While telecommunications operators around the globe remain remarkably cautious about plunging headfirst into artificial intelligence tokens, major Chinese carriers are charging forward with an aggressive new strategic direction. China Telecom shattered industry precedent back in May by becoming the world’s very first telecommunications operator to commercially sell AI tokens, with its two primary domestic rivals quickly following suit.
In the months since that historic rollout, China Telecom has signaled that it intends to place tokens at the absolute center of its entire corporate business model. The company’s leadership firmly believes that these units of computational and linguistic work will inevitably become the next great economic driver for the telecommunications sector as a whole, shifting the fundamental nature of telecommunications infrastructure and revenue generation.
This sweeping strategic pivot is already yielding massive engagement figures across enterprise and consumer lines. Since making its official debut in July, China Telecom’s proprietary TeleAgent office assistant has rapidly amassed an impressive base of 1.2 million active users. More remarkably, daily token consumption across the platform has skyrocketed to top 200 billion, illustrating a voracious appetite for enterprise-grade generative intelligence tools integrated directly into traditional telecom pipelines.
The sheer scale of this transition highlights a profound evolution in how network operators view their core utility. Speaking at a prominent industry event, Li Xuelong, head of China Telecom’s Institute of AI (TeleAI), articulated the paradigm shift currently sweeping through the nation’s digital infrastructure sector.
"The industry is currently undergoing a transformation from traditional bit traffic management to token value management," Li told attendees, emphasizing that the days of merely moving raw data packets are rapidly giving way to a new commercial era built on managing, routing, and monetizing intelligent computational output.
To realize this vision, major telecom operators are actively constructing an end-to-end operational loop designed to support the seamless flow of AI tokens. According to Li, this comprehensive infrastructure encompasses every critical stage of the AI lifecycle, ranging from upstream model production and dynamic resource scheduling down to final enterprise settlement.
Demonstrating this capability at scale, China Telecom has successfully aggregated 170 distinct large language models onto its centralized token management hub. This massive aggregation layer is specifically engineered to enable large-scale token production and distribution. Furthermore, the company has leveraged these foundational assets to build 420 industry-specific vertical agents tailored explicitly for complex sectors such as government administration, public transport, and advanced manufacturing.
Rising Usage Rates and Architectural Breakthroughs
A persistent engineering hurdle in deploying generative AI at scale has historically been the friction between heavy cloud-based token platforms and end-user hardware devices. However, China Telecom reports that it has successfully cleared a major technological bottleneck in bridging this divide.
Li noted that the operator’s proprietary solution relies on an advanced end-edge-cloud architecture. This integrated framework creates a closed-loop system capable of facilitating intelligent transmission and localized applications simultaneously, ensuring that high volumes of tokens can be processed and delivered without introducing crippling latency or overwhelming network bottlenecks.
This aggressive push into artificial intelligence is not merely a corporate-driven venture; it is deeply intertwined with broader national industrial policy. China Telecom, alongside its major rivals China Mobile and China Unicom, has openly acknowledged that it is executing according to a centralized national AI strategy. This strategic roadmap places telecom operators squarely at the physical and digital center of the country’s burgeoning AI data center ecosystems and massive cloud computing expansions, while simultaneously tasking them with leading the charge in bringing commercialized AI services directly to the mass market.
For these carriers, immediate profitability is not necessarily the primary short-term target. Instead, they are playing a long-game strategy, heavily banking on substantial long-term returns as artificial intelligence becomes deeply embedded in the daily operations of businesses and consumers alike.
The Financial Reality and the Cost of Scale
Despite the immense enthusiasm surrounding token-based business models in China, international industry observers are urging caution regarding the underlying economics of scaling AI consumption. A comprehensive research note published by Bain & Co. highlights significant financial downsides and operational risks that tokens present to telecommunications operators globally.
The research points to troubling figures indicating that customer usage rates are currently rising at a much faster pace than the rate at which underlying AI production costs are falling. The fundamental dilemma facing modern telecom executives, according to the report, is figuring out how to scale up artificial intelligence services "without simply adding token costs to an already-heavy legacy operating cost base."
While the headline price of individual foundational models has historically declined by roughly tenfold each year, the economic reality on the ground paints a much more complicated picture for enterprise deployers. The research note emphasizes that the effective cost per completed task frequently remains stubbornly flat, while the total corporate token bill balloons in an unpredictable fashion as user adoption scales upward.
Even in markets like China, where domestic large language models may offer lower baseline operational costs, the AI model itself remains only a single component of a much larger, multifaceted expense ledger. To navigate these treacherous financial waters, Bain urges telecom operators to closely measure the granular cost per task for both standard AI implementations and advanced agentic AI workflows, while simultaneously establishing a dedicated, ring-fenced AI compute budget to prevent runaway expenditures from eroding traditional telecommunications margins.
