Moody’s Ratings told investors this week that the buildout of artificial intelligence infrastructure is draining free cash flow and raising balance-sheet risk at six of the largest technology companies in the world. In a research note issued Wednesday, the agency wrote that the shift from asset-light to asset-heavy business models “requires unprecedented levels of investment,” and said the transition threatens credit quality at Microsoft, Amazon, Alphabet, Meta, Oracle, and CoreWeave.
Moody’s projects combined capital expenditures of $785 billion across the six firms in 2026, rising to roughly $1 trillion the following year, against direct debt that has already reached approximately $460 billion.
The number that deserves the most attention, though, is the one that does not appear on any balance sheet.
The leases nobody is counting
Rather than owning every new facility outright, hyperscalers have leaned on off-balance-sheet financing through long-term data center leases. Moody’s puts those lease commitments across the group at $1.2 trillion, with more than $820 billion tied to leases that have not started because the buildings are still under construction. The agency treats those obligations as debt-equivalent liabilities — commitments that will bind these companies to substantial rent payments regardless of what happens to AI demand.
Moody’s analysts pegged the unstarted-lease figure at $662 billion back in February. It has climbed roughly 24 percent in five months, and the off-balance-sheet total now runs to nearly twice the group’s combined direct borrowing.
That is the structural point. A company can slow capital spending in a downturn. It cannot walk away from a signed twenty-year lease on a facility that is halfway built.
Where the pressure actually sits
The top of the group is not in trouble. Moody’s noted that Microsoft, Alphabet, Amazon, and Meta still hold among the strongest corporate balance sheets in the world, and their investment-grade ratings are not at immediate risk.
The strain is concentrated below them. Oracle carries a Baa2 rating with a negative outlook, two notches above junk, while CoreWeave sits in high-yield territory at Ba3 and finances its Nvidia GPU fleets through complex private debt structures. Those are the two names where a demand shortfall translates into a financing problem rather than a headline.
The circularity problem
Moody’s also flagged what it calls a circular AI ecosystem: a portion of the multibillion-dollar contract backlogs hyperscalers cite as evidence of demand comes from strategic agreements with pre-IPO AI labs including OpenAI and Anthropic — labs that have received large investments from the same tech giants now selling them cloud capacity.
That structure inflates reported demand without necessarily producing independent revenue. The Bank for International Settlements has characterized some of the financing behind the buildout as resembling shadow borrowing. Whether the backlog is real customer demand or recycled capital is the question that determines whether $1.2 trillion in future rent gets covered.
Moody’s has separately noted a two-to-three year lag between capital spending and AI revenue actually arriving.
What the market did with it
The test came almost immediately. Alphabet reported its first negative free cash flow quarter since going public, despite Google Cloud revenue growing 82 percent, and the stock fell 7 percent as investors pressed for evidence that the capital being deployed will be monetized. Quarterly capital expenditures hit $44.9 billion, and the company raised its 2026 capex guidance by $15 billion at the midpoint to a range of $195 billion to $205 billion, signaling a significant further increase in 2027.
Strong cloud growth was not enough. That is a genuine shift in how this spending is being received.
Why it matters outside Silicon Valley
Three implications for businesses that are not building data centers.
Enterprise AI pricing is not permanently subsidized. Companies that have built workflows around current per-token and per-seat pricing should understand that those rates reflect a land-grab phase funded partly by debt. Providers facing pressure to demonstrate returns on $785 billion have an obvious lever.
The regional development pipeline carries real risk. Data center projects promised to municipalities across the Northeast and mid-Atlantic depend on developers and tenants who are financing speculatively. Local officials negotiating tax abatements and grid interconnection commitments are counting on rent that Moody’s is now describing as an obligation rather than a certainty.
Credit markets are repricing tech, and that flows downstream. When bondholders demand wider spreads on the largest technology borrowers, the cost of capital rises for every vendor and integrator in that supply chain.
None of this argues the buildout is a mistake. Moody’s itself points to robust computing demand and long-term contracts that give real revenue visibility. The argument is narrower and harder to dismiss: the economics of the most profitable business model of the last two decades are being rebuilt in public, on borrowed money, and the bill arrives on a schedule nobody controls.
JBizNews Desk | New York
© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.
