AI demand remains strong, while rising capital commitments and China’s low-cost competition keep cycle risk elevated.

The underlying risk picture is broadly unchanged. Recent results continue to show strong cloud demand and AI-related monetisation at Amazon, Alphabet and Meta, while their infrastructure spending continues to depress free cash flow or raise capital intensity. Financing is broadening through large equity and debt issuance, and China’s lower-cost model ecosystem and domestic-chip gains remain a competitive constraint, although no new China-driven development materially changed the cycle today.

• Capital intensity remains high: Amazon, Alphabet and Meta reported substantial AI infrastructure outlays alongside negative quarterly or trailing free cash flow.
• Financing reliance is widening: Alphabet, Amazon, Meta, Microsoft and Oracle have raised $326 billion through stock and bonds this year, increasing sensitivity to funding conditions.
• Anthropic disclosed roughly $895 billion of compute- and infrastructure-related commitments, illustrating the scale of off-balance-sheet and contractual exposure relative to currently recognised revenue.
• Grid, power and capacity constraints remain a conversion risk: Alphabet said it is using third-party capacity and expects near-term cloud-margin headwinds from supply constraints and higher operating costs.

• Cloud conversion remains tangible: AWS revenue rose 39% year over year and AWS operating income rose to $17.8 billion in the latest quarter.
• Google Cloud revenue rose 103% and segment operating income more than tripled; Alphabet expects to recognise more than half of its $561 billion cloud backlog over the next 24 months.
• Meta reported AI-driven advertising gains through engagement, recommendations and monetisation, while quarterly revenue grew 26% year over year.
• China’s cheaper capable models can expand adoption and inference volumes, potentially increasing global demand for cloud capacity, networking and accelerators even as model-layer pricing weakens.

• Whether AI infrastructure debt continues to migrate toward lower-quality or less protected borrowers and structures.
• Whether power and permitting delays alter tenant payments, project milestones or lender protections at Project Jupiter or comparable campuses.
• Whether cloud backlogs convert into recognised revenue and operating cash flow without lower prices or more concessional terms.
• Whether Chinese model distribution translates into durable paid overseas enterprise adoption, and whether domestic chips gain further share without constraining compute growth.