A Nikkei Asia investigation has found that five US tech giants — Alphabet, Microsoft, Amazon, Meta, and Oracle — are sitting on an estimated $1.65 trillion in off-balance-sheet debt from AI infrastructure spending. That figure exceeds the $1.35 trillion in debt the same companies officially reported in their most recent quarterly filings. The accounting technique being used is the same one that helped bring down Enron.
🔍 THE BOTTOM LINE
The AI infrastructure boom is being financed not just with reported debt and equity, but with special purpose vehicles and off-balance-sheet arrangements that keep the full liability off the companies’ financial statements. Meta alone has roughly $420 billion in hidden debt, according to Nikkei. The total across the five companies — $1.65 trillion — is larger than the GDP of most countries. Accounting experts are explicitly comparing the pattern to Enron’s pre-collapse structure, and the comparison is not flattering.
What the Investigation Found
Futurism reported on the Nikkei Asia investigation, which examined the financial filings of the five companies most aggressively building AI infrastructure. The key finding: these companies are using special purpose vehicles (SPVs) and legally distinct subsidiaries to fund data centre construction without recording the full debt on their balance sheets.
The technique is legal. It is also the same technique that Enron used before its 2001 collapse — creating shell entities to hide liabilities and make financial reporting look healthier than reality. Enron’s off-balance-sheet structures concealed billions in debt until the entire edifice collapsed. The scale here is orders of magnitude larger.
Meta’s $420 billion in off-balance-sheet debt is particularly striking. The company has been the most aggressive AI infrastructure spender among the five, committing to data centre builds that dwarf its annual revenue. The hidden debt suggests Meta’s actual financial exposure is significantly greater than its public filings indicate.
The Enron Comparison
Technical accounting consultant Tom Selling told Bloomberg Tax that the accounting treatment is “in fashion” but raised the core risk: “What if one of these companies was a house of cards and was propping itself up with this accounting treatment? To me, that’s the risk.”
The Enron comparison matters because it identifies the specific failure mode. Enron didn’t collapse because its business model was wrong — it collapsed because the hidden debt became unsustainable and confidence evaporated overnight. When investors realised the company’s actual liabilities were far larger than reported, the stock went to zero in weeks. The same dynamic could apply to any of these five companies if AI revenue fails to materialise fast enough to service the real debt load.
Yahoo Finance has already published a direct comparison piece framing the AI sector as “the new Enron sector.” The framing is provocative but the underlying math is sobering.
How This Connects to What We Already Know
This investigation contextualises several stories we have already covered. When S&P downgraded Oracle to BBB- — one notch above junk — the rating agency cited a $42 billion free cash flow deficit and called OpenAI a “central credit risk.” That was based on Oracle’s reported debt. The off-balance-sheet figure is presumably larger.
When Google burned through more cash than it earned for the first time in a decade — negative $5.9 billion in free cash flow — CEO Sundar Pichai said the company would “continue to invest” as long as demand outpaces supply. Alphabet plans to spend $205 billion on AI this year. The Nikkei investigation suggests that figure may understate the actual capital commitment by a significant margin.
Meanwhile, Moonshot AI’s Kimi K3 is bringing price competition to the model market, potentially compressing the revenue side of the equation while the spending side accelerates. If model prices fall faster than infrastructure costs, the gap between AI spending and AI revenue widens — and the hidden debt becomes harder to service.
Why the Off-Balance-Sheet Trick Works (For Now)
Special purpose vehicles allow companies to finance large capital projects without consolidating the debt on their balance sheets. The SPV borrows money, builds the data centre, and the tech company pays usage fees. Under current accounting rules — particularly ASC 842 for leases and the various consolidation exceptions — much of this debt stays off the parent company’s books.
The technique works as long as two conditions hold: the SPVs generate enough revenue to service their own debt, and investors don’t demand full transparency. Both conditions are under pressure. Four of the five companies analysed are reporting Q2 earnings in the coming weeks, and analysts are increasingly asking about AI capital expenditure returns. If revenue disappoints, the off-balance-sheet structures become the next logical question.
What Happens If the Bubble Pops
The vulnerability is not that any single company goes bankrupt. These are the largest companies in the world, with diversified revenue streams. The risk is contagion: if one company’s hidden debt becomes unsustainable and forces a fire sale of AI infrastructure, the market re-prices the assets of all five. Data centres that were valued based on projected AI demand suddenly trade at a fraction of their build cost. The SPVs default. The parent companies face consolidation pressure they have been deferring.
This is the Enron scenario. Not a single collapse, but a systemic re-pricing event where investors discover that the reported financials were a best-case presentation of a much worse reality.
❓ FAQ
What is off-balance-sheet debt?
Debt held by entities that a company controls but does not fully consolidate in its financial statements. Special purpose vehicles (SPVs) are the most common structure. The SPV borrows money and builds an asset — say, a data centre — while the parent company pays usage fees. The debt appears on the SPV’s books, not the parent’s.
Is off-balance-sheet debt illegal?
No. Under current US accounting rules (ASC 842 and related standards), companies can legally structure financing to keep certain liabilities off their balance sheets. Enron’s use of SPVs was also technically legal until the rules were tightened after its collapse. The question is whether the current rules are adequate for the scale of AI infrastructure spending.
Why should a New Zealand reader care?
Two reasons. First, if the AI bubble pops and these companies’ spending contracts, the model market that NZ businesses rely on — Claude, GPT, Gemini — could see sharp price increases as providers try to recoup infrastructure losses. Second, NZ’s own sovereign AI ambitions depend on the economics of the global AI market. If the infrastructure bubble deflates, the cost of compute could fall — or spike, depending on how the default cascade plays out.
What would trigger a crisis?
The most likely trigger is a revenue disappointment. If Q2 and Q3 earnings show that AI revenue is not scaling proportionally with infrastructure spending, investors will re-examine the balance sheets. Once analysts start asking about SPV-level debt, the off-balance-sheet structures become a liability rather than a convenience.
🔍 THE BOTTOM LINE
The Nikkei investigation is not predicting a collapse. It is documenting that the financial foundations of the AI boom are significantly less solid than the companies’ reported numbers suggest. $1.65 trillion in hidden debt is not a rounding error — it is a structural feature of how the AI infrastructure buildout is being financed. The Enron comparisons will sound alarmist until they don’t. The same accounting technique, the same scale of hidden liability, the same confidence-dependent structure. The only missing ingredient is the trigger — and with Q2 earnings arriving in weeks, that may not be missing for long.
