AI demand remains strong while financing and power constraints persist; China adds price pressure but no new cycle break.

The underlying risk picture is broadly unchanged. NVIDIA’s reported data-centre demand remains exceptionally strong, while large-scale third-party financing, wider hyperscaler borrowing costs and power-project screening show that the buildout is becoming more capital-intensive and execution-dependent. China continues to pressure model pricing through capable lower-cost open-weight offerings, but no material China-driven change was verified today.

• Aggregate hyperscaler capex growth and weaker aggregate free-cash-flow growth leave returns dependent on future monetisation rather than current cash conversion.
• Hyperscaler bond supply has risen sharply; Reuters reported wider spreads, lower order cover and higher new-issue concessions as investors absorb greater issuance.
• Texas and other jurisdictions are filtering data-centre power requests; utilities have reduced demand pipelines after imposing deposits and other financial safeguards, exposing potential overstatement in proposed capacity.
• Chinese open-weight models offer a credible low-cost alternative for some workloads. If adoption shifts materially, lower inference prices could weaken the revenue base underwriting Western infrastructure commitments.

• NVIDIA reported $89.0 billion of quarterly data-centre revenue, up 117% year over year, and forecast $108.0 billion of total revenue for the next quarter, consistent with still-strong infrastructure demand.
• Amazon and Qualcomm’s collaboration on inference chips and 1.6Tbps connectivity indicates continued investment in lower-cost, higher-throughput inference rather than a retreat from deployment.
• Power and cooling suppliers report rising orders and backlogs, supporting the view that physical infrastructure demand is real even as individual grid-request pipelines are cleaned up.
• Cheaper Chinese models can broaden adoption and token demand, potentially increasing overall compute consumption even if model-provider pricing and profit pools compress.

• Hyperscaler quarterly capex guidance, free cash flow and disclosed cloud or AI revenue conversion.
• Bond pricing, order books and refinancing terms for major AI infrastructure borrowers.
• Grid-connection approvals, deposits, cancellations and contracted-versus-requested power capacity.
• Evidence of enterprise workload migration to lower-cost Chinese or open-weight models, including effects on token volumes and cloud demand.