Five of the world’s largest technology companies have collectively pushed an estimated $1.65 trillion in artificial intelligence infrastructure debt into off-balance-sheet vehicles, according to an analysis highlighted by TheNextWeb, raising comparisons to the accounting maneuvers that brought down Enron in 2001.
The report argues that as companies race to build the data centers, chips, and power capacity needed to train and run AI models, many are structuring the financing through joint ventures and special purpose vehicles (SPVs) rather than borrowing directly. That keeps the debt off the parent company’s balance sheet, even though the tech giant remains the primary customer and, in many cases, the entity ultimately responsible for the underlying assets.
How the structure works
Special purpose vehicles are legally separate entities created to isolate financial risk. In the AI buildout, a tech company typically partners with a private capital firm — asset managers like Blue Owl Capital, Blackstone, and PIMCO have all struck such deals — to jointly finance a data center. The private partner raises the debt and owns a stake in the facility, while the tech company signs a long-term lease or compute-purchase agreement to use it.
On paper, the arrangement lets the tech company report lighter capital expenditures and a cleaner balance sheet, even as it commits to years of payments that function much like debt service. Meta, for instance, has structured a joint venture with Blue Owl Capital to fund its Hyperion data center campus in Louisiana, a deal reported to be worth roughly $27 to $30 billion, with Blue Owl holding the majority equity stake and raising most of the capital through debt.
Echoes of Enron
Enron collapsed in 2001 after regulators and investors discovered the energy company had used thousands of off-balance-sheet special purpose entities to hide billions in debt and inflate reported profits, wiping out shareholder value and triggering the Sarbanes-Oxley reforms that tightened corporate disclosure rules. Analysts cited in the report say the AI financing structures are legal and disclosed in regulatory filings, unlike Enron’s fraudulent concealment, but they warn the effect is similar: outside observers can’t easily see the full scale of financial exposure tied to a company’s operations.
The core concern isn’t illegality — it’s opacity. Investors evaluating a company’s health may be looking at a balance sheet that understates the financial commitments propping up its AI ambitions.
Why it matters now
The AI infrastructure race has become one of the largest capital-spending waves in corporate history. Companies including Microsoft, Google, Amazon, Meta, and Oracle have collectively pledged hundreds of billions of dollars annually toward data centers, custom chips, and the power infrastructure needed to run them. Financing that buildout entirely with cash on hand or conventional corporate debt would weigh heavily on credit ratings and quarterly earnings, giving companies an incentive to lean on SPVs and joint ventures instead.
Oracle has taken a similar path in financing compute capacity tied to its cloud contracts with OpenAI, while CoreWeave and other AI infrastructure providers have relied heavily on debt-backed GPU financing arrangements. The pattern, according to the analysis, means the true liabilities tied to the AI boom are dispersed across a web of joint ventures, lease agreements, and private credit funds rather than concentrated on a handful of easily scrutinized balance sheets.
What could go wrong
Supporters of the structure argue it’s a sensible way to share risk with specialized infrastructure investors who are better equipped to own and finance long-lived physical assets like data centers. Private capital firms actively seek these deals because they offer stable, long-term returns backed by creditworthy tech tenants.
But critics warn the arrangement could leave markets under-informed about systemic risk if AI demand fails to meet the aggressive growth projections companies are using to justify the spending. If usage of AI compute slows or profit margins on AI products disappoint, the lease and purchase commitments backing this debt don’t disappear — they still have to be paid, even as the assets they’re tied to potentially lose value faster than expected. That dynamic is part of why some Wall Street analysts have begun flagging AI infrastructure spending as a potential source of financial fragility, even as the sector’s public earnings reports continue to look strong.
The debate over hidden leverage in tech’s AI spending arrives alongside broader scrutiny of how Silicon Valley’s biggest players are deploying capital, from data center financing to political spending; Google cofounder Sergey Brin has spent over $100 million fighting a California wealth tax, underscoring how much personal and corporate wealth is now tied up in the industry’s fortunes.
What comes next
Regulators have not signaled plans to change disclosure rules specifically targeting AI-related SPVs, and the companies involved maintain that their filings meet existing accounting standards. Still, the comparison to Enron is likely to keep pressure on tech giants to provide clearer breakdowns of their AI-related financial commitments, particularly as investors try to gauge whether current AI spending levels are sustainable or represent a bubble building outside public view.
