Tech Giants' Off-Balance-Sheet AI Commitments Reach $3 Trillion as Morgan Stanley Warns of Hard-to-Assess Liability Risks
Nashnova编辑部
Nine major tech firms have amassed $3 trillion in off-balance-sheet AI commitments, dwarfing their reported capex; Morgan Stanley warns the hidden liabilities are making true risk harder for investors to gauge.
Where is the $3 trillion hiding?
A Wall Street Journal analysis of earnings footnotes across nine companies — Alphabet, Meta, Amazon, Microsoft, Oracle, Nvidia, Broadcom, SpaceX, and AMD — puts their off-balance-sheet commitments (contractual payment obligations that don't appear on the balance sheet) at $3 trillion.
The same group spent roughly $600 billion in capex over the past year; off-balance-sheet obligations are growing far faster.
This means → the visible "spending rate" is only the tip of the iceberg — the locked-in total is roughly five times larger.
What is the money committed to?
$1.2 trillion comes from leases on unfinished data centers — about four times the level recorded in 2025.
$1.9 trillion comes from long-term procurement contracts locking in Nvidia chips, memory, and grid power. Alphabet alone has $81.1 billion in purchase commitments; its energy contracts run through 2054.
In plain terms = half is "pre-renting the buildings," the other half is "pre-buying the parts and the electricity" — all on decades-long terms.
How extreme is Meta's "Hyperion" project?
Meta's Hyperion data-center campus in Louisiana covers an area equal to 1,700 American football fields and carries a 20-year lease.
The initial lease commitment alone is $12.3 billion; unrecognized lease obligations have climbed to $347 billion.
This reflects a single project whose hidden liabilities rival the entire market cap of many listed companies.
Is Nvidia itself expanding off-balance-sheet?
Nvidia has pledged $27 billion in equity investments before its current fiscal year ends, widening its own off-balance-sheet exposure.
This means → it is not just the chip buyers adding leverage — the chip seller is also betting on AI through off-balance-sheet vehicles.
What happens if AI revenue disappoints?
Alphabet and Amazon have already seen free cash flow turn negative as capex accelerates — meaning cash earned no longer covers cash spent.
Morgan Stanley warns that the increasingly complex structure of these obligations will make it harder for investors to assess big tech's true liability risk.
In plain terms = the books look healthy on the surface, but the long-term contracts are like invisible mountains — if AI monetization lags expectations, those mountains land squarely on the income statement.
Content is for reference only, not financial advice.