As a fundamental unit for processing information in artificial intelligence (AI) large models, tokens are increasingly entering financial services scenarios, with some banks adopting them as one of the reference indicators for credit assessment. Recently, several banks have successively launched "Token Loan" products, opening new financing avenues for asset-light tech enterprises that lack traditional collateral.
How do "Token Loans" work? Which enterprises are they primarily targeting? And what new trends in bank credit assessment do they reflect?
Token consumption is now being incorporated into credit reference indicators. Recently, Wenzhou Jinku Network Technology Co., Ltd. secured a 200,000 yuan loan from Agricultural Bank of China, leveraging its historical token settlement data, computing power procurement contracts, downstream business orders, and independent intellectual property rights. The funds are earmarked for computing power procurement to address the enterprise's working capital needs.
"As our independently developed interactive film and television creation platform gains wider adoption, our monthly token call volume continues to grow. Computing power procurement requires advance payment, but downstream project collections involve billing cycles. This loan helped alleviate our fund mismatch pain point," said Ge Binbin, the company's head. In the wave of the digital economy, computing power has become a core production factor for AI enterprises, and token consumption is rising rapidly.
Data from the National Data Administration shows that China's average daily token call volume exceeded 140 trillion in March, more than 1,000 times the level of two years ago. Behind the robust demand for computing power, the financing challenges for AI enterprises are becoming increasingly prominent. Over the past two years, many banks have tailored financing products such as "Token Loans" for AI companies, incorporating token consumption, computing power service contracts, and business orders into credit indicators, breaking away from the traditional risk control reliance on physical collateral like factories and equipment.
Expanding from "assessing collateral and financial statements" to "assessing computing power and tokens," multiple banks have accelerated the rollout of "Token Loans": Agricultural Bank of China has refined its tech innovation evaluation system, launching dedicated "Token Loan" solutions in places like Shanghai and Zhejiang; Bank of China has introduced the "Computing Power Token Loan," incorporating enterprise token settlement and computing power service contract value into credit assessment; Bank of Jiangsu has launched "Computing Power Loans," integrating computing power efficiency, team structure, R&D investment, and intellectual property into its scoring system.
Notably, several surveyed banks stated that "the amount of tokens used does not directly determine the loan amount." Instead, they cross-validate data such as computing power contracts and accounts receivable from multiple dimensions, and determine the final credit line based on the enterprise's operations, financials, and credit status. Lou Feipeng, a researcher at Postal Savings Bank of China, noted that incorporating token consumption into credit reference indicators does not alter the fundamental logic of bank lending; rather, it adds more evaluation metrics for innovation capability on top of traditional indicators like business operations, representing an exploratory practice of tech finance extending into emerging industries.
Segmenting business scenarios to strengthen industry chain services. In mid-August, Guangdong Province released a specialized financial product, the "Token Loan," with Bank of China Guangzhou Branch simultaneously launching three sub-products: "Computing Power Token Supply Loan," "Computing Power Token Application Loan," and "Computing Power Token Service Loan," offering a maximum credit line of 30 million yuan per borrower.
"Launching targeted products for different business scenarios can enhance service capabilities for diverse innovation entities in computing power supply, application, and services," said a representative from Bank of China Guangzhou Haizhu Sub-branch. To date, the branch has extended over 28 million yuan in "Computing Power Token Loans," serving six small and micro enterprises across the computing power industry chain.
The supply, application, and services of computing power are becoming three key directions for the financial industry's support of tech enterprises. From upstream enterprises building and training large models, to midstream application entities using tokens for content innovation, to downstream service providers offering computing power operations and technical support, the segmentation of business scenarios requires finance to achieve precise services across the industry chain. Dong Ximiao, chief economist at Zhaolian, believes that token consumption across different scenarios serves as alternative evidence of an enterprise's operational status, reflecting customer activity, product market acceptance, and business sustainability to a certain extent. It can serve as an important reference for identifying the true business value and growth potential of tech enterprises along the industry chain.
A representative from Agricultural Bank of China's Corporate Banking Department stated that banks exploring the inclusion of new production factors like computing power into financial products and credit evaluation models, positioning for early-stage tech enterprises, and building full-chain, full-cycle tech financial service solutions, helps construct a financial service model more suited to the development of the computing power industry.
Government-bank collaboration is continuously optimizing risk control systems. In August, the Chengdu Municipal Economy and Information Technology Bureau and the Municipal New Economy Commission jointly developed the "Computing Power Loan" with Bank of Chengdu, a dedicated pure credit loan product using "computing power vouchers" fiscal funds as credit enhancement. The maximum credit line per enterprise is 5 million yuan, with the first disbursement of 1.14 million yuan successfully made. Earlier, the "Computing Power Loan" product launched by Zhangjiagang Rural Commercial Bank also adopted a "basic credit plus computing power voucher enhancement" model, linking local computing power support policies with government risk compensation mechanisms.
Zhou Zhen, associate professor at Tsinghua University's PBC School of Finance, noted that the linkage model of "computing power vouchers plus computing power loans" effectively amplifies the leveraging and guiding role of fiscal funds. Supporting measures such as risk compensation and interest subsidies can further reduce banks' trial-and-error costs. In the future, banks need to cultivate more composite talents who understand both risk control and AI business models to better promote deep integration of finance and industry. It should also be recognized that the risk control model for "Token Loans" is still being refined, and the industry lacks unified, auditable standards for token attestation. Therefore, multi-dimensional verification of data authenticity and accuracy is needed, along with continuous optimization of intelligent risk control systems and credit rules.
Lou Feipeng expressed confidence that as banks continue to unlock key channels for converting data into credit, financial "living water" will more precisely irrigate the digital industry, injecting strong momentum into the high-quality development of AI enterprises.