Traditional businesses of telecom operators are hitting a ceiling, while emerging businesses fail to adequately fill the gap, leaving them trapped in the dilemma of "volume growth without revenue growth." However, the advent of the AI (Artificial Intelligence) Token era seems to provide them with a crucial leverage.
As the core business shifts from "traffic operation" to "token operation," can telecom operators leverage this to regain their past glory?
Telecom Operators March Towards "Token Operation"
At the 2025 annual performance briefings of the three major telecom operators, "token operation" was repeatedly mentioned. So, why has Token become a development focus for these operators? Experts analyze that there are two main reasons.
First, the imperative of their own transformation. Looking at the 2025 financial reports, the revenue and net profit growth rates of the three major telecom operators have continued to slow down, with sluggish growth in traditional businesses and limited volume expansion in emerging businesses. The emergence of Tokens provides a new path for operators to transition towards computing power and data value operation.
Second, the explosion in market demand. In March this year, the scale of Token invocations exceeded 140 trillion tokens, a more than thousand-fold increase over the two years since early 2024. Meanwhile, the invocation cost of mainstream AI models has dropped from $20 per million Tokens in November 2022 to $0.07 in October 2024. This surge in volume and drop in cost are pushing operators into the new track of AI infrastructure and inference services.
Under the confluence of multiple factors, the three major telecom operators have sent out a common signal—shifting from traditional "traffic operation" to "token operation."
Ke Ruiwen, Chairman of China Telecom, pointed out that China Telecom will take Token services as its core business line. Centering on the goal of "building a leading AI service provider," it will integrate elements such as technology, talent, and channels to provide integrated, digitalized, intelligent, and localized AI services, creating a new form of intelligent economy. It is reported that China Telecom will build around the integrated intelligent cloud system of "computing power, platform, data, model, and application," strengthening its own Tokens, expanding ecosystem Tokens, and actively exploring international Token operation. It will provide differentiated services and pricing for general, professional, and customized Tokens.
Chen Zhongyue, Chairman of China Mobile, pointed out that Token is the most critical the most critical factors of production in the communications industry in the AI era, and China Unicom is about to enter a new stage of dual high-speed growth in Token usage and BAT (Baidu, Alibaba, Tencent) traffic. Management of China Mobile stated that the Token market is rapidly opening up. In the future, it will strengthen Token operation, promote the integration of high-quality models by Mobile Cloud to build trusted inference services, and connect the service chain of "Agents using Tokens, and Tokens driving computing power."
Although China Unicom has not explicitly mentioned Token operation, its AI layout also points to the business closed-loop of "computing power + model + intelligent agent." Currently, China Unicom has accumulated over 400TB of high-quality datasets, provides more than 140 mainstream models, and will launch 200 AI applications and intelligent agents in the future, serving over 400,000 customers in government affairs, healthcare, finance, and other sectors.
Can the Three Major Paths Outrun Cloud Vendors?
For telecom operators, the current priority is to grasp both aspects: first, consolidating the foundation of computing power and continuously expanding investment; second, building a sustainable profit model.
In terms of computing power, the three major telecom operators have entered the stage of large-scale deployment, making full preparations for the Token transformation. The total scale of self-owned and accessed intelligent computing power of China Telecom has reached 91 EFLOPS, the mutual access latency among the eight major hubs has been reduced to within 12 milliseconds, and the rack power scale of AIDC (AI Data Center) exceeds 3.2GW. The total intelligent computing power scale of China Mobile has reached 92.5 EFLOPS, achieving full-specification computing capabilities from hundreds to over ten thousand cards, perfecting the three-level computing power latency circle of 1 millisecond in metropolitan areas, 5 milliseconds in provincial areas, and 20 milliseconds nationwide, and providing external services with over 1.5 million standard IDC (Internet Data Center) racks. China Unicom has a standard rack scale of over 1.1 million, has built seven 100-megawatt-level AIDC parks, and its intelligent computing scale has reached 45 EFLOPS.
Meanwhile, in 2026, the capital expenditures of the three major telecom operators are clearly leaning towards computing power: China Mobile's capital expenditure plan is 136.6 billion RMB, of which investments in computing power networks and intelligent networks will increase by 62.4% and 19.8% respectively; China Telecom's investment plan is 73 billion RMB, with 25.5 billion RMB for computing power infrastructure investment, a year-on-year increase of 26%. China Unicom's capital expenditure is about 50 billion RMB, and the proportion of computing power investment will exceed 35%.
In terms of the profit model, China Comservice Consulting & Design Institute Co., Ltd. pointed out that Token operation is an upgraded version of the transition from past IaaS (Infrastructure as a Service) operation to Token PaaS (Platform as a Service) operation. Telecom operators have three major profit paths: direct Token sales, ecosystem cooperation revenue sharing, and using Tokens as settlement tools.
First, direct Token sales. Targeting individuals, SMEs (Small and Medium Enterprises), and government and enterprise customers, it will launch pre-sale services such as Token packages, office packages, and private deployment, billing based on Token volume.
Second, ecosystem cooperation revenue sharing, integrating third-party AI models and applications, and extracting a share of Token invocation revenues.
Third, using Tokens as small-value settlement units for data trading and data exchange, such as settlement tools for digital employees.
But can telecom operators seize the AI window period just by holding substantial computing power resources? In the traffic era, operators missed out on social tools and were reduced to dumb pipes by internet vendors. In the AI era, while operators advocate Token operation, cloud vendors have already used real money to seize AI entry points.
Token operation is essentially still a "pay-as-you-go" model. If operators merely replace "traffic" with "Tokens" without mastering the entry points for large models and intelligent agents to interact with users, history is likely to repeat itself: changing from "traffic pipes" to "Token pipes." The name has changed, but the destiny remains the same.
Of course, from another perspective, the competition in the AI era is far from settled. Holding massive resources such as machine rooms and transmission networks, operators still need to truly break away from old thinking in the future and outrun cloud vendors in terms of ecosystem and innovation.
Managing Editor/Layout: Zhu Wenfeng
Proofreader: Mei Yaxin
Supervisor: Liu Qicheng