AI demand still validates hardware spending, but locked-in capacity and financing raise cycle risk; China adds price pressure.

No reporting changes the rules-based scores. The measured change is the lower open-model score. Recent evidence tests the inputs: Dell reported strong AI-server orders and wider customer demand; Microsoft disclosed large Azure revenue; meanwhile, Anthropic capacity contracts, hyperscaler leases and Nvidia-linked financing increase the amount of future demand required to justify infrastructure. China-specific developments reinforce price-performance competition and domestic capacity-building, but do not verify a new China-driven break in global AI demand today.

• Aggregate hyperscaler capex growth and declining aggregate free-cash-flow growth remain the dashboard's primary imbalance.
• Large uncommenced data-centre leases and growing use of debt/project-style financing increase fixed-cost and refinancing sensitivity if AI utilisation disappoints.
• Texas and other power markets are filtering electricity requests after evidence that some data-centre demand is duplicative or weakly committed, challenging headline capacity pipelines.
• Cheap capable Chinese open-weight models can reduce proprietary-model pricing power and shift value from model vendors toward infrastructure, applications or users.

• NVIDIA Data Center revenue growth remains exceptionally strong in the authoritative semiconductor-demand proxy.
• Dell reported more than $130 billion of AI-server bookings over the prior 12 months and said demand was broadening beyond a narrow customer group.
• Microsoft's new Azure disclosure provides evidence of substantial cloud revenue at scale, improving transparency for testing infrastructure returns.
• Alibaba reported strong AI cloud and compute revenue growth, indicating that Chinese AI investment is accompanied by commercial demand, not only subsidised capacity.

• Whether new AI debt, lease and project-finance commitments obtain durable customer contracts and acceptable credit terms.
• Power-queue clean-ups and whether cancelled or delayed grid requests translate into slower realised data-centre construction.
• Azure's subsequent disclosed revenue trajectory versus capex, alongside AWS and Google Cloud growth.
• Whether Chinese open-weight vendors can sustain performance and monetise through licensing, cloud hosting and enterprise services.