Key Takeaways:
More than $200 billion in data-center debt was raised in 2025, creating a closed-loop financing system with echoes of past bubbles.
More than $200 billion in data-center debt was raised in 2025, with projections topping $1 trillion by 2028, as cross-investments among Nvidia, Microsoft and OpenAI created financing structures one analyst likened to a blend of 1990s infrastructure spending and 1920s-style lending.
"AI is real, but investors may be pricing the future far too early," Ray Dalio, founder of Bridgewater Associates, said.
Nvidia passed a $5 trillion valuation in October 2025 and is pouring $100 billion into OpenAI to help build data centers loaded with its chips. Microsoft owns 27 percent of OpenAI and represents nearly a fifth of Nvidia's revenue. OpenAI partners with CoreWeave, a company Nvidia also holds a large stake in. When CoreWeave issues billions in debt to build new capacity, Nvidia guarantees it will buy whatever CoreWeave cannot sell through 2032.
The risk is that the entire structure depends on constant capital inflows. OpenAI, valued at $500 billion, expects $13 billion in revenue and a $5 billion loss in 2025 and may burn more than $140 billion before turning profitable — more than Amazon, Tesla and Uber's cumulative early losses combined. If credit conditions tighten or a major player needs to sell instead of buy, the system could face a rolling reset similar to 2000-2002.
A $1 Trillion Debt Market Takes Shape
Roughly $200 billion in data-center debt was raised in 2025 alone. By 2028, the market could exceed $1 trillion, with as much as $750 billion coming from private credit, according to industry estimates. Meta secured $29 billion for a Louisiana facility, with $26 billion structured as debt from Blue Owl and PIMCO. Many of these deals involve off-balance-sheet financing, GPU-backed collateral and complicated leasing structures with borrowers who have never been tested in a downturn.
AXA said in December it would "avoid financing technological gambles" after watching lending volumes explode, a sign that even major institutions are starting to worry. Wisconsin regulators in July denied We Energies' petition to loosen financial guarantees for hyperscale data center users, suggesting oversight is tightening.
The Revenue Gap That Could Break the Loop
The most awkward reality: the companies building the foundation for AI are not profitable. OpenAI may burn more than $140 billion before turning profitable — more than Amazon, Tesla and Uber's cumulative early losses combined. An MIT study found 95 percent of companies see zero return on their generative-AI investments despite spending $30 billion to $40 billion. Bain estimates that AI will need $2 trillion in annual revenue by 2030 just to justify current infrastructure spending — more than the combined revenues of America's largest tech firms in 2024.
Goldman Sachs notes that $19 trillion in market cap is running ahead of economic impact, citing five danger signals reminiscent of the 1990s: peaking investment, falling profits, rising debt, Fed cuts and widening credit spreads. The S&P tech sector trades at 30 times forward earnings — high, but nowhere near dot-com extremes. Still, AI has driven 75 percent of S&P returns, 80 percent of earnings growth and 90 percent of capex growth since late 2022, making the market increasingly dependent on a handful of companies.
The most likely scenario is not a sudden collapse but a rolling reset where the weakest players fail first and drag parts of the system with them. Big Tech will survive. The casualties will be the unicorns, the developers relying on off-balance-sheet structures and the investors who believed the hype without checking the math. Data-center demand is projected to grow more than 19 percent annually through 2030, and Nvidia sees global capex rising from $600 billion to as much as $4 trillion. If demand keeps up, today's spending rush will not look reckless — it will look early. The question for 2026 is whether the financial structure built around it can hold.
This article is for informational purposes only and does not constitute investment advice.
