AI demand holds, while financing and power constraints persist and China’s low-cost models keep rent pressure elevated.

The underlying risk picture is broadly unchanged. Supermicro raised its preliminary quarterly revenue and margin expectations on stronger AI-platform demand, supporting near-term infrastructure demand; this is counterbalanced by persistent power-delivery and financing frictions around large campuses. China-driven competitive pressure remains material but unchanged: low token costs and rapid adoption of Chinese models may compress model economics while also broadening AI use and compute demand.

• Aggregate hyperscaler capex growth remains far above aggregate free-cash-flow growth, leaving returns dependent on sustained demand and efficient capacity deployment.
• Cloud-revenue growth still trails the capex growth used in the monetisation proxy, so evidence of demand does not yet establish proportionate cash returns.
• The Oracle/Blue Owl power delay shows that contracted projects can face delayed economics and tighter lender scrutiny even where tenant commitments remain intact.
• Reported power shortages could defer deployment of memory, optical and other secondary infrastructure even if leading accelerator demand remains resilient.

• Supermicro’s preliminary revenue and margin update points to stronger-than-expected AI-platform demand rather than an immediate order break.
• AWS management cited a nearly $500 billion order backlog, sub-three-year server paybacks and rapidly growing AI revenue, though these are company statements rather than independently audited project returns.
• Micron, AMD and HPE reported strong data-centre-related growth and profitability, consistent with broad infrastructure demand rather than solely speculative construction.
• Chinese low-cost models can accelerate enterprise adoption and token consumption, potentially enlarging total compute demand even as they pressure model pricing.

• Whether hyperscaler results demonstrate cloud and AI revenue conversion sufficient to offset capex and free-cash-flow pressure.
• Whether Project Jupiter or other power-constrained campuses show changes to tenant payments, debt terms, construction milestones or lender protections.
• Whether power bottlenecks push out component deliveries or produce cancellations beyond secondary semiconductor categories.
• Whether Chinese model adoption translates into commercial revenue and compute demand, or primarily compresses global model pricing and shifts profit pools.