Fiscal Acceleration Window Approaching, AI Capex Supercycle Continues

Overview

The report assesses China’s economy and the technology investment cycle by combining policy signals from the Politburo meeting, fiscal execution progress, PMI and price indicators, as well as AI capex and generative AI business models. The core conclusion is that policy remains focused on accelerating the deployment of fiscal tools and major projects rather than shifting comprehensively toward demand stimulus; meanwhile, the AI capex supercycle is still advancing, and investment by hyperscale cloud service providers, inference demand, and application usage are expected to continue growing over the next two years.

Core views

The two-speed nature of China’s economy remains evident: exports remain resilient, while domestic demand continues to weaken, with the combined slowdown in June to July PMI exceeding seasonal levels. July government bond issuance and cement shipments have not yet shown a clear acceleration in fiscal deployment, and the gap between input and output prices for industrial products also reflects weak price pass-through, with profit improvement concentrated in upstream and a few downstream industries. There is still substantial room for policy execution, including accelerating government bond issuance, deploying new policy-based financial instruments, moderately expanding the scope of interest subsidies, and advancing the 109 major projects under the 15th Five-Year Plan. If economic activity fails to stabilize in July to August, the probability of additional easing in September to October will rise. In the technology cycle, AI-related capex is expected to increase significantly over the next two years, inference demand continues to grow, and the incremental unit economics and return on capital of the three major AI inference business models are attractive.

Analysis framework

The report uses policy text interpretation, comparisons of fiscal quotas and execution progress, tracking of PMI as well as sector price and profit indicators, and cross-analysis with AI infrastructure capex, inference demand, application usage time, and business model returns on capital.

Methodology notes

• Policy AnalysisPolicy Signals and Fiscal Execution Analysis

  Distinguishing changes in policy direction from accelerated implementation of existing policies

  By examining Politburo meeting language, remaining government bond quotas, policy-based financial instruments, and requirements to advance major projects, the report judges that current policy is more tilted toward strengthening budget execution rather than a comprehensive policy pivot.

• Macro AnalysisHigh-Frequency Economic Indicator Tracking

  Verifying economic momentum from growth, fiscal, and price dimensions

  Using PMI, government bond issuance, cement shipments, the gap between input and output prices, and the distribution of industrial profits, the report evaluates the actual progress of domestic demand, fiscal transmission, and reflation.

• Industry and Return AnalysisAI Unit Economics and Return on Capital Analysis

  Assessing capex sustainability through inference demand and business models

  By combining factors such as capex by hyperscale cloud service providers, AI application usage time, compute leasing costs, and operating costs, the report assesses the incremental return on capital potential of generative AI business models.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• AI Infrastructure and Computing Power Supply Chain

  Benefits from hyperscale cloud service providers increasing AI-related capex over the next two years and growth in inference demand.

  Strengths

  The capex cycle remains strong, AI application usage time is growing, and inference business models show relatively attractive incremental unit economics.

  Weaknesses

  Upfront investment is high, and returns depend on utilization, pricing power, energy costs, and continued commercialization.

  Comparison

  Compared with overall China macro demand, the AI investment cycle shows stronger structural growth momentum.

  Risks

  Demand growth falls short of expectations, compute leasing and energy costs rise, or business model returns are lower than estimated.

• Chinese Government Bonds and Policy-Based Financial Instruments

  They are the main vehicles for accelerated fiscal implementation and growth-stabilization policy transmission in the second half of 2026.

  Strengths

  The remaining government bond quota is relatively large, and the scale of new policy-based financial instruments is higher than the previous year.

  Weaknesses

  July issuance and actual project transmission have not yet clearly accelerated.

  Comparison

  The remaining government bond quota in the second half of 2026 is higher than in the same period of 2025.

  Risks

  Delays in issuance or use of funds, slower-than-expected project implementation, or a fiscal multiplier below expectations.

• China Cyclical and Domestic Demand-Related Assets

  Their performance depends on fiscal execution, stabilization of domestic demand, and improvement in price pass-through.

  Strengths

  There is still substantial policy space, and additional easing catalysts may emerge in September to October.

  Weaknesses

  PMI is weakening, domestic demand is soft, fiscal deployment has not yet accelerated, and the scope of profit improvement is limited.

  Comparison

  Compared with the export sector, economic momentum in domestic demand-related areas is weaker.

  Risks

  Domestic demand continues to decline, the scope of reflation remains narrow, or policy implementation is later than expected.

Key data

• Potential fiscal and quasi-fiscal stimulus space in the second half of 2026Approximately RMB 2 trillionRefers to unused budgetary and quasi-fiscal policy space.
• Remaining government bond quota in the second half of 2026RMB 7.5 trillionHigher than RMB 5.7 trillion in the second half of 2025.
• Scale of new policy-based financial instruments in 2026RMB 800 billionThe corresponding scale in 2025 was RMB 500 billion.
• Interest subsidy budget rangeWithin RMB 100 billionPolicy may moderately expand coverage within the established budget range.
• Potential return on capital for generative artificial intelligence25% to 50%The report believes the three major AI inference business models can achieve an incremental return on capital of at least 25%, but actual results depend on demand, costs, and commercialization assumptions.
• Number of major projects109The Politburo meeting called for advancing the implementation of major projects listed in the 15th Five-Year Plan.

Impact & implications

At the macro level, the large remaining fiscal quota provides support for growth stabilization in the second half of 2026, but the ultimate effect depends on whether bond issuance, policy-based financial instrument deployment, and project construction can accelerate. If domestic demand remains weak, a stronger policy response may emerge in September to October. At the industry level, AI capex, inference demand, and application activity jointly support the prosperity of computing infrastructure and related supply chains, but returns on capital still need to be realized through revenue growth, higher utilization, and cost control.

Risks

• Economic activity fails to stabilize in July to August, and weak domestic demand continues further.
• Government bond issuance, deployment of policy-based financial instruments, or implementation of major projects is slower than expected.
• Insufficient pass-through from input prices to output prices, with reflation and corporate profit improvement continuing to be concentrated in a few industries.
• AI inference demand or application commercialization falls short of expectations, putting pressure on capex returns.
• Rising energy, depreciation, compute leasing, and other operating costs weaken the return on capital of generative AI projects.
• The report’s forecasts are based on assumptions that may not materialize, and public information may also be incomplete or subject to subsequent revisions.

What to watch

• Whether PMI, domestic demand, and economic activity can stabilize in July to August.
• Whether additional fiscal or monetary easing measures are introduced in September to October.
• The issuance and use progress of the RMB 7.5 trillion remaining government bond quota in the second half of 2026.
• The deployment pace and project coverage of the RMB 800 billion in new policy-based financial instruments.
• Implementation progress of the 109 major projects under the 15th Five-Year Plan.
• Whether cement shipments, price pass-through, and industrial profit improvement spread from upstream industries to more downstream industries.
• AI capex guidance, inference demand, and usage time for major AI applications from hyperscale cloud service providers over the next two years.
• Whether generative AI business models can achieve an incremental return on capital above 25%.