The operator's word package has been launched for several months, and market feedback is limited. It will take time to popularize it on a large scale.

It has been several months since the operator's Token package was launched. What is the market feedback? In May this year, China Mobile, China Unicom, and China Telecom launched word package packages, which packaged artificial intelligence (AI) computing power into standardized products that can be ordered and billed, which immediately attracted widespread attention in the market. However, despite the popularity, the number of users who have truly formed the habit of paying continuously is still limited. Most consumers are still in the stage of trying out the experience or waiting to see how it works. It will take some time for operators' Ci Yuan packages to be popularized on a large scale.

In the long run, operators are likely to become important promoters of expanding the word economy. With the improvement of industry standards and the expansion of application scenarios, Ciyu is expected to move from the developer group to mass consumers.

Seize the new entrance to AI

In 2026, Ciyuan economy is moving from a technical concept to an industrial reality. It is not accidental for operators to launch word package, but an inevitable choice due to the superposition of multiple factors such as changes in the industry environment, pressure to monetize computing power investment, and the need for strategic transformation. The operator's entry into the market also marks that it is accelerating its transformation from the traditional "selling traffic and communication services" to "selling computing power and intelligent services".

Ruan Jing, a professor and doctoral supervisor at the School of Statistics and Data Science at Capital University of Economics and Business, pointed out that on the surface, operators launched word package to replicate the glory of mobile data packages in the AI ​​era and get a share of the AI ​​dividend era. But in essence, this is a strategic battle for the super entrance in the AI ​​era.

Facing the wave of large models, operators are trying to rely on their huge sinking networks and massive user base to package words into products similar to traffic. Their core intention is to control the computing power distribution rights of AI services and seize the first entry point for ordinary users to contact AI.

Operators have structural advantages when deploying word packages - a nationwide computing network, a huge user base, a mature billing system, and carrier-grade security. Accordingly, operators are expected to reduce unit word costs through scale effects and effectively revitalize existing idle computing power. Taking China Mobile as an example, it has built a MoMA unified operation platform to achieve one-stop services for billing, settlement, and authentication. For example, Hubei Mobile launched two monthly packages of 40 yuan and 200 yuan, corresponding to 18,000 and 90,000 calls respectively. Beijing Mobile launched a 5.99 yuan package and a 24.99 yuan monthly package. The monthly package includes 10 million Token computing power quota.

Bai Runxuan, an analyst at CCID Consulting's Artificial Intelligence and Big Data Research Center, said that the three operators have different development paths. China Telecom promotes platform-based operations with unified group pricing and invests the most in computing power. China Mobile has access to multiple large models and is gradually rolling out through provincial pilots. China Unicom is more focused on small, medium and micro entrepreneurship scenarios. Overall, the operator's word business has moved from the conceptual stage to the stage of substantial resource investment.

There are still difficulties in landing

Transforming intangible computing power into standardized tradable commodities, whether the market really buys it, the actual use experience gives a realistic answer.

"I can use 10 million Tokens in half a day to do simple questions and answers and information retrieval." Mr. Jia, who is engaged in light technology services in Beijing, has been using API services for a long time and is very sensitive to the cost of word consumption. He said that a seemingly simple question-and-answer search may consume a large number of word elements in a short period of time, and some agents can consume tens of thousands of tokens in a single question. In order to control costs, he needs to subscribe to multiple monthly products at the same time and select the best services from different model manufacturers as needed.

A simple conversation consumes an unusually high amount of money. There is a significant gap in consumption between the same model on different platforms, which directly affects user experience and consumer trust. Ruan Jing pointed out that the opacity of billing does not mean that the text of the billing rules is not disclosed, but that the underlying generation mechanism of word elements is "black boxed". "The data processing mechanisms and tokenizer algorithms of different large models are different, and there is a lack of industry consensus on token generation. Directly applying this kind of product with large consumption fluctuations and opaque generation mechanism to a standardized billing model similar to mobile phone traffic is misaligned with the actual consumption pattern."

Li Wei, deputy director of the Cloud Computing and Digital Research Institute of the China Academy of Information and Communications Technology, believes that the current word package is still in the early stages of development. Although there is a unified product concept, it lacks unified, transparent, and horizontally comparable measurement rules. She suggested that the industry should clarify basic concepts, provide users with queryable consumption details, establish a reminder mechanism for the entire consumption process, and speed up the introduction of unified measurement and service specifications.

Experts interviewed generally judged that in the long run, Ci Yuan is expected to become a new basic resource for digital life. It will gradually shift from independent explicit payment to packaged and embedded, and eventually become an imperceptible underlying digital right.

"In the future, consumers may use word elements every day, but they may not necessarily need to directly perceive and calculate word elements." Li Wei said that as artificial intelligence evolves from question and answer tools to intelligent agents, AI will gradually be embedded in mobile phones, computers, cars and home terminals to continuously perform tasks such as information retrieval, real-time translation, content generation, and schedule management. The shift in AI usage from occasional calls to continuous online use will bring stable and normalized demand for word elements.

"Only developers or professionals need cost control. It is unnecessary for ordinary consumers to delve into the underlying consumption, and it goes against the original intention of product ease of use." Ruan Jing said, this is just like the early users of the mobile Internet who carefully calculated traffic by MB (megabytes). As the traffic scale increases, the public no longer dwells on single web page consumption, and only focuses on the overall quota of the package.

Diversified supply begins to appear

Judging from the market structure, word unit sales have gradually formed three models: the first type is the original application programming interface (API) of large model manufacturers, such as DeepSeek, Zhipu, Baidu Qianfan, Huoshan Engine, etc., with significant pricing differences; the second type is cloud vendor aggregation platforms, such as Tencent Cloud, Alibaba Cloud Bailian, Huoshan Ark, etc., which call multiple models through a unified interface; the third type is telecom operators, and the three major operators have launched their own word unit packages and accessed multiple mainstream models.

Experts interviewed believe that the entry of operators is not simply to seize the original market, but to promote the evolution of the industry from direct sales by manufacturers to a model of "direct sales plus distribution, single model plus multi-model aggregation".

Li Wei pointed out that operators have massive personal, government and enterprise customer resources and can transform APIs originally intended for professional developers into package products that are understandable and affordable for small and medium-sized enterprises and ordinary consumers. In the long run, through centralized procurement and multi-model intelligent scheduling, operators are expected to expand the market pie of the entire word industry, and at the same time launch industry competition around customer access, bargaining power, and profit margins.

However, the current industry structure has not been finalized and is still subject to practical constraints such as insufficient self-developed model capabilities, inconsistent pricing among provinces, offline channels that need to be expanded, and lack of industry standards.

Faced with the phased contradiction between the rapid expansion of the industry and the out-of-synchronization of the construction of a standardized system, industry governance has been launched, and pricing standards, hierarchical assessments, and differentiated pricing models are being stepped up. Experts suggest that the competent authorities should establish a mandatory information disclosure system and require platforms to fully disclose billing rules and consumption details. With the continuous improvement of supporting systems, the existing chaos in the industry will be gradually rectified, and the general direction of operators' deployment of word-based operations will not change. (China Economic Net reporter Li Fang)