AI's $1 Trillion Bet Runs on Borrowed Time

The top five U.S. hyperscalers will spend roughly $800 billion on capital expenditures in 2026, climbing toward $1.1 trillion in 2027, according to Morgan Stanley. That's triple what they spent in 2024. But here's the part equity investors keep missing: corporate profits can't cover it. The AI infrastructure buildout is increasingly a credit story, and credit markets are starting to price in doubt.

The top five U.S. hyperscalers are projected to spend approximately $600 billion in capital expenditures in 2026, representing a 38% increase over 2025's already stellar 68% growth, driven by continued AI infrastructure buildout , S&P Global Ratings reported in February. Hyperscaler gross bond issuance topped $100 billion in 2025 , per the Bank for International Settlements. Credit markets will finance more than $1 trillion in global data center spending through 2028 , Morgan Stanley estimates. The entire AI supercycle now depends on the smooth functioning of credit markets—at precisely the moment those markets are flashing warning signs.

Can Credit Markets Absorb This Much AI Debt?

Just a handful of hyperscalers and AI data center deals now account for an enormous share of duration-adjusted issuance, with Oracle becoming one of the largest risk-weighted names in the investment-grade universe and Meta rocketing up the rankings in less than a year , Goldman Sachs noted. The investment-grade market is starting to look less like a diversified bond portfolio and more like a leveraged bet on AI infrastructure.

The problem isn't just volume. Credit default swap spreads rose, especially for hyperscalers with lower credit ratings, reflecting both the volume of supply and uncertainties around the projects' payoffs , BIS researchers found. One Oracle-linked financing reportedly took more than 6 months to distribute because demand simply was not deep enough , according to the Financial Times. Some lenders explored selling portions at a discount just to free up balance sheet space.

Meanwhile, Big Tech is getting creative with its financing structures. Hyperscalers have turned to off-balance sheet arrangements involving dedicated vehicles that acquire or develop data center assets, with the hyperscaler holding a minority stake and committing to long-term operating leases or capacity offtake agreements while keeping most of the associated debt off the balance sheet . These arrangements amount to "shadow borrowing" that strengthens links between hyperscalers and non-bank investors such as private credit vehicles and insurers .

Translation: leverage doesn't disappear by moving it off the balance sheet. It just becomes harder to see.

What Happens When AI Disrupts Its Own Lenders?

The credit risks run deeper than hyperscaler debt. The massive sell-off in syndicated loans issued by technology companies was fueled by fears that rapid advancements in generative AI would upend traditional software business models , Bloomberg reported in March. In Technology, loan spreads widened by 40.26% compared with 23.84% for high-yield bonds .

Software companies—many of them leveraged buyouts financed by private credit—are suddenly facing an existential question: what's their business model worth when AI agents can automate their core functions? Roughly half of software loans carry ratings of B- or lower, with 26% rated CCC and only 7% in the comparatively safer BB tier , LPL Research found. Approximately 46% of software debt is due within the next four years, with 25% of the market needing to be refinanced in 2028 alone .

UBS analysts laid out a baseline scenario in which borrowers of leveraged loans and private credit see a combined $75 billion to $120 billion in fresh defaults by the end of this year , CNBC reported in February. The tail risk? Defaults could jump by twice those estimates, cutting off funding for many companies , UBS warned.

Private credit is where the structural risks are most deeply embedded—and least visible. Direct lenders funded 40% to 70% of leveraged buyouts between 2022 and 2023, and software and technology companies now represent over 20% of BDC investments, with market estimates that between 25% to 35% of private credit portfolios carry some degree of AI-related disruption risk , according to LPL Research.

The irony is almost too perfect: the AI boom is being financed partly by loans to companies that AI threatens to disrupt.

Who Bears the Risk If This Goes Wrong?

AI firms, traditionally reliant on internal cash flows and equity, now face higher leverage, which could amplify shocks and affect the health of financial intermediaries if expected returns on AI investments fail to materialize , the Bank for International Settlements warned. This raises concerns about the potential for systemic spillovers, not least given the rapid growth of less transparent private credit markets .

But the risks aren't confined to tech lenders and private credit funds. Utilities and their ratepayers are absorbing infrastructure costs too. At least 40 U.S. states now allow utilities to use Construction Work In Progress financing, letting companies recover some project costs from customers before infrastructure is completed—roughly double the number from a decade ago , Reuters reported. Critics believe households are absorbing risks that utilities and large technology companies previously carried themselves .

Dominion's 2024 resource plan projects nearly 27 GW of new generation by 2039, including 21 GW of renewable energy and 5.9 GW of gas , the Belfer Center noted. In February 2025, Dominion proposed its first base-rate increase since 1992, adding about $8.51 per month in 2026 . Consumers are increasingly blaming data-center power demand for higher electricity bills, with national attention on this issue expected to grow heading into the midterm elections in the U.S. , Morgan Stanley analysts said.

The question isn't just whether AI investments pay off for Big Tech. It's who gets stuck with the bill if they don't.

What Changed This Week

Enbridge announced it's developing a 365-MW solar and 200-MW/1,600-MWh battery storage project near Cheyenne, Wyoming, expanding its partnership with Meta to approximately 1.6 GW of contracted capacity across North America . Enbridge expects to invest $1.2 billion to construct the project, which is expected to enter service by the end of 2027 . The deal illustrates how hyperscalers are locking in dedicated power supplies through long-term contracts—and how much capital that requires. Meanwhile, credit markets continue digesting the implications of AI-driven debt issuance at a scale that would have seemed unthinkable two years ago.

What to Watch

S&P Global Ratings is monitoring whether primary bond markets can absorb tech and AI capital expenditure plans through 2026. UBS will update its default forecasts for leveraged loans and private credit as AI disruption accelerates. Watch for Q2 earnings calls from hyperscalers in late July—particularly their commentary on infrastructure financing and return expectations. The PJM capacity auction results, expected in June, will provide insight into how grid operators are pricing the data center demand surge. And keep an eye on private credit redemption pressure: if Business Development Companies face liquidity stress, the AI financing model could face its first real test.