Summary of Key Points
In August 2026, Haizhu District in Guangzhou introduced the "Token Loan" initiative, which uses the amount of Tokens consumed by AI companies (the smallest unit for processing text using AI) as a core criterion for granting bank loans. This approach addresses the financing challenges faced by AI firms with limited assets. It represents a fundamental shift in banking risk assessment from focusing on fixed assets to evaluating business activity. Many cities across the country, including Beijing, Shanghai, Anhui, and Chengdu, have followed suit. However, there are also risks associated with this model, such as the possibility that high Token consumption does not necessarily equate to profitability, data fraud, and insufficient bank capabilities.
I. Token Loans: The New Form of Financing for AI Companies
Traditional bank loans rely on tangible assets like factories and equipment as collateral, but AI companies are typically characterized by their lack of physical assets, with only a few servers and funds invested in computing power and research and development. Banks struggle to understand the value behind this "money-consuming" activity and are therefore hesitant to provide loans.
Token Loans aim to change this situation:
- New criteria for lending: Instead of focusing on fixed assets, banks consider four key factors: Token consumption (indicating business activity), the value of computing power services, accounts receivable, and the amount of Token commissions earned.
- Coverage of the entire industry chain: There are three types of Token Loans available: for providers of computing power, users of computing power, and companies involved in Token distribution. Businesses of all sizes can apply.
- Flexibility and inclusiveness: Loan amounts range up to 30 million yuan with a term of 3 years. Guarantees can include credit or the pledge of accounts receivable; even new companies with orders or guarantees can obtain loans.
For example, the Haizhu Branch of the Bank of China has already provided over 400 million yuan in loans to AI companies, with a trial loan amount of 28 million yuan. Even startup teams from large corporations (which consume more Tokens) find it easier to get funding.
II. Why Can Tokens Be Used as a Basis for Loans?
Tokens are akin to the "electricity consumption" in the AI industry: high Token usage indicates that business operations are ongoing, as customers are using models and services are being delivered.
Haizhu District is able to implement this approach because it has a robust AI ecosystem with over 8,000 companies in various fields, ensuring that Token consumption is based on actual business activities. Additionally, supportive policies have been established:
- Government incentives: The district offers subsidies for R&D (1.5 million yuan per year), application integration projects (5 million yuan), Token consumption (2 million yuan), and loan interest discounts (2 million yuan). This combination of government support, bank funding, and financial assistance forms a solid foundation.
The National Data Administration has recognized Tokens as a "value anchor in the intelligent era," providing official recognition for this metric.
III. National Competition: Different Approaches to Token Loans
Token Loans are not unique to Guangzhou; cities across the country are competing to develop their own models:
- Guangzhou Haizhu: Combines financial products with industry policies to target AI companies' Token consumption and financing needs.
- Beijing Economic Development Zone: Offers incentives such as "Eight Measures for Word Elements" (50% funding for R&D, 5 million yuan for application integration projects, and 2 million yuan for Token consumption subsidies), with a goal of reaching an intelligent economy worth 400 billion yuan by 2030, focusing on infrastructure development.
- Shanghai: Turns data assets into tradable tokens, allowing them to be used as collateral (e.g., the Beijing Bank provided a 20 million yuan loan to Shudian Technology).
- Anhui: Encourages Token Loans and Model Loans, offering up to 500,000 yuan in personal guarantees for startups with interest discounts.
- Chengdu: Offers pure credit-based computing power loans, using "computing power tokens" as a form of credit support, with instant approvals (first loan amount: 1.14 million yuan).
The common goal is to help banks better understand the AI industry by shifting their focus from hardware (computing power loans) to business activities (Token Loans).
IV. Risk Warning: High Token Consumption Does Not Necessarily Mean Profitability
While the logic behind Token Loans seems sound, there are several potential issues:
1. Tokens as a measure of activity, not profitability: A company that consumes 1 billion Tokens per day may be experiencing rapid growth or inefficient model usage (repeated calls), or it could be using scripts to inflate Token numbers for fraudulent loans.
2. Systemic risks: If banks create a cycle where increased Token consumption leads to more loans, which in turn drives further investment in computing power and more Token consumption, the industry's downturn (e.g., reduced customer demand or model prices) could cause a sudden drop in Token usage, leading to financial distress for companies.
3. Insufficient bank expertise: Traditional bankers are accustomed to reviewing financial reports and collateral; they need to understand advanced AI technologies and the commercial significance of Tokens. Using a one-size-fits-all approach may lead to misjudgments.
4. Lack of data verification: There are no established standards for auditing Token usage, raising concerns about potential fraud if companies overreport their consumption.
Conclusion
Innovation is welcome, but we must not mistake new methods for perfect solutions. Token Loans have addressed the issue of AI companies' difficulty in obtaining loans, but they do not solve the problem of whether these companies can actually generate profits. High Token consumption merely indicates the use of AI; profitability depends on business models, customer willingness to pay, and cost control.
As cities race to introduce Token Loans, the key is to ensure proper governance to prevent potential issues. Improving the verification of Token data and integrating Token metrics with traditional risk assessment methods (such as orders, payments, and profits) is essential for long-term success. While Token Loans represent a step forward, they are not the ultimate solution.
(The entire text is explained in plain language to make financial concepts accessible to a general audience.)