S&P Global Ratings warns that the top six US hyperscalers are projected to spend over $7tn on AI and data centre capex by 2030, driving half of US private-sector growth. Despite rising capex and opaque financing, these firms do not quantify AI returns on investment, forcing S&P to assume investments will pay off. The agency sets guardrails: Microsoft’s debt-like commitments can only reach 1x EBITDA before AAA rating risk, while Oracle’s BBB- rating allows up to 4.5x EBITDA before junk status.

Amazon risks losing its AA rating by 2027 unless earnings rise or capex slows. Alphabet has the most headroom in its AA+ rating. Rating decisions are committee-based, with potential ameliorating circumstances, but leverage thresholds remain critical for credit outcomes.