The $3 Trillion Shadow Ledger: Off-Balance-Sheet AI Commitments Dwarf Reported Debt
Underneath the reported capital expenditures of the major technology giants lies a massive, hidden financial ledger of off-balance-sheet commitments. As hyperscalers race to build out artificial intelligence infrastructure at a trillion-dollar pace, they are increasingly turning to off-balance-sheet financing to keep direct debt off their balance sheets and preserve their stellar credit ratings.
Ballooning Lease Commitments
According to a late July 2026 research note by Moody’s Ratings, off-balance-sheet lease commitments across the major hyperscalers (including Microsoft, Alphabet, Amazon, and Meta) have ballooned to $1.2 trillion.
Crucially, $820 billion of this total is tied to leases that have not yet started, indicating that these data centers are still under construction. While these obligations do not appear as traditional direct debt on balance sheets, ratings agencies treat them as debt-equivalent liabilities that legally bind the companies to massive, long-term rent payments.
As Moody's noted:
"To keep direct debt off their balance sheets, hyperscalers are leaning on off-balance-sheet financing, mostly through long-term data center leases1... While these obligations don’t show up as traditional debt, Moody’s says it considers them as debt-equivalent liabilities that will bind companies to significant rent payments down the line."
Crux AI: The Off-Balance-Sheet TPU Vehicle
A prime operational example of this strategy is Crux AI (formerly known internally as "Project Braid"), a $5 billion neocloud joint venture between Google and Blackstone launched in mid-2026.
Led by former Google VP of Engineering Benjamin Treynor Sloss, Crux AI is funded with $5 billion in equity from Blackstone to build out 500 megawatts of TPU-powered AI data center capacity by 2027. This structure allows Google to scale its custom Tensor Processing Unit (TPU) hardware footprint to rival Nvidia, while moving the direct capital expenditures and construction risks entirely off Alphabet's balance sheet.
The Circular AI Ecosystem
This shadow ledger is further complicated by structural circularity. Hyperscalers have invested billions of dollars in pre-IPO artificial intelligence labs such as OpenAI and Anthropic. These labs, in turn, contract to spend billions on cloud computing from these same tech giants. This creates a circular loop where reported cloud backlogs are funded by the hyperscalers' own investments, masking the true commercial demand and capital risks of the AI buildout.
-
An instance of The AI buildout's profits are paper gains and its true debts sit off balance sheet. — Moody's $1.2 trillion of debt-equivalent lease commitments — $820 billion not yet started — plus circular lab-to-cloud spending keep the buildout's real leverage and demand hidden from reported financials. ↩︎