Chinese banks, telecom carriers and a district government have begun packaging AI tokens as consumer and business products. Moonshot's Kimi and Agricultural Bank of China launched an AI-native credit card; China Telecom sells tiered token plans starting at 9.9 yuan/month for 10 million tokens, bundling its Xingchen model, DeepSeek V3.2, and GLM5 for developers. Guangzhou's Haizhu district introduced a Token Loan program that uses token consumption as a credit-review signal, plus subsidies up to 2 million yuan/year toward token bills. Separately, research across 25 European languages shows tokeniser fertility varies from 1.23 tokens per word in English to about 3.1 in Greek and Maltese, meaning a fixed token bundle delivers roughly 2.5x less usable service in some languages than others.
Questions this post answers
How much does China Telecom's cheapest AI token plan cost and what do you get for it?
China Telecom's trial token packages, started 17 May, offer an entry consumer tier reportedly priced at 9.9 yuan a month for 10 million tokens. Individual customers get access to the carrier's own Xingchen model plus DeepSeek V3.2, while developer tiers also include GLM5. The carrier is also building a Tianyi Token system letting loyalty points be redeemed for token packages via a platform called TokenHub. Developers weighing regional AI pricing plans can follow token-cost shifts like this on daily.dev.
Why does the same number of AI tokens buy less text in some languages than others?
Because tokeniser fertility differs sharply by language: research across 25 European languages found it ranges from 1.23 tokens per word in English to about 3.1 in Greek and Maltese. Romance languages fall between 1.5 and 1.7, Germanic between 1.7 and 1.9, Slavic between 2.2 and 2.5, and Uralic/Baltic languages between 2.7 and 3.0, meaning a fixed token bundle delivers roughly 2.5 times less usable output in Athens than in Dublin. Teams pricing or budgeting multilingual LLM usage can track tokeniser fertility research via daily.dev.
What is the Haizhu district Token Loan program in Guangzhou?
It is a lending program launched in August by Guangzhou's Haizhu district that uses AI token consumption, platform qualifications, and payment-collection progress as credit-review criteria instead of traditional collateral like plant and equipment. Bank of China's Guangzhou branch can set a credit limit based on a contract or a company's token use, and the district also offers subsidies up to 2 million yuan a year toward companies' token bills. Founders and investors tracking novel AI-financing models can follow developments like this on daily.dev.