In the rapidly evolving landscape of artificial intelligence, China’s technology sector is experiencing a dramatic shift in how companies approach AI token consumption. What began as a badge of honor for employees has now become a strategic challenge for management as token usage explodes.
The token the fundamental unit of computing power required for AI processing, has become a critical metric in the tech industry. Initially, high token counts were seen as indicators of productivity and initiative. However, the unprecedented growth in token consumption has forced companies to reevaluate their strategies.
AI tokens become ubiquitous in China
By 2026, AI tokens have permeated everyday life in China, transforming from a niche technical concept to a ubiquitous currency. They are now offered as incentives with purchases, included in credit card rewards, and sold in monthly packages similar to mobile data plans. This widespread availability has made tokens accessible to the general public, often without them actively seeking them out.
To understand this phenomenon, it’s essential to grasp what a token represents. In the context of AI, a token is the smallest unit of text that a model processes to generate a response. It serves as both a measurement unit for consumption and the functional currency of AI services. The more tokens consumed, the higher the cost, making token giveaways an effective marketing strategy.
The rapid growth of token consumption
The surge in token usage in China has been nothing short of extraordinary. At the start of, the National Data Administration reported daily consumption of 100 billion tokens. By June of the same year, this figure had skyrocketed to 500,000 billion tokens per day. While much of this traffic originates from businesses and advanced users, popular adoption is accelerating at an impressive pace.
The key factor enabling this proliferation is the cost advantage of Chinese open-source models. These models are significantly cheaper to operate, with cost margins estimated to be 60-90% lower than those of OpenAI or Anthropic. This cost efficiency makes it feasible for companies to offer tokens as promotions, associating their brands with cutting-edge technology.
Innovative token distribution strategies
Chinese companies have developed creative ways to distribute tokens to consumers. Moonshot AI, for instance, has partnered with the Agricultural Bank of China and American Express to offer tokens as rewards for credit card purchases. Merchants Bank has launched a developer-focused card that provides up to 1.8 billion tokens for models like MiniMax, while Shanghai Pudong Development Bank offers up to 3 billion tokens for Qwen models.
The hospitality industry has also embraced this trend. In Beijing, the AGI Bar offers unlimited access to DeepSeek V4 Flash with beverage purchases. Even restaurants are getting involved, with some offering token vouchers to diners, encouraging them to ‘feed’ their AI agents.
As with any valuable commodity, a parallel market has emerged. On Xianyu, China’s second-hand platform, individuals resell token packages obtained through promotions or giveaways. Some users even share AI accounts to distribute the cost among multiple parties.
Telecom operators enter the token economy
China’s major telecom operators have integrated tokens into their service offerings, treating them similarly to data traffic. China Telecom provides monthly plans starting at 10 million tokens for approximately 1.16 euros, while China Unicom offers 6 million tokens for about 2 euros. China Mobile goes even further, with recharge options as low as 400,000 tokens for 13 cents.
The financial sector has also adapted to this new reality. Banks in Guangzhou’s Haizhu district have introduced ‘token loans’ for AI startups, using token consumption and production data to determine creditworthiness. This innovative approach reflects the growing importance of tokens in China’s economic landscape.
AI gateways: managing the token explosion
As AI adoption grows, telecom operators face new economic challenges. While model prices are decreasing, This paradox has led to the development of AI gateways which act as intermediaries between applications and AI models.
AI gateways provide centralized control over AI traffic, enabling operators to route requests to the most appropriate models based on price, performance, quality, and latency. This approach allows for more efficient resource allocation and cost management. AT&T, for example, has implemented a gateway system that has reportedly reduced AI spending by up to 90% by optimizing token usage.
Chinese operators are taking this concept further by incorporating AI gateways into their commercial offerings. China Mobile’s Mixture of Models and Agents (MoMA) platform aggregates computational resources and services, allowing clients to access different models through a unified interface. This service has reportedly reduced token costs by about 30% for participating businesses.
China Telecom has developed TokenHub, a distributed infrastructure for AI inference that combines proprietary and third-party models. They offer tiered access plans, with the highest tier providing 680 million tokens per month for approximately 699 euros. This approach transforms tokens into a measurable and marketable unit of AI access.
The evolution of AI gateways presents both challenges and opportunities for telecom operators. While the technology itself may become commoditized, operators can differentiate themselves by integrating it with their existing assets, such as connectivity, cloud services, and cybersecurity. This integration could potentially transform AI gateways into a valuable enterprise service offering.
