State capital plays a decisive role in China's AI financing system

Overview

This report discusses the financing structure of China's AI industry. The core judgment is that China's AI capital expenditure is clearly different from the US model dominated by hyperscale cloud providers; instead, it is jointly driven by multi-level state capital, policy finance, capital markets, and some private internet giants. Facing advanced chip restrictions and the China-US technology gap, China is strengthening its self-reliant supply chain through semiconductors, AI chips, data centers, cloud services, and LLMs.

Core views

The report believes that central government funding is especially critical for bottleneck segments in the AI industry chain, including chip design, manufacturing, equipment, materials, and computing infrastructure. The Big Fund, ultra-long central government special treasury bonds, the National Venture Capital Guidance Fund, new policy financial instruments, and industrial subsidies together form the main financing axis. Overall direct funding support from local governments is relatively weak, mainly due to the property downturn and hidden debt resolution, but some cities with industrial foundations may still make concentrated efforts. The private sector is mainly focused on cloud services, LLMs, AI agents, and computing infrastructure.

Analysis framework

The report uses a method of policy fund mapping, financing instrument breakdown, and public-private sector comparison, dividing China's AI financing sources into central fiscal funds and state funds, policy bank equity capital, local government funds, capital market financing, industrial subsidies, and private enterprise capital expenditure, and compares this with the US AI capex structure dominated by private hyperscale cloud providers.

Methodology notes

• Industrial financing structurePublic-private sector capex comparison

  Compare the US private-sector-led AI investment model with China's state-capital-led model.

  The US AI ecosystem is mainly financed by hyperscalers, while China relies more on central and local governments, SOEs, policy banks, and state funds to compensate for constraints in advanced chips and core technologies.

• Policy funding analysisMulti-level state capital framework

  Assess AI financing through channels such as state funds, special treasury bonds, policy financial instruments, and local government bonds.

  This framework emphasizes the role of different policy tools at different stages of the industry chain, such as early-stage venture investment, project capital, data center networks, and key semiconductor segments.

• Capital market financingObservations on domestic and overseas listing financing for technology companies

  Observe financing support from the STAR Market and Hong Kong market for AI, chip, and PCB companies.

  Against the backdrop of an overall weak A-share IPO environment, technology companies still obtain funding through the STAR Market and Hong Kong IPOs, reflecting targeted capital market support for China's AI and semiconductor industries.

Asset mapping & comparison

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

• China's AI industry chain

  Core research subject

  Strengths

  State capital, policy finance, and capital markets form multi-channel financing support, with rising policy priority.

  Weaknesses

  Advanced chips are constrained, and some core technologies still lag behind the US.

  Comparison

  Compared with the US model dominated by hyperscalers, China relies more on government and state-owned capital investment in bottleneck segments.

  Risks

  Risks related to the pace of policy fund implementation, difficulty of technological substitution, external restrictions, and capital efficiency.

• Semiconductors and AI chips

  Bottleneck segments prioritized by state capital

  Strengths

  Big Fund Phase III, industrial subsidies, the STAR Market, and Hong Kong IPOs all support chip design, manufacturing, equipment and materials, and advanced packaging.

  Weaknesses

  High-end chips and key equipment/materials still face technological and external supply constraints.

  Comparison

  Compared with cloud services and LLM applications, chip manufacturing and upstream segments rely more on state capital.

  Risks

  Long R&D cycles, high capex intensity, and uncertain progress in import substitution.

• Data centers and power grid infrastructure

  Important carriers of AI computing capex

  Strengths

  The potential RMB2.0trn data center plan and at least RMB5.0trn total investment could significantly expand the AI computing base.

  Weaknesses

  The investment scale is large and requires high standards for power, networks, chips, and operational efficiency.

  Comparison

  Compared with traditional infrastructure, data center networks are beginning to receive clearer strategic funding support.

  Risks

  Risks related to project approval, funding sources, utilization rates, energy consumption, and meeting domestic technology standards.

• Local government funds

  Auxiliary source of support

  Strengths

  Some cities can launch concentrated large-scale support plans around advantaged industry chains, such as Wuhan's memory chip initiative.

  Weaknesses

  The property drag and hidden debt resolution constrain local fiscal capacity, resulting in a low overall share of investment.

  Comparison

  Compared with central fiscal funding and policy finance, direct local government funding support is clearly weaker.

  Risks

  Local fiscal pressure, debt swap priority, and homogenization of regional projects.

• STAR Market and Hong Kong equity market

  Equity financing channels for technology companies

  Strengths

  GPU, advanced packaging, PCB, and LLM companies obtain funding through listings.

  Weaknesses

  The overall A-share IPO environment remains cautious, and financing windows may be affected by market sentiment.

  Comparison

  The Hong Kong market attracts mainland technology companies by relaxing IPO rules, while the STAR Market maintains targeted support for key technology companies.

  Risks

  Valuation volatility, regulatory pace, post-listing liquidity, and pressure to deliver earnings.

• Cloud service providers and LLM companies

  Main force of private-sector AI capex

  Strengths

  ByteDance, Alibaba, Tencent, and Baidu continue to expand investment in AI infrastructure, cloud services, and model R&D.

  Weaknesses

  The payback period for capex may be long, and model commercialization and computing utilization still need to be verified.

  Comparison

  Compared with semiconductor hardware bottleneck segments, private capital is more concentrated in cloud, LLMs, AI agents, and application infrastructure.

  Risks

  Intensifying competition, computing cost, regulatory changes, and demand growth falling short of expectations.

Key data

• Registered capital of Big Fund Phase IRMB139bnEstablished in September 2014, with a focus on downstream chip manufacturing, packaging, and testing.
• Registered capital of Big Fund Phase IIRMB204bnEstablished in October 2019, with focus shifting toward upstream semiconductor equipment and materials.
• Registered capital of Big Fund Phase IIIRMB344bnEstablished in May 2024, covering core bottleneck areas and expanding to PCB and AI.
• Total support scale of the three phases of the Big Fund近RMB700bnUsed for state capital support related to semiconductors and AI.
• National AI Industry Investment FundRMB60bnParticipated in its establishment by Big Fund Phase III in January 2025.
• Total amount of 2026 ultra-long central government special treasury bondsRMB1.3trnOf this, RMB800bn is used for the "Two Major" areas, the same scale as in 2025.
• Second batch of 2026 "Two Major" fundsRMB217bnAnnounced in April 2026, with AI explicitly listed as one of the key support directions.
• Initial capital of the National Venture Capital Guidance FundRMB100bnSupported by ultra-long central government special treasury bonds, with a 20-year term and investment in integrated circuits, AI, biopharma, quantum technology, and 6G, among others.
• Bloomberg-reported data center planRMB2.0trnReportedly for a nationwide data center network over the next five years, with at least 80% localization of core technologies.
• Potential total investment including grid supporting infrastructure至少RMB5.0trnFunding may come from ultra-long central government special treasury bonds, national industry funds, bank loans, and private capital.
• Initial quota of the new policy financial instrumentRMB500bnLaunched in September 2025, with an additional RMB800bn quota in the 2026 fiscal budget.
• Potential investment leveraged by the initial RMB500bn policy financial instrument超过RMB7trnAccording to NDRC figures, it has strong leverage effects when used as project capital.
• Net financing of local government bonds in 2025RMB5.4trnOf this, only RMB140bn, or 2.7%, went to new infrastructure and strategic emerging industry infrastructure.
• Relevant proportion of local funding allocation in 1Q261.0%Showing that overall direct local government funding support for AI investment is limited.
• Wuhan memory chip support planRMB260bnLaunched in April 2026 to support YMTC and XMC capacity expansion.
• Decline in financing costs for AI-related bond issuers in 2025约35bpA larger decline than the 24bp drop in average corporate loan rates, showing policy-driven financing advantages.
• Direct financing for AI companies in 2025约RMB250bnMedium- and long-term bond issuance expanded significantly.
• Issuance growth of sci-tech innovation bonds and similar instruments in 2025超过71%Broadening financing channels for AI companies.
• Moore Threads STAR Market IPORMB8bnA financing case for a domestic GPU design company.
• MetaX STAR Market IPORMB3.9bnA financing case for a domestic GPU design company.
• SJ Semiconductor STAR Market IPORMB4.8bnA financing case for an advanced packaging company.
• Victory Giant Technology Hong Kong IPOHKD23bnA financing case for a PCB company and the largest IPO in Hong Kong that year.
• Montage Technology Hong Kong dual listing IPOHKD8bnA financing case for a computing chip design company.
• Zhipu AI and MiniMax Hong Kong financingHKD4.4bn; HKD5.5bnFinancing cases for LLM companies.
• Annual AI capital expenditure of major domestic hyperscalers约RMB500bnForming a private-sector AI investment driver beyond public funding.
• ByteDance 2026 AI capital expenditure plan超过RMB200bn25% higher than the preliminary RMB160bn budget discussed at the end of 2025.
• Alibaba three-year cloud computing and AI hardware investment plan超过RMB380bnAnnounced in February 2025, and later stated that its five-year AI infrastructure budget would significantly exceed the previous plan.
• Tencent AI-related capital expenditure in 1Q26RMB37bnUp 16% year-on-year, higher than Alibaba's RMB27bn in the same period.
• Baidu cumulative AI capital and R&D investment超过RMB100bnSupporting Ernie Bot, autonomous driving, and AI computing infrastructure.

Impact & implications

The main investment implication of the report is that the financing resilience of China's AI industry chain depends to a large extent on the pace of policy funding, project implementation of policy financial instruments, the capital market financing window for technology companies, and sustained capital expenditure by cloud service providers and LLM companies. If Beijing accelerates bond issuance and fiscal spending, AI may gain higher priority in the strategies for stabilizing investment and achieving technological self-reliance, thereby benefiting semiconductor equipment and materials, AI chips, advanced packaging, PCB, data centers, grid supporting infrastructure, and cloud infrastructure.

Risks

• The actual implementation speed of ultra-long special treasury bonds, policy financial instruments, and state funds may be lower than expected.
• AI-related projects involve large-scale capital expenditure, and insufficient utilization could lead to lower capital efficiency.
• The US and other external technology restrictions may continue to affect the supply of advanced chips, equipment, and materials.
• Local fiscal pressure may limit sustained direct local government investment in the AI industry.
• Although private internet companies have high AI capex, uncertainty remains around commercialization returns, cloud demand, and LLM profit models.
• Capital market financing windows may be affected by regulatory pace, market sentiment, and valuation volatility.

What to watch

• The issuance pace of subsequent 2026 ultra-long central government special treasury bonds and the share allocated to AI-related projects.
• The allocation, project list, and actual leveraged investment scale of the additional RMB800bn quota for the new policy financial instrument.
• Specific investment progress of Big Fund Phase III in AI chips, semiconductor equipment and materials, PCB, and AI companies such as DeepSeek.
• The pace of capital deployment by the National Venture Capital Guidance Fund into early-stage AI, integrated circuit, quantum, and 6G projects.
• Whether the RMB2.0trn data center network plan is formally implemented, and the required ratio of domestic core technologies.
• The financing window in the STAR Market and Hong Kong market for chip, advanced packaging, PCB, LLM, and cloud infrastructure companies.
• Whether AI capex by private companies such as ByteDance, Alibaba, Tencent, and Baidu continues to be revised upward.