sheet liabilities reached approximately $420 billion, nearly triple its reported debt, per the study. The bulk traces to long-term data center commitments tied to Meta's infrastructure expansion. Individual breakdowns for the other four companies were not available in the retrieved reporting. Treat the $1.65 trillion aggregate as an analyst estimate rather than a formal GAAP disclosure.

Three institutional voices had flagged this category of exposure before the Nikkei study assigned it a named total. Per AIWeekly's coverage of the findings, the Bank for International Settlements warned about shadow-borrowing structures in tech as a market stability risk, Morgan Stanley flagged the lease-risk exposure, and Moody's raised concerns about pre-operation lease obligations.

Why it matters

Conventional balance-sheet analysis of these five companies captured roughly half their total commitments. Adding the off-balance-sheet layer brings the combined obligation stack to approximately $3 trillion. The gap has a direct operational dimension beyond equity valuation. The GPU supply contracts and data center leases that produced the hidden debt are the same commitments underpinning cloud compute availability. A hyperscaler that needs to unwind those contracts carries costs not visible in most public filings today, and that unwinding runs through every inference API and hosted AI service built on top of that infrastructure.

What to watch next

Alphabet, Microsoft, and Meta report Q3 2026 earnings in October. Those calls are the first structured opportunity for analysts to press CFOs on off-balance-sheet infrastructure figures by name. A formal rating action from Moody's or S&P citing AI-infrastructure lease obligations would mark a higher-stakes escalation: credit ratings move equity, and a named downgrade risk tied to GPU contracts has not materialized yet.

Sources