Goldman expects hyperscalers to issue $420 billion of debt in 2027, more than 60% above its 2026 estimate, while Bank of America estimates approximately $5.1 trillion of AI capital expenditure through 2030, including $3 trillion of growth spending and $2 trillion of maintenance expenditure. The AI boom is increasingly about who finances the buildout, what that capital costs, and how far credit markets can stretch. External funding needs could reach $1.2 trillion to $1.5 trillion, with the strongest operators largely able to fund themselves while Oracle, SpaceX, CoreWeave and Nebius face a combined shortfall of roughly $700 billion. Vendor guarantees from Nvidia and Broadcom are lowering borrowing costs through revenue support and residual value guarantees, but move part of the cycle’s risk onto the chip suppliers themselves.