With the US model already in place, when will the AI financing boom flow into the domestic bond market?

With the US model already in place, when will the AI financing boom flow into the domestic bond market?

When Tencent's free cash flow turned negative for the first time in the second quarter of 2026 and Alibaba launched its first share placement in seven years since its IPO, a question was put on the table for bond investors: When will the AI financing wave truly flow into the domestic bond market?

On August 26th, the fixed income team at Huatai Securities (Zhang Jiqiang, Qiu Wenzhu, Wen Chenxin, etc.) analyzed in their research report, "AI Financing: From Cash Machine to Capital Machine," that overseas AI giants have shifted from "cash machines" to "capital machines." The domestic bond market has no problem absorbing this demand; the issue is insufficient financing demand. Increased supply presents more of an opportunity for investors, but whether key variables can loosen will determine when this opportunity will materialize. The report estimates:

The proportion of debt financing in China's AI blockchain sector remains low overall, but the financing demand from major domestic companies has increased since Q2 2026. Based on the assumptions of a 2%, 10%, and 20% CAGR for Capex over 2026-2030, the total domestic AI financing demand from 2026 to 2030 is estimated at 1.27-3.5 trillion yuan, corresponding to a new domestic bond supply of 91.3-260.8 billion yuan, with an average annual supply of 18.3-52.2 billion yuan . The bond market has no worries about absorbing this supply. The key is how to address bottlenecks related to the intended use, currency, and maturity of AI-related bonds to release their supply. In the short term, AI financing may lead to structural changes in bond supply, impacting various types of bonds such as STAR Market bonds, Panda bonds, REITs, and offshore bonds.

Turning point signal: Financing demand from major domestic companies begins to rise

Tencent:

In Q2 2026, Tencent's capital expenditure reached RMB 52.784 billion, a year-on-year increase of 176%, mainly used for AI computing power procurement and infrastructure construction. As a result, the quarterly free cash flow dropped to -RMB 13.8 billion—the first time Tencent has experienced negative free cash flow in a single quarter since it disclosed this data in Q2 2014.

Meanwhile, on June 17, Tencent issued RMB 15 billion in dim sum bonds and USD 2.45 billion in bonds .

Alibaba:

Alibaba's capital expenditures in Q1 of fiscal year 2027 (Q2 of calendar year 2026) reached RMB 67.678 billion, a year-on-year increase of 75%, while free cash flow in the quarter fell to -RMB 44.67 billion.

On August 23, Alibaba announced its plan to place HK$80 billion in new shares , all of which will be invested in AI development. This is the first time it has launched a new share placement since its listing on the Hong Kong Stock Exchange in 2019.

Huatai Securities pointed out in its research report: "The second quarter of 2026 already showed a trend of accelerated capital expenditure and increased demand for external financing."

This closely matches the evolutionary path overseas—except that China is at an earlier stage.

The US Model: The Three-Phase Process Completed, Risks Are Shifting

Overseas AI financing has gone through three clear steps:

Phase 1 (2022-2023) : Model validation period, with funding primarily tied to strategic equity and cloud resources. Microsoft's investment in OpenAI, Google's investment in Anthropic, and Amazon's investment in Anthropic all fall into this category. At that time, the six giants (Microsoft, Alphabet, Amazon, Meta, Oracle, and Nvidia) were still "cash cows," with their combined operating cash flow increasing from approximately $199.8 billion to approximately $368.2 billion between 2019 and 2023.

Phase Two (2024-2025) : Capex devours free cash flow, and investment-grade bonds transform from low-frequency instruments to the norm. The Capex/OCF ratio of the six major firms rises from approximately 36% in 2019 to approximately 60% in 2025. In 2025, they will collectively issue over $100 billion in investment-grade bonds, nearly three times the average from 2020 to 2024.

Phase Three (2026 to Present) : Capital Stack Mobilization. As of July 2026, the six giants had issued approximately $220 billion in bonds this year, more than double the total for 2025. Financing instruments have expanded horizontally to equity, hybrid equity, GPU asset-backed securities, project financing, SPVs, and revolving credit; and vertically, different repayment tiers have been established.

A research report from Huatai Securities points out: "Risks are shifting from the balance sheet to contracts, guarantees, and structures."

A prime example is CoreWeave—based on GPU hardware and long-term customer contracts, it packages computing assets into collateralizable and ratingable financing vehicles, with its $8.5 billion deferred drawdown term loan ultimately receiving an A3 investment-grade rating from Moody's. Another example is Nvidia: providing approximately $250 billion in financing guarantees for OpenAI , transforming itself from a chip supplier into a credit provider.

The gap between China and the US: not just a stage, but also a structure.

Huatai Securities' research report summarized three major structural differences:

First, they are at different stages.

Capital expenditures of leading domestic technology companies are still largely covered by internal revenue. In 2025, the combined capital expenditures of eight leading internet giants, including ByteDance, Alibaba, and Tencent, amounted to approximately 455 billion yuan, equivalent to one-sixth of the expenditures of the six major US cloud vendors during the same period. The overall internal revenue coverage rate was approximately 90%, while the three major telecom operators approached 100%.

Second, the investment entities are different.

Overseas, the market is dominated by large-scale cloud companies; domestically, private internet companies and central state-owned operators are the main players, supplemented by third-party IDC and local state-owned computing power platforms.

Third, the financing structures are different.

  • Internet giants: Approximately 50% of their revenue is covered by internal cash flow; debt financing is mainly comprised of overseas bonds (55%) and overseas syndicated loans (31%), with domestic bonds accounting for only about 1%.
  • The three major telecom operators primarily rely on lease liabilities and have not yet issued bonds.
  • Third-party IDC: Primarily relying on bank loans, they are beginning to explore multi-tiered REITs.
  • Semiconductor companies: primarily relying on bank loans

A research report by Huatai Securities points out that the reasons restricting domestic bond issuance are multifaceted: "The bulk of AI capital expenditure is in the procurement of high-end chips, which is currently generally settled in US dollars... If domestic RMB financing is used to cover foreign currency expenditures, there are exchange costs and cross-border capital risks." In addition, Tencent, Alibaba, and others are all red-chip VIE structures, which presents institutional frictions for domestic bond issuance; the maturity of domestic private enterprise bonds is generally short, and the base of long-term buyers is weak.

Key estimates: Average annual new supply of domestic bonds is 18.3-52.2 billion yuan.

Huatai Securities calculated the financing needs for domestic AI construction from the perspective of investment entities, covering 8 internet platforms, the three major telecom operators, 2 third-party data centers, and 4 semiconductor companies.

Three scenarios are set, with 2026 as the base year:

Under the three scenarios, the cumulative Capex will be approximately RMB 5.56 trillion, RMB 6.52 trillion and RMB 7.95 trillion in 2026, 2027 and 2030 respectively, corresponding to an annual average of approximately RMB 1.11 trillion, RMB 1.30 trillion and RMB 1.59 trillion.

Based on this calculation, the cumulative funding gaps for 2026-2030 under the three scenarios are approximately RMB 1.08 trillion, RMB 1.94 trillion and RMB 3.27 trillion respectively, with cumulative financing needs of approximately RMB 1.27 trillion, RMB 2.15 trillion and RMB 3.5 trillion respectively.

The research report stated:

Based on the assumptions of a CAGR of 2%, 10%, and 20% for Capex over 2026-2030, we estimate the cumulative financing demand for AI infrastructure development in China to be approximately RMB 1.27 trillion, RMB 2.15 trillion, and RMB 3.5 trillion, respectively. This financing demand will primarily be met through bank loans and domestic and international bonds. Under these three scenarios, the cumulative new supply of domestic bonds is estimated at RMB 913 billion, RMB 1587 billion, and RMB 2608 billion, respectively, from 2026 to 2030. The domestic bond market is large enough to absorb this demand; the issue lies in insufficient financing demand, and the increased supply presents more of an opportunity for investors.

Domestic bank loans account for more than 70% of the total financing through the two channels, and remain the main way for the domestic financial system to meet the financing needs of AI. Based on the aforementioned financing structure assumptions, the proportion of domestic bonds will gradually increase during the calculation period, with an average annual increase of approximately RMB 18.3-52.2 billion.

By entity, internet platforms remain the largest source of domestic debt supply, contributing RMB38.7-123.2 billion under the three scenarios, accounting for 42%-47%. The three major telecom operators show the greatest elasticity, with their share rising from 7% to 23% under the high-growth scenario.

Where are the opportunities? Four types of products worth paying attention to.

A research report from Huatai Securities points out that short-term AI financing will bring more structural changes than an overall impact.

Science and Technology Innovation Bonds : Highest Alignment with AI Financing Needs. Since the launch of the "Technology Bond Market" in May 2025, science and technology innovation bonds bearing labels such as "computing infrastructure," "token computing power factory," and "integrated storage and computing" have been issued. However, currently, newly issued science and technology innovation bonds are still mainly concentrated in traditional industries such as power, banking, and construction engineering, with a relatively low proportion from electronics, computers, and communications. Research reports predict: "In the future, as debt financing for hard-tech companies increases, the 'technological content' of science and technology innovation bond issuers is expected to improve."

Panda bonds : Catering to the RMB financing needs of red-chip Chinese and foreign entities. From January to July 2026, the issuance of foreign-funded Panda bonds already exceeded the total for 2025, with net financing reaching RMB 70.35 billion, and foreign ownership accounting for 54%. Currently, foreign technology companies have not yet participated. If foreign AI industry chain companies combine RMB financing with their computing power investments in China, both the scale and issuer structure of Panda bonds are expected to benefit.

Multi-tiered REITs : Addressing the exit and recycling of existing computing power assets. In August 2025, Southern Wanguo Data Center REIT and Southern Runze Technology Data Center REIT became the first batch of publicly offered data center REITs in China, raising a total of 6.9 billion yuan, and both have issued announcements regarding planned expansion. Regarding inter-institutional REITs, companies such as Century Internet, Wanguo Data, and Aofei Data already have products in operation, with a cumulative issuance scale of 10.767 billion yuan from March 2025 to August 2026.

Dim Sum Bonds/Chinese USD Bonds : Offshore Markets Cater to Long-Term Demand. In June 2026, Tencent issued 30-year Dim Sum bonds, with sovereign wealth funds and insurance companies subscribing to 47%, exceeding the subscription ratio of banks. A research report points out: "Currently, the supply of ultra-long-term corporate bonds in the domestic market remains relatively scarce. If the supply of long-term corporate bonds from high-grade technology companies expands, it will help meet the allocation demand of insurance funds."

Key variables: Chip bottlenecks and regulatory loosening

A research report from Huatai Securities clearly states that whether the supply of domestic bonds can increase significantly in the future depends on two key variables:

Firstly, the progress of domestic production of advanced chips. Currently, due to export restrictions, the actual computing power that Chinese AI companies can deploy is limited by chip availability. Even if they are willing to borrow, the funds are difficult to fully convert into effective investment. The research report points out: "If breakthroughs are achieved in key technologies such as domestically produced GPUs, some of the financing needs currently being met by overseas syndicated loans and overseas bonds may also be shifted to domestic channels." The estimated overseas bond financing needs of internet platforms, ranging from approximately 470 billion to 1.2 trillion yuan during the calculation period, have the potential to shift to domestic channels.

Secondly, the ease of domestic bond issuance by internet platforms has improved. This includes: reducing institutional friction in bond issuance by red-chip VIE structures, improving credit enhancement mechanisms for private enterprise bond issuance, and cultivating a base of domestic long-term bond buyers.

The research report concludes: "The shift in financing structure from equity-dominated to a combination of equity and debt will become an important variable in the supply of Chinese corporate bonds."

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.