Following the Politburo meeting, policy remains supply-side oriented and the AI capex cycle continues
AI summary card
Following the Politburo meeting, policy remains supply-side oriented and the AI capex cycle continues
Morgan Stanley believes that China's policy focus remains on accelerating the implementation of fiscal and quasi-fiscal tools, with September and October potentially serving as a catalyst window for easing, while global AI capex and inference demand remain strong.
- Beijing still has approximately RMB 2 trillion in unused budgetary and quasi-fiscal impulses available for the second half of 2026.
- If economic activity fails to stabilize in July and August, the likelihood of additional easing in September and October will increase.
- The remaining government bond quota for the second half of 2026 is RMB 7.5 trillion, higher than the RMB 5.7 trillion in the second half of 2025.
- The scale of new policy financial instruments in 2026 is RMB 800 billion, higher than RMB 500 billion in 2025.
- Hyperscalers will significantly increase AI-related capex over the next two years, while AI inference demand and usage duration of key AI applications continue to grow.
Report interpretation
Overview
This report discusses the policy path following China's Politburo meeting, potential fiscal catalysts in the second half of 2026, and the impact of the AI capex cycle on the economy and assets. The report's core view is that policy remains focused on the supply side and budget execution, with no comprehensive policy shift yet; meanwhile, the AI capex supercycle continues, and growth in inference demand supports investment returns for related business models.
Core views
The report believes that the policy signals released by the Politburo meeting primarily call for accelerating fiscal deployment, advancing the use of policy financial instruments and local government special-purpose bonds, and increasing the share of people's livelihood projects in government investment. China's economy continues to display a "two-speed" pattern: exports remain resilient, but domestic demand is weakening, fiscal deployment did not accelerate noticeably in July, and reflation remains narrow. On AI, the report expects hyperscalers to significantly increase AI capex over the next two years, with usage duration of key AI applications and inference demand continuing to rise, while GenAI-related business models could achieve incremental ROIC of more than 25%.
Analysis framework
The report combines analysis of policy signals, observation of fiscal implementation progress, assessment based on PMI and high-frequency activity data, analysis of price pass-through and corporate profit structures, and analysis of AI capex and unit economics to evaluate China's macro policy path and the AI investment cycle.
Methodology notes
Whether policy is shifting
The report assesses the policy focus through the wording of the Politburo meeting, the deployment of fiscal tools, local government special-purpose bonds, and the advancement of major projects. Its conclusion is that fiscal implementation is accelerating, but no comprehensive policy shift has emerged.
Two-speed economy
Taking into account the slowdown in PMI from June to July, resilient exports, and weak domestic demand, the report concludes that China's economy continues to exhibit a divergent pattern of strong external demand and weak domestic demand.
AI Capex Supercycle
The report evaluates the investment return potential of GenAI business models through analysis of cloud providers' capex, AI inference demand, application usage duration, and unit economics.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Chinese macroeconomic and policy-related assetsAffected by fiscal deployment and policy implementation
- Strengths
- Significant unused fiscal and quasi-fiscal capacity remains, with potential easing catalysts in September and October.
- Weaknesses
- Fiscal implementation had not accelerated noticeably in July, while domestic demand remains weak.
- Comparison
- The remaining government bond quota for the second half of 2026 is higher than in the same period of 2025.
- Risks
- Continued delays in fiscal implementation, policy strength below expectations, and domestic demand recovery falling short of expectations.
- AI infrastructure and cloud service providersDriven by growth in AI capex and inference demand
- Strengths
- Hyperscalers plan to significantly increase AI-related capex over the next two years, while usage duration of key AI applications continues to grow.
- Weaknesses
- Capex intensity is high, and returns depend on the continued release of inference demand and commercialization efficiency.
- Comparison
- The report describes AI capex as remaining in a supercycle.
- Risks
- AI application monetization slower than expected, excess computing capacity, and rising energy and depreciation cost pressures.
Key data
- Unused fiscal and quasi-fiscal impulseApproximately RMB 2 trillionThe report states that Beijing still has approximately RMB 2 trillion in unused budgetary and quasi-fiscal support capacity in the second half of 2026.
- Remaining government bond quota for H2 2026RMB 7.5 trillionCompared with RMB 5.7 trillion in the second half of 2025, this indicates room for fiscal issuance to accelerate.
- Scale of 2026 policy financial instrumentsRMB 800 billionHigher than RMB 500 billion in 2025.
- Interest subsidy budgetBelow RMB 100 billionThe report notes that interest subsidies may be expanded moderately, but remain within the budgeted amount.
- Incremental GenAI ROICAbove 25%, with a potential path to 25%–50%The report believes that three major business models in the AI inference era offer attractive unit economics and investment returns.
Impact & implications
For markets, September and October may become an important window for observing additional Chinese policy support; if activity data remains weak in July and August, the implementation of fiscal and quasi-fiscal tools may accelerate. In the AI value chain, continued growth in capex and inference demand supports demand for cloud services, computing infrastructure, energy, and related equipment. At the macro level, weak domestic demand, slow fiscal implementation, and insufficient price pass-through continue to constrain the strength of reflation.
Risks
- If economic activity fails to stabilize in July and August, this would indicate that weak domestic demand may persist.
- If fiscal deployment and government bond issuance remain slow, the transmission of policy support to the real economy may be insufficient.
- Weak price pass-through may cause reflation to remain concentrated in upstream and select downstream industries.
- Returns on AI capex depend on inference demand, application usage duration, and business model implementation, creating a risk that investment returns fall short of expectations.
What to watch
- Whether additional easing or faster fiscal implementation emerges in September and October.
- The progress of government bond issuance, policy financial instruments, and local government special-purpose bond deployment.
- Whether PMI, domestic demand, high-frequency activity, and cement shipments improve.
- AI-related capex plans, inference demand, and usage duration of key AI applications.
- Whether improvements in industrial profits broaden from upstream to downstream industries.