The underlying risk picture is broadly unchanged. Strong cloud demand and constrained capacity continue to support the build-out, while heavy spending, weaker free-cash-flow conversion and power-permitting friction keep returns exposed. China’s DeepSeek-Huawei software work remains a competitive development, but no material China-driven change was reported since the prior briefing.
DAILY BRIEFING
AI risk remains elevated as China’s domestic stack advances while returns and power constraints remain unresolved.
The risk picture is unchanged at 3.125: cloud and accelerator demand remain strong, but capital spending and free-cash-flow pressure remain material. China is incrementally strengthening its domestic model-and-chip software stack, which could widen low-cost competition and shift future AI profits, though no material China-driven change to the investment cycle was verified today.
- Hyperscaler cloud demand remains robust, but Amazon and Alphabet both paired strong cloud results with sharply higher capital plans and weaker free cash flow.
- Project Jupiter keeps power availability, contractual risk allocation and project-finance timing in focus; Oracle and Blue Owl say commitments remain intact, but the force-majeure notice tests whether capacity can convert into revenue on schedule.
- Enterprise deployment remains uneven: Gartner found only 22% of surveyed organizations had scaled AI across multiple business units, despite continued planned spending.
- China is reducing friction in its domestic AI stack through DeepSeek-Huawei tools and Ascend ecosystem development; if adoption follows, capable lower-cost alternatives could pressure model and cloud economics outside China.
- AWS reported 37% year-on-year revenue growth, a $496 billion backlog and continuing capacity shortages, supporting the view that a meaningful portion of infrastructure demand is contracted rather than speculative.
- BEA research finds evidence consistent with AI use being associated with stronger output and productivity, although it does not establish causality.
- China’s domestic-chip progress is not yet proof of global displacement: Huawei’s performance and ecosystem claims are company statements, and developers still face switching and interoperability costs.
- Western open-weight releases are narrowing China’s recent lead, preserving alternatives for customers unwilling or unable to use Chinese model ecosystems.
- Whether Project Jupiter’s power delay changes rent, lender protections, construction milestones or funding terms.
- Whether hyperscaler backlogs convert into recognized cloud revenue and free cash flow rather than further contracted capacity.
- Evidence of enterprise AI scaling, pricing power and measured productivity gains beyond surveys and early correlations.
- Developer adoption and commercial deployment of DeepSeek-Huawei Ascend tools outside announced partnerships.
Competitive position: Verified evidence shows China advancing a more integrated domestic stack: DeepSeek has adapted a leading model to Huawei hardware, Huawei reports a growing Ascend software ecosystem, and Chinese official reporting cites domestic supercluster deployment and broad industrial AI use. Inference: this improves China’s ability to compete in cost-sensitive models and infrastructure, but does not establish parity with the leading global frontier stack or broad overseas commercial adoption.
Effect on the cycle: Cheaper capable Chinese models and a more usable Ascend stack could compress rents for proprietary US models and redirect some infrastructure spending and profits toward Chinese suppliers. They could also lower deployment costs and expand global AI adoption, increasing inference demand rather than reducing total compute demand. Current evidence supports competitive pressure, not a demonstrated global revenue displacement.
SOURCE LINKS AT PUBLICATION
Relevant reporting
These links point to third-party publishers. Availability, access and paywalls remain under each publisher’s control.
Oracle moves to shield itself from AI buildout risks
Oracle’s Project Jupiter force-majeure notice makes power availability and contractual allocation of delay risk directly relevant to data-centre revenue timing and project finance; Oracle and Blue Owl said commitments were unchanged.
CAPEX · CREDIT · MACRO ↗Tech consortium launches National Compute Grid for AI
The proposed shared-compute network targets idle capacity and fragmentation. If implemented at scale, it could improve utilization and temper incremental infrastructure demand; it is not yet evidence of reduced spending.
CAPEX · SEMIS · MACRO ↗AI Utilization and Economic Performance, October 2026
BEA says the evidence is consistent with stronger output and productivity where AI is used, but cautions that causal attribution remains difficult—supportive but incomplete evidence for returns.
MONETISATION · MACRO ↗Gartner Survey Finds Only 22% of Organizations Have Successfully Scaled AI Across Multiple Business Units
The survey indicates AI spending is rising faster than broad operational scaling and financial measurement, testing the pace at which enterprise demand converts into durable returns.
MONETISATION · CAPEX ↗Amazon lifts investment plans after strong cloud sales; shares jump
AWS revenue rose 37% and backlog reached $496 billion, but Amazon raised capital spending to $220 billion and reported negative trailing-12-month free cash flow—simultaneous demand support and cash-return pressure.
CAPEX · MONETISATION · SEMIS ↗Google increases capex forecast again after cloud-driven quarterly beat
Alphabet raised its 2026 capex outlook to $195 billion-$205 billion despite strong cloud demand and reported negative quarterly free cash flow, underscoring the investment-versus-cash-conversion tension.
CAPEX · MONETISATION ↗Huawei and DeepSeek release open source AI tools in a bid to lower Nvidia exposure
DeepSeek released Ascend-targeted infrastructure components with Huawei support. This lowers software-porting friction for domestic accelerators, but early repository traction and production adoption remain unproven.
OPEN-MODELS · SEMIS ↗Advancing the Agentic World, Building a Solid Silicon Foundation
Huawei says more than 40 models have been pre-trained on Ascend/CANN and that its ecosystem supports major open-source projects. These are company claims but indicate an expanding domestic software layer.
SEMIS · OPEN-MODELS ↗Digital News_English
Official reporting describes a 100,000-card domestic AI supercluster, model adaptation to Chinese chips and AI use in manufacturing. It supports evidence of capital intensity and adoption, while the underlying performance measures are not independently audited here.
SEMIS · OPEN-MODELS · MACRO ↗DeepSeek unveils new AI model tailored for Huawei chips as China pushes for tech autonomy
DeepSeek’s Huawei-adapted V4 demonstrates that a prominent Chinese model developer can use domestic accelerators for part of its stack, supporting the domestic-substitution thesis.
OPEN-MODELS · SEMIS ↗Revision to License Review Policy for Advanced Computing Commodities
The US policy permits case-by-case review for specified lower-performance advanced chips to China under supply, volume, security and end-use conditions. It may moderate, but does not remove, incentives for domestic substitution.
SEMIS · OPEN-MODELS ↗