AI spending and credit structures remain stretched, while China’s capable lower-cost models add competitive pressure without a new break.

The underlying risk picture is broadly unchanged. Recent developments show continued large commitments to AI chips, data-centre capacity and financing, alongside tighter scrutiny of whether proposed power demand is real and financeable. China’s impact remains two-sided: Chinese model capability, funding and cloud commercialisation add pressure to model economics, but domestic chip bottlenecks and rising memory costs limit the speed and cost advantage of China’s infrastructure buildout.

• Aggregate capex growth of 72.29% sits alongside a 16.11% decline in aggregate free-cash-flow growth, leaving returns vulnerable if demand or pricing disappoints.
• AWS and Google Cloud revenue growth of 24.30% remains well below the 64.73% capex growth comparator, sustaining the investment-versus-monetisation gap.
• Reuters Breakingviews reports roughly $500 billion of data-centre debt issuance this year, with risk differentiating sharply for projects lacking permits or secured power.
• Texas and other jurisdictions are filtering data-centre interconnection requests after evidence of duplicative or weakly funded demand; this raises execution and power-cost risk for marginal projects and developers without firm commitments or deposits at risk of loss today? Actually note omit deposit loss uncertain.,

• NVIDIA Data Center revenue grew 117% year over year in the latest reported quarter, indicating that accelerator demand has not yet broken.
• Amazon’s long-term arrangement to purchase up to $60 billion of Qualcomm AI data-centre products indicates sustained demand for inference infrastructure and diversification beyond incumbent accelerator supply.
• Firmus signed OpenAI as an anchor customer for Malaysian capacity, lifting its contracted capacity above 900 MW; contracted offtake supports infrastructure utilisation, though contract values were not disclosed.
• Alibaba reported 45% growth in AI Cloud and Compute Services revenue, a 133% increase in cloud adjusted EBITA and continuing triple-digit growth in AI-related revenue, evidence that AI commercialisation can produce revenue and margin expansion in at least one major Chinese cloud platform.

• Whether large data-centre proposals convert into binding power connections, permits, deposits and operating facilities rather than being culled as speculative requests.
• Whether hyperscaler and AI-lab financing migrates further toward guarantees, leases, project debt and public support, and whether lender protections or pricing tighten.
• Whether enterprise AI revenue and cloud margins accelerate enough to narrow the capex-to-monetisation gap.
• Whether HBM constraints and higher domestic accelerator prices constrain China’s compute buildout, or whether domestic supply expansion offsets them.