Summary of Key Points
Recently, Guangdong introduced the first “Token Loan” financial product in China, with institutions such as the Bank of China and CITIC Bank also following suit and launching similar products. In simple terms, AI companies can now use their Token consumption as one of the indicators for loan approval. This represents a new approach by banks to address the challenges faced by small and medium-sized AI companies, which often struggle to obtain loans due to their “light assets” and lack of collateral. However, experts point out that Token consumption is only a supplementary indicator and not the sole basis for lending decisions. There are still gaps in the process from Token consumption to the actual ability of companies to repay their debts. Additionally, using Tokens to assess credit carries risks such as poor comparability and potential distortion of value. To make Tokens a reliable credit indicator, it is necessary to improve systems for stratified calibration and third-party verification.
Detailed Explanation
1. What is Token Loan, and why do banks focus on Token consumption?
Tokens can be considered the “basic units” of information processing in AI. For example, processing a Chinese character in AI may consume approximately 0.6 Tokens, while processing an English character consumes even fewer. Token consumption occurs whenever AI companies train models or users utilize AI tools (such as ChatGPT). Banks have found it difficult to provide loans to small and medium-sized AI companies because these companies typically have “light assets” and lack tangible assets like factories, machinery, or real estate, which can be used as collateral. Traditional financial indicators (such as transaction volumes and orders) may not be reliable. Token consumption, on the other hand, is an essential metric for AI companies, as it directly reflects their operational activity (similar to how electricity consumption reflects a factory’s activity). Therefore, banks are using Token consumption as an additional criterion for lending, but the ultimate decision still depends on the company’s profitability and ability to repay.
2. What are the advantages of using Token consumption as an indicator compared to traditional data?
Compared to traditional financial indicators, Token data has three significant advantages:
- Timeliness: Token data is updated in real-time, allowing banks to quickly detect changes in a company’s operations (for example, a sudden increase in the frequency of users using AI tools).
- Objectivity: Token consumption represents actual resource usage, making it less susceptible to fraud compared to transaction volumes.
- Suitability for light-asset companies: Since these companies do not have collateral, Token consumption provides a basis for banks to assess their creditworthiness.
3. Challenges in transitioning from Token consumption to actual repayment ability
Although high Token consumption indicates activity, it does not necessarily mean the company can repay its debts. There are two key hurdles:
- Token consumption does not equal revenue: Many AI companies offer free or low-cost services to attract users, which can lead to higher Token consumption and greater losses.
- Revenue does not equal repayment ability: Even if Token consumption is converted into revenue, the quality of that revenue (e.g., from long-term government contracts or low-profit activities) must be considered to determine the company’s ability to repay.
4. Risks associated with using Tokens as a credit indicator
Using Tokens as a credit indicator poses several challenges:
- Poor comparability: The meaning of Tokens varies significantly across different AI companies. For example, companies in the training phase may consume more Tokens due to higher R&D costs, while mature companies with stable operations may consume fewer Tokens.
- Value distortion: The cost of processing data has been decreasing, which can affect the accuracy of Token values. For instance, processing 1000 characters may now cost only 0.50 yuan, potentially leading to misleading comparisons between companies with different business volumes.
- Limited perspective: Token consumption only reflects resource usage and does not account for other critical factors, such as the loss of key customers or regulatory penalties, which can impact a company’s repayment ability.
5. Needed improvements to make Token credit system mature
To make Tokens a reliable credit indicator, experts suggest addressing the following issues:
- Stratified calibration: Differentiate Token value standards for different types of AI companies (e.g., those in the training phase versus those with mature applications, or those serving B2B versus B2C markets).
- Third-party verification: Ensure that Token data is verified by neutral organizations to protect both banks’ interests and companies’ privacy.
- Comprehensive evaluation: Combine Token data with traditional financial indicators (such as cash flow, gross profit margins, and customer composition) for a more accurate assessment.
Conclusion
Token Loan represents a new approach for banks to lend to AI companies, addressing the financing challenges of light-asset companies. However, the system is still in its pilot phase and requires further development to become truly effective.