July 22, 2026 — A bombshell study from Nikkei Research has revealed that America's five largest technology companies are sitting on a hidden debt pile of $1.65 trillion — debt that doesn't appear on their balance sheets but behaves exactly like it. The figure has grown roughly eightfold in just four years, driven entirely by the AI infrastructure arms race [citation:1][citation:4].

Even more alarming: this hidden debt now exceeds the companies' actual reported debt. If you thought Alphabet, Microsoft, Amazon, Meta, and Oracle were fortress-balance-sheet tech giants, you were only seeing half the picture.

The Numbers: $1.65 Trillion and Counting

According to the Nikkei study, the five companies' combined visible and hidden obligations total roughly $3 trillion — nearly double what appears in standard credit models [citation:1][citation:4].

To put that in perspective, $1.65 trillion is larger than the GDP of all but about a dozen countries on Earth. It's more than the entire market capitalization of most national stock exchanges. And it's all tied to one thing: building AI data centers, buying GPUs, and securing the computing power needed to train and run the next generation of artificial intelligence models [citation:1].

The growth trajectory is staggering. In 2022, this hidden debt stood in the low hundreds of billions. By 2025, it had already doubled. The 2026 figure is anticipated to be roughly twice as high as 2025 [citation:4].

How the Debt Is Hidden: SPVs, Leases & Shadow Borrowing

The key mechanism is off-balance-sheet financing — structures that allow tech giants to fund massive capital expenditures without the debt appearing as a straightforward liability on their corporate balance sheets [citation:6][citation:7].

The specific instruments identified by Nikkei are:

• Data center leases — long-term operating leases for massive AI facilities
• GPU supply contracts — multi-billion-dollar commitments to secure Nvidia and other accelerator chips
• Special Purpose Vehicles (SPVs) — separate legal entities that own the physical assets while the tech company leases them back
• Private credit joint ventures — partnerships with firms like Blue Owl Capital to fund infrastructure

Here's how it works in practice: Meta's $30 billion Hyperion data center in Louisiana was funded through an SPV called Beignet Investor. Meta owns just 20% of the joint venture. Blue Owl-managed funds own 80%. The project issued $27 billion in debt anchored by PIMCO and BlackRock. Meta keeps operational control but not the debt on its books. It simply leases back the finished campus under a long-term operating lease [citation:7].

The Bank for International Settlements (BIS) — the central bank for the world's central banks — has labeled this phenomenon "shadow borrowing" and warned that it creates "a complex web of private arrangements" with risks that are "typically poorly disclosed" [citation:3][citation:8].

Meta & Oracle: The Most Exposed Giants

Not all five companies carry the burden equally. The Nikkei study singles out two firms as particularly exposed:

Meta: $420 Billion Off-Balance-Sheet

Meta's off-balance-sheet liabilities come in around $420 billion — nearly triple its reported debt of approximately $151.6 billion (as of Q1 2026) [citation:1][citation:9]. Most investors would still describe Meta as effectively net-cash, given its $81.2 billion in cash and marketable securities. But in reality, the obligation stack sits roughly on the order of the infrastructure it has committed to build [citation:1].

Meta's 2026 capital expenditure guidance has already been raised to $125–$145 billion, up from a prior range of $115–$135 billion, reflecting higher component pricing and additional data center costs [citation:9].

Oracle: 30x Debt Growth in 4 Years

Oracle's hidden debt has soared roughly 30-fold over four years, reaching $273.3 billion by the end of May 2026 [citation:4]. This is particularly concerning because Oracle's ability to pay back this debt is largely dependent on the future profitability and success of OpenAI — its single largest customer [citation:4].

Oracle carries a Baa2 credit rating, just two rungs above junk bond territory. It has disclosed more than $248 billion of not-yet-commenced data center lease commitments and borrowings of about $124 billion. The company recently pledged to raise $45–$50 billion more this year, split between debt and equity, and has reportedly planned layoffs of thousands of employees to finance its data center buildout [citation:10].

BIS Sounds the Alarm: "This Has Happened Before"

The Bank for International Settlements didn't mince words in its Annual Economic Report 2026, released June 28. It drew direct parallels between the current AI investment boom and historical episodes that ended in economic catastrophe:

• The canal mania of the 1830s
• The British railway bubble of the 1840s
• The dot-com crash of 2000

"The scale and pace of the current AI investment boom, accompanied by expectations of large productivity payoffs, bear resemblance to these precedents," the BIS wrote. "These episodes ended with an eventual reversal in investment, inducing economy-wide recessions" [citation:8].

The BIS's core concern isn't that AI is a fraud — the technology is real, and task-level studies show productivity gains of 20% to 50%. The concern is that every major hyperscaler is making the same massive bet simultaneously, driven by the perception that only a handful of players will ultimately dominate. That logic, the BIS warns, is a recipe for collective overcommitment [citation:8].

Using contest-theory modeling, BIS economists found that as competitive pressure drives capital expenditure higher, the net economic surplus for the sector as a whole declines and could turn negative in adverse scenarios. A disappointment in returns, the report warns, "could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust" [citation:8].

Michael Burry Rings the Bell

On July 21, 2026, Michael Burry — the legendary investor who famously predicted the 2008 subprime mortgage crisis, immortalized in "The Big Short" — took to X (formerly Twitter) to amplify the Nikkei findings [citation:4].

"Hidden debt at U.S. tech giants swelled eightfold in roughly four years to an estimated $1.65 trillion as artificial intelligence investments ballooned, a Nikkei study shows, exceeding actual debt and making it tougher for investors to assess risk," Burry posted, quoting the Nikkei report directly [citation:4].

Burry's endorsement of the study carries weight. He has built a career on spotting systemic risks that the market misses. When the man who saw the 2008 crash coming starts waving red flags about tech debt, investors listen.

What Happens If AI Doesn't Pay Off?

The critical question is what happens if AI revenues fail to materialize at the scale needed to service this debt. Several scenarios are in play:

The BIS was careful to note that immediate risks to financial stability appear modest — but with a significant asterisk: the situation stays contained only if the AI sector actually delivers strong financial outcomes [citation:3].

As one analyst put it: "In 2008, the 'trash' was hidden in housing debt. In 2026, the 'trash' is unproductive AI infrastructure debt" [citation:11].

Key Takeaways