Morgan Stanley projects that global AI-linked debt is on track to nearly double to $570 billion in 2026, financing the AI infrastructure build across Microsoft, Google, Amazon, Anthropic, and OpenAI through a mix of equity, private debt, and increasingly public bond markets. The note lands during a week when the four hyperscalers' combined 2026 capex commitments now exceed $650 billion. AI is now a credit cycle as much as an equity story.
The number that frames the move is the trajectory: AI-linked debt was roughly $290 billion in 2025 by Morgan Stanley's count, putting the 2026 ramp at 97% growth in a single year. That rate matches the AI capex ramp at the hyperscalers, which is the point. Equity issuance and operating cash flow alone cannot fund a half-trillion dollars of annual AI infrastructure, so the marginal dollar has shifted to debt. Microsoft and Amazon have issued investment-grade bonds at scale. CoreWeave and the GPU-cloud tier raised $7+ billion in private debt against forward GPU contracts. The bond desks now run an "AI infrastructure" category they did not run eighteen months ago.
The operator read is in the interest-rate sensitivity that this introduces into the AI thesis. If the 10-year sits at 4.5%, AI infrastructure financing math works at current returns. If yields move higher, the refinancing cost on $570 billion of debt creates margin pressure that no AI-revenue ramp can mask quickly. The Fable takedown, the Cursor acquisition, and the talent shakeups of the last two weeks were all corporate-strategic events. Macro just rejoined the AI picture as a real input.
Bottom Line
The AI build is now a credit cycle as much as a capex cycle. If you are modeling hyperscaler margins or AI-vendor valuations into 2027, the curve at the 10-year point matters more than it did six months ago.