A new report from Morgan Stanley highlights a critical shift in the AI arms race: the massive capital expenditures required are now outpacing the free cash flow of even the largest technology companies. This has forced hyperscalers like Meta, Alphabet, and Amazon to turn to the debt markets on an unprecedented scale, raising questions about a potential credit bubble. The investment bank now projects that the top five hyperscalers will collectively spend $800 billion on AI infrastructure in 2026, a figure that is nearly double the spend from 2025.
The core of the issue is that building the infrastructure for artificial intelligence is an incredibly capital-intensive endeavor. While these companies have long been seen as "asset-light" cash-generating machines, the pivot to AI requires building and equipping a global network of massive data centers. This spending spree is happening so fast that it is outstripping profits, forcing a reliance on borrowed money to fund the expansion and maintain shareholder returns through buybacks and dividends.
The result is a flood of new debt that is beginning to test the limits of the credit market. The tech sector now accounts for a record 18% of the US investment-grade bond supply, double its share from the same period last year. Specific deals underscore the scale: Meta recently secured a $13 billion financing package for a single data center in Texas, and Alphabet launched a multi-tranche euro bond offering worth at least €3 billion. However, signs of fatigue are showing; Meta’s recent $25 billion bond sale saw significantly lower peak orders than its prior offering, suggesting investor appetite may be waning.
This concentration of debt is creating a "wall of worry" for credit markets and the banks that underwrite these deals. According to a report from the Financial Times, major banks like JPMorgan are struggling to syndicate the massive loans, with one $38 billion package for an Oracle data center taking over six months to sell down. Banks are now hitting internal risk limits and are being forced to use "significant risk transfer" (SRT) tools to offload exposure to non-bank lenders and private credit funds. The market is signaling its concern in other ways; even as Meta’s stock has soared, the cost to insure its debt against default has climbed to a record high.
The bullish counter-argument is that this debt-fueled boom is a necessary investment in a technology that will unlock trillions in economic value, easily justifying the current spending. The broad adoption of AI-enabling technologies, as seen in the semiconductor space with companies like AMD, suggests a fundamental and sustainable demand driver. Proponents argue that the productivity gains from AI will generate more than enough revenue to service the debt, making the current risk assessments overly pessimistic.
Looking forward, Morgan Stanley has outlined four key signals that could indicate the AI credit cycle is turning sour: debt growth outpacing earnings growth, a sharp rise in M&A activity, faster growth in the leveraged-loan market, and a decline in the equity portion of private equity deals. For now, the credit markets continue to finance the AI buildout. The critical question is for how much longer, and at what price.
