$1.65 Trillion Off-Balance-Sheet Bomb! The "Invisible Debt" of the Five Tech Giants Soared Eightfold in Four Years—Could the AI Arms Race Trigger the Next Liquidity Crisis?

According to Zhitong Finance, the frenzied "arms race" among America's five major tech giants over artificial intelligence (AI) infrastructure is giving rise to a "hidden debt empire" whose scale far exceeds their disclosed on-balance sheet liabilities.

Recently, an analysis of the latest financial statement footnotes from Alphabet (GOOGL.US), Microsoft (MSFT.US), Amazon (AMZN.US), Meta (META.US), and Oracle (ORCL.US) revealed that the combined "invisible debt" accumulated by these five companies via long-term procurement commitments and data center leases has soared to $1.65 trillion. This not only surpasses their officially reported liabilities of about $1.35 trillion, but has also surged eightfold over the past four years.

The massive scale of these off-balance sheet obligations is increasingly worrying Wall Street and global regulators: if AI demand falls short of expectations, these future payment commitments, tucked away in the footnotes, could rapidly turn into "debt bombs" consuming cash flows.

Deferred GPU Purchases Conceal Hidden Debt Risks

The so-called "invisible debt" refers to future payment obligations incurred by enterprises through investments, equipment procurement, or long-term leasing agreements. In accordance with current accounting standards, these obligations need not be immediately recognized as liabilities until the actual delivery of assets or completion of facilities. They neither immediately consume cash on hand nor appear in the main columns of the balance sheet, but are instead tucked away as small-print footnotes at the end of financial statements.

Currently, as the battle for global AI dominance intensifies, tech giants are spending massive sums to bulk-purchase high-end GPUs and build ultra-large-scale data centers. To avoid a sharp swell in immediate liabilities, they typically employ long-term procurement commitments, sale-leaseback, and long-term leasing as data center financing models. This approach legally and compliantly leverages the accounting rule of "capitalization upon delivery," deferring the hefty capital expenditure burden, while also building up a huge mountain of off-balance sheet invisible debt.

Statistics show that among the five companies, Meta’s hidden debt is the most staggering, estimated at about $420 billion—almost 2.8 times its on-balance sheet liabilities. Meta is co-building a massive data center with Blue Owl Capital (OWL.US) in Louisiana, with the total project development cost having risen from an initial $27 billion to over $50 billion. Meta has only put in 20% equity and will utilize the facilities via leasing after completion, thus reducing its upfront investment on the surface. However, it is reported that Meta has agreed to bear all losses as guarantor if the partnership falls apart—this potential rigid obligation is precisely what does not show up on the balance sheet.

Oracle’s hidden debt has seen explosive growth, increasing more than 30-fold over four years to $273.3 billion. This burden is mainly from large-scale lease arrangements, with a significant portion of the funds earmarked for the "Stargate" AI data center project in partnership with OpenAI. Meanwhile, the backlog of orders from Microsoft, Alphabet, and Amazon’s cloud services and other businesses, although totaling about $1.45 trillion and seemingly underpinning future revenues, nevertheless represents a growing overhang of deferred payment commitments.

If Demand underperforms, "Invisible Debt Bombs" Will Morph into Liquidity Disasters

The core industry concern is that, should global AI demand fail to keep up with the sector’s sky-high collective expectations, these deferred-recognition debts will suddenly become real cash flow black holes.

Some tech industry insiders have pointed out that while commercial real estate leases can retain considerable asset value, AI data centers are full of rapidly iterating semiconductors, and technological advances quickly depreciate hardware. If tepid AI demand causes utilization rates to fall, the tech giants will face massive asset impairment losses.

Even more concerning is the currently popular "circular investment" model—tech companies investing in each other’s AI services to support reported revenues and demand figures. This, in the short term, obscures uncertainties at the consumption end. However, if real demand growth cannot absorb the vast computing power under construction, deferred payments will explode all at once as these data centers come online.

Despite tech giants themselves maintaining optimistic outlooks—with executives such as the CEO of Amazon AWS insisting current AI infrastructure investments are not speculative—the explosive growth in off-balance sheet liabilities is an undeniable reality.

The “Central Bank of Central Banks” Issues Historic Warning

As tech giants pile up both on- and off-balance sheet leverage, the Bank for International Settlements (BIS), often dubbed the "central bank of central banks," issued an explicit warning in its latest annual economic report. The BIS directly compared today’s trillion-dollar AI investment frenzy to the 19th-century canal mania, the British railway bubble, and the dot-com bubble of 2000, stating that “these precedents all ended with investment reversals and ensuing economic recessions.”

The report estimates that the five U.S. hyperscalers’ AI-related capital expenditures for 2025-2026 will exceed $1 trillion, already surpassing what their profits and free cash flows can support, forcing some firms to take on debt for these investments.

The BIS specifically highlighted the "complex private transaction networks" lurking behind the scenes: tech giants securing long-term compute procurement through equity investments in AI labs, third-party contractors building data centers and leasing them back under long-term contracts embedded with exit clauses. Disclosure of these terms is highly insufficient, creating risks of multiple guarantees on the same assets.

If spending slams on the brakes, the entire supply chain—from chip companies to infrastructure builders, and private and direct investment funds—faces the prospect of simultaneous revenue streams drying up. Combined with the fact that direct lending to the AI sector has ballooned fourfold in five years, elevated inflation and ongoing geopolitical shocks could trigger chain reactions in the financial system that hit harder than traditional banking crises.

The BIS report clearly stated: "The size and speed of the current AI investment boom, coupled with widespread expectations of dramatic productivity gains, closely resemble the historical precedents above. And in all those cases, the end result was a sudden reversal of investment, which in turn triggered economy-wide recessions."