The $120 Billion Reality Check: AI Capex Hits the Wall of Worry
The honeymoon phase for artificial intelligence financing is officially over. As 2025 draws to a close, the bill for the massive infrastructure build-out, projected to exceed $400 billion this year alone, is coming due, and public credit markets are beginning to choke on the supply.
For the past year, the narrative has been dominated by equity valuations. But the real story is now in the plumbing of the financial system. Hyperscalers issued $121 billion in bonds in 2025, more than four times the five-year average. The market’s capacity to absorb this paper is being tested, driving a wedge between companies with pristine balance sheets and those relying on financial engineering to fund the future.
Public Markets Signal Indigestion
The clearest signal of market stress is evident in the divergence of credit spreads. While Microsoft and Alphabet retain their fortress balance sheets, other players are seeing the cost of capital rise sharply.
Oracle has become the canary in the coal mine. Despite record demand for cloud services, the company’s debt-to-equity ratio has ballooned to nearly 500%. In response, secondary markets are now pricing Oracle’s debt near junk levels, with credit default swap spreads widening to levels not seen since the 2009 financial crisis. This widening suggests that public bond investors are no longer willing to fund AI capex at any price; they are demanding a significant risk premium for leverage.
The Rise of “Shadow” Financing and SPVs
With public markets demanding higher premiums, tech giants are increasingly turning to private markets and complex structuring to offload risk. A new trend of “off-balance-sheet” financing has emerged, reminiscent of pre-2008 financial structures.
- Meta’s Strategic Pivot: Rather than taking debt onto its own books, Meta recently secured nearly $30 billion in financing through a partnership with Blue Owl Capital. This Special Purpose Vehicle (SPV) allows Meta to lease back data centers without damaging its credit rating.
- The Circular Economy: Companies like CoreWeave are raising billions in debt collateralized by the very Nvidia chips they are purchasing, chips that depreciate rapidly.
This shift to private credit and SPVs indicates that the “easy money” period of the AI boom has ended. Lenders are now the gatekeepers of the infrastructure build-out, and they are becoming increasingly selective.
Implied Volatility: What This Means for Borrowers
The volatility in the AI financing sector is a microcosm of the broader credit environment. When liquidity tightens for the largest tech companies in the world, the downstream effects are felt by mid-market borrowers across the economy. Banks, already wary of regulatory capital requirements, may retreat further as they assess systemic exposure to these intertwined AI credit risks.
The Bond Capital View: At Bond Capital, we distinguish between productive capex and financial engineering. The physical build-out of the AI economy (the steel, concrete, and energy infrastructure) requires massive, flexible capital. However, the current reliance on circular financing structures and asset-backed loans against depreciating hardware introduces systemic fragility.
For business owners and CFOs, the lesson is clear: Access to committed capital is paramount. In a market where public sentiment can shift spreads overnight, private credit offers certainty. While the public markets worry about Oracle’s leverage or Meta’s SPVs, we remain focused on lending against strong cash flows and tangible assets, ensuring our partners have the resources to grow regardless of Wall Street’s mood.
