Industry Experts Connect | Exclusive Interview with the President of China Chengxin International: How AI Credit Ratings Maintain Risk Boundaries Amid Explosive Growth in Tech Innovation Bonds

Ask AI · How do rating agencies balance support for technological innovation with risk prevention?

China News Service Jingwei, May 7 (Luo Kun) — One year after the launch of the “Technology Board” in the bond market, the tech innovation bond market has entered a rapid development phase.

In this round of reshaping the bond market ecosystem led by technological innovation, what role do rating agencies play? How to balance supporting tech innovation and preventing risks? In response to these hot topics in the bond market, China News Service Jingwei recently interviewed Yue Zhigang, President of China Chengxin International Credit Rating Co., Ltd. (hereinafter referred to as China Chengxin International).

Photo of Yue Zhigang, President of China Chengxin International, provided by the interviewee

The quantity and quality of the tech innovation bond market are both rising

“The launch of the ‘Technology Board’ in the bond market has provided an efficient platform to guide market funds into the field of technological innovation,” Yue Zhigang told Jingwei.

Data from China Chengxin International show that in 2025, the issuance of tech innovation bonds exceeded 2.28 trillion yuan, and from 2026 to date, it has doubled year-on-year.

Yue Zhigang pointed out that over the past year, the tech innovation bond market has experienced positive changes across multiple dimensions. In terms of issuers, the dominant position of state-owned enterprises remains stable, but the participation of private enterprises has significantly increased. Since 2026, the number of private tech innovation bond issuances has risen from 10.7% in 2025 to about 14.5%. Regarding bond maturities, over 60% of tech innovation bonds have maturities concentrated in 3-5 years, with some long-term varieties over 10 years emerging, better matching the long-cycle funding needs of technological innovation. The use of funds has also become more diversified, covering R&D, project construction, mergers and acquisitions, and restructuring.

Addressing concerns about whether “policy hotness and fast issuance” might lead to inflated ratings or underestimated risks, Yue Zhigang said: “The current structure of issuers and the distribution of credit ratings for tech innovation bonds show that credit ratings are generally aligned with risk levels.”

He cited data indicating that in 2025, the total proportion of bonds issued by central enterprises and state-owned enterprises was about 85%, with credit ratings mainly above AA+. Among them, bonds issued by AAA-rated entities accounted for about 75%. Regarding risk premiums, taking 2025’s medium-term notes as an example, the average issuance cost for AAA and AA+ rated tech innovation bonds was 2.27%, slightly lower than traditional industry bonds of the same grade; for entities rated AA and below, the issuance cost was 2.74%, 11 basis points higher than traditional industry bonds of the same grade. This partly reflects that high-grade tech innovation bonds are more accepted in the market with lower risk premiums, giving them some financing cost advantages; whereas bonds rated AA and below have slightly higher issuance costs, indicating investors consider more credit risk factors and require moderate risk premiums for these bonds.

He pointed out that the credit risk of tech innovation companies has stage-specific characteristics, especially for startups and growth-stage companies, which tend to have high risks and high growth potential, with industry risks also varying. When serving the financing needs of tech innovation companies, credit rating agencies need to balance supporting enterprises and revealing risks—improving value discovery and risk pricing through professional capabilities, while maintaining risk bottom lines and safeguarding rating credibility.

“Rating agencies must adhere to principles of independence, objectivity, fairness, and prudence, strengthen internal governance and compliance management, establish mechanisms such as firewall isolation and conflict of interest prevention, avoid external interference from market competition or business cooperation, and ensure ratings reflect the fundamentals of the companies. This provides a credible pricing benchmark and risk signals for the tech innovation bond market, safeguarding the agency’s reputation and industry credibility,” Yue Zhigang said.

“AI results are never directly used as final rating conclusions”

Accurately rating high-growth, high-risk tech innovation companies is a common challenge faced by the industry. Traditional rating frameworks focus heavily on assets and revenue size, which are less applicable to asset-light, R&D-intensive tech firms.

“We have introduced a core indicator system called ‘Technological Innovation Competitiveness,’ breaking away from traditional rating ideas,” Yue Zhigang explained. This system shifts the evaluation focus toward the company’s technological value, growth potential, and liquidity risk. Specifically, dimensions such as R&D strength (R&D investment and personnel ratio), technological成果转化情况 (patent count and new product launches), are incorporated into quantitative assessments.

Amid the wave of fintech, China Chengxin International is deeply integrating AI, big data, and other technologies into credit evaluation. Regarding risks like AI hallucination and overfitting, Yue Zhigang said: “We require AI models to strictly rely on authoritative company databases and design them with the principle of ‘no answer if uncertain.’” He emphasized that all AI-assisted materials must be reviewed and approved by analysts one by one, and AI results are never directly used as final rating conclusions.

To address overfitting, he noted that overfitted models perform well on historical data but tend to fail in new entities, new cycles, or “black swan” events. Therefore, China Chengxin International strictly controls data quality, cleanses abnormal data, constructs training samples across cycles and industries, and has rating experts select features with economic significance. The models’ performance is regularly monitored and optimized.

“Establishing a sound risk mitigation and sharing mechanism”

Looking ahead to the “14th Five-Year Plan,” Yue Zhigang proposed several mid- to long-term suggestions for supporting technological innovation through bonds.

First, strengthen product innovation and expand the scope of bond market services. Continue to deepen the construction of the “Technology Board,” build a tiered issuance system, consider moderately lowering issuance thresholds for small and medium-sized tech enterprises, and allow more companies to access bond financing, expanding service scope. Accelerate the development of bond varieties linked to equity markets, enhancing financing flexibility. Additionally, further encourage market institutions to issue medium- and long-term tech bonds to better match investment cycles and financing needs.

Second, cultivate a diverse professional investor base and optimize the investment ecosystem. Promote various funds entering the market, especially encouraging commercial banks, social security funds, and other long-term capital to increase investments in tech innovation bonds, fostering long-term and patient capital. Use window guidance, tax incentives, and moderate subsidies to incentivize investment in tech bonds. Continue developing tech innovation bond ETFs and index products, supporting the creation and tracking of diversified indices, enriching the supply of tech bond investment products.

Third, improve risk mitigation and sharing mechanisms, and strengthen risk identification and management. On one hand, optimize risk mitigation and sharing through a “government-guided + market operation” model, promoting government-backed financing guarantee agencies to increase guarantees, and allowing commercial guarantee institutions to play a supplementary role, establishing a multi-layered guarantee system to share default risks. On the other hand, the government should increase support for guarantee agencies, providing risk compensation for tech bond guarantees to boost their willingness to provide guarantees. Encourage issuers to utilize credit risk mitigation tools such as CDS and CRMW alongside bond financing, leveraging risk-sharing mechanisms. As the market expands, more refined risk identification and management are needed, focusing on the issuer’s R&D transformation ability, technological progress, and credit quality changes to improve risk prediction, response, and disposal, while clarifying responsibilities among relevant parties after risk exposure.

Fourth, optimize supporting service mechanisms and create a favorable market environment. Strengthen information disclosure, simplify non-core disclosures, and allow flexibility for core confidential information, exploring targeted disclosure mechanisms under confidentiality commitments, and improving full-chain bond information disclosure management. Encourage market institutions to optimize valuation models, incorporate more tech innovation factors, and improve valuation scientificity and flexibility. Regulate third-party valuation agencies, promote the use of multiple valuation products, and enhance cross-checking of valuation results.

(More reporting tips, please contact the author Luo Kun: luokun@chinanews.com.cn) (China News Service Jingwei APP)

(The viewpoints in this article are for reference only and do not constitute investment advice. Investment involves risks; please proceed with caution.)

Editor: Xue Yufei, Li Zhongyuan