The Computing Power Network is here, and operators are awakening: Will TaaS be the next trillion-dollar business for operators?

The Computing Power Network is here, and operators are awakening: Will TaaS be the next trillion-dollar business for operators?

1. What happened? Computing Power Network strategy heats up, operators take center stage

1. Computing Power Network: From technical concept to national strategic infrastructure

Recently, the "Computing Power Network" has gained unprecedented market attention. The driving force behind this is not a single technological breakthrough, but a series of intensive national policy signals.

In 2026, the "15th Five-Year Plan" outlined "National Integrated Computing Power Network" as the very first of 109 major projects**, clearly stating the need to "build a multi-tier computing power infrastructure system and a national integrated computing network." Previously, the 2025 "15th Five-Year Plan" proposals included the computing power network for the first time as part of the modern infrastructure system, alongside information and communication networks and major scientific infrastructure.

What’s even more noteworthy is the elevated positioning. The State Council executive meeting listed the computing power network alongside water networks, new power grids, new-generation communication networks, urban underground pipeline networks, and logistics networks—collectively known as the “Six Networks.” This means the computing power network is now regarded as strategically important as water, electricity, and transportation—traditional “lifelines” of the nation.

The underlying logic of the computing power network is: in the AI era, computing power is becoming a fundamental resource like electricity. However, China currently faces a severe structural mismatch in computing resources—demand is high in the East but land and electricity costs are high, while the West is rich in energy but lacks applications for computing power. The goal of the computing power network is, just like the national power grid, to integrate scattered computing resources into a unified, on-demand-dispatched “computing power grid.”

The National Development and Reform Commission has disclosed that, by 2026, total investment in the “Six Networks” and related fields is expected to exceed 7 trillion RMB. This top-down policy push is the most direct reason for market attention to the computing power network concept.

2. Telecom operators: Irreplaceable core players in computing power network construction

The essence of the computing power network is the integration of “connectivity + computing.” Telecom operators have unrivaled natural advantages in the connectivity aspect.

Firstly, networks are the physical foundation for computing power scheduling. The three major operators hold the world’s largest fiber optic and mobile communication networks, with coverage breadth and depth unmatched by any cloud provider or third-party data center. The Ministry of Industry and Information Technology’s “Computing Power Interconnection Action Plan” and “Millisecond Computing” special action directly named telecom enterprises as core implementing units—not by chance, but because only operators’ backbone networks can meet the low latency and high reliability required for computing power scheduling.

Secondly, operators are already designated as main builders in the "East Data, West Computing" project. Since the comprehensive launch in 2022, the three operators have played key roles in constructing data centers in eight hubs and ten clusters. This project lays the geographical blueprint and resource foundation for the computing power network.



Thirdly, operators' "cloud-network integration" capabilities form differentiated barriers. Unlike pure cloud service providers, operators possess full-stack capabilities from underlying network and infrastructure to upper scheduling platforms. China Telecom’s “Xirang” platform can schedule nearly 91 EFLOPS of computing power, China Mobile’s “Compute-Net Brain” achieves daily token scheduling in the hundreds of trillions—these platforms are the embryonic operating systems for the computing power network itself.

Finally, their "national team" status is a substantial barrier in computing power network construction. As the computing power network is defined as national strategic infrastructure, it involves data sovereignty, network security, and energy coordination. When serving critical industries like government, finance, and energy, operators' central enterprise status and compliance capabilities are irreplaceable admission tickets compared to internet companies.

2. Why is it important? Strategic shift—from “selling traffic” to “selling Tokens”

1. From “Computing Power Network” to “Token Economy”: The natural extension of operators' business logic

The traditional business of the three telecom operators has reached its ceiling—user numbers are near saturation, traffic is rising, but revenues (ARPU) are falling.

Computing power network construction opens a new business door for operators. When computing power can be scheduled and traded like electricity, Token—the smallest unit of information processing for large models—naturally becomes the “pricing unit” and “transaction medium” for computing power.

The Token economy is exploding faster than expected. By early 2024, China’s daily Token invocation volume was 100 billion; by the end of 2025, it surged to 100 trillion; by March 2026, it broke through 140 trillion—a thousand-fold increase in two years.

(Chart: Weekly call volume growth trends for mainstream large models on OpenRouter)

This growth curve reinforces progress in building the computing power network: it supplies low-cost, wide-coverage computing power, lowering the threshold for Token usage; the explosion in Token demand, in turn, validates the urgency of building the network. Operators as the core builders and operators of the network naturally stand at the hub of this commercial closed loop.

Against this backdrop, the three operators in their March 2026 earnings briefings, for the first time, explicitly made Token operation a core strategy: China Telecom proposed “transforming operations from traffic to Token,” China Mobile emphasized “strengthening Token operations,” and China Unicom proposed building a new model for computing power operation as “intelligent + Token + AI cloud.” Meanwhile, despite a general reduction in total capital expenditure, operators’ Token-related capital spending in 2026 soared—China Mobile reached 37.8 billion RMB (+62%), China Telecom 25.5 billion RMB (+26%), China Unicom 17.5 billion RMB (+35%).

The computing power network is the “road,” Token is the “car,” and operators are both builders of the road and dispatch centers for the cars. Establishing this role is the core clue to understanding operators' ongoing strategic transformation.

2. Telecom Operator TaaS (Token-as-a-Service) already has real cases:

Recently, the three major operators have frequently launched Token pricing packages for both consumers and enterprises nationwide, successfully completing the practical leap from "technical indicator" to "consumer commodity."

  • Unified measurement: No longer selling abstract “computing cores/GPU hours,” but converting computing power into Token count—an understandable metric for users, code, and applications.
  • Simplified billing: Integrating complex AI interface payments into familiar phone bills and broadband packages.
  • Platform-based distribution: For example, China Telecom’s recently launched Xingchen TokenHub 1.0 operation platform, scheduling nationwide computing power (like the Xirang platform), packaging it into unified interfaces for user consumption.

3. Compared to direct API sales by major model vendors, what is the industry value and differentiation offered by operators?

Major model vendors are “Token producers,” with their API direct sales constituting pure DTC (direct-to-consumer) software retail. Telecom operators act as “new-type digital element operators” and “universal AI traffic hubs.” They do not replace model vendors, but serve as the super agent and central dispatcher for computing power, networks, data, and channels.

① For “users (especially SMEs and individual developers)”, operators provide:

  • Multi-model aggregation, removing “choice anxiety.” A user buys a Token package from Shanghai Telecom and can instantly access 30+ mainstream models with a single click, without needing to register, top-up, and bind credit cards with multiple vendors.
  • Ultra-low threshold, no tech burden: By binding Token to cloud computers and smart devices, users start up and use immediately, avoiding local deployment, environment setup, and pricey hardware purchases.
  • Convenient payment, clear public/private separation: Supports phone bill payments. Employees and individuals can reimburse via telephone bills, making settlement far easier than cumbersome approval processes for overseas or Internet cloud vendors.

② For “major model vendors”, operators provide:

Exceptional customer acquisition channels and “suburban network”: The three operators possess the largest paying user base in China (hundreds of millions of individual users and tens of millions of enterprise clients). By connecting models to operators' TokenHub, vendors can instantly reach vast markets unfamiliar with APIs.

Reducing production costs through “computing-electric synergy”: Model vendors who buy GPUs and run Tokens face high power and network costs. Operators use the national integrated computing network and platforms like Xirang to dynamically schedule computing power across regions—leveraging surplus green electricity in the West—to lower Token production costs to a minimum.

Nation-owned-level security and trusted data space guarantees: For government and enterprise clients, operators (like Tianyi Cloud, GuoYun) offer fully autonomous, controllable central-enterprise trusted data spaces, providing endorsement for vendors and solving pain points around privatized deployment and data privacy, driving usage of vertical models in key scenarios to the millions of Tokens.

3. What's next? Can operators “sell” Tokens well?

1. Structural advantages: Why can operators sell Tokens?

Operators joining TaaS have innate structural advantages:

① Network-computing integration and edge inference:

This is operators' most critical differentiator. Cloud vendors (AliCloud, Volcano Engine) focus on centralized supercomputing centers and struggle with nationwide low-latency coverage. Operators, with nationwide backbone networks and thousands of edge nodes, can push MaaS capabilities close to end users. For industrial IoT, vehicle-road collaboration, city governance needing nearby inference, operators' low latency and stability is unmatched by cloud vendors.

② Extreme cost control—computing-electric synergy:

Power comprises 60%-70% of Token production cost. Through the "East Data, West Computing" project, operators deploy massive clusters in low-power-cost western areas, using direct green electricity to lower power costs by 30%-50%. This “watts-for-bits” vertical integration gives operators theoretical industry-best unit cost for Tokens. China Telecom’s Tianyi Cloud leverages Xirang’s integrated scheduling to continually push Token costs down.

③ Natural trust barrier for government and enterprise markets:

For government, finance, healthcare, and energy sectors involving data sovereignty and compliance, operators as central enterprises enjoy credibility unmatched by Internet vendors. They can offer all-stack domestic solutions—“5G dedicated networks + edge intelligence + industry models”—meeting top-level security and innovation standards. This “national team” status is a practical admission barrier for Tokens in government/enterprise markets.

2. Structural disadvantages: Operators’ Achilles’ heel

Operators also have obvious weaknesses in market business:

① Weak model capabilities and algorithm accumulation:

Tokens are fundamentally outputs of model capabilities, not simply monetizing computing power. Operators, though strong in computing power, their in-house models lag first-tier competitors in performance, data quality, and user experience. In most scenarios, operators are “second landlords”—reselling Tokens bought from third-party models, which means additional “licensing fees” and limits pricing power.

② Internet product DNA and developer ecosystem lacking:

Operators are accustomed to contract sales of SIMs and broadband, lacking the “simplicity, rapid iteration, data-driven” gene of consumer internet products. Their AI app interfaces, interaction logic, and developer docs lag behind cloud vendors. Developers prefer calling APIs directly on platforms like AliCloud or Tencent Cloud, and operators' ecosystems are weaker in usability and community activity.

③ Organizational inertia—decision processes and talent structure:

As central enterprises, operators struggle with “weekly iteration” speeds in AI, unable to respond to market changes as quickly as startups. Top AI talent prefers tech companies or unicorns with higher pay limits and flexible incentives, limiting operators’ capacity to optimize Token production efficiency.

3. How to look ahead for operator TaaS business?

According to IDC, China’s total AI inference Token consumption will reach 39,000,000 trillion by 2030. Currently, wholesale API prices for mainstream models are around 1 RMB per 1 million Tokens. With benefits from Moore’s Law, ASIC chip optimizations, model distillation (miniaturization), and falling green power costs in the West, it’s forecast that by 2030, retail Token prices will deflate to about 1% of today’s—about 1 RMB per 100 million Tokens.

We present three scenarios to project potential TaaS gross margin for operators (assuming 10% gross margin, primarily with third-party models). Since TaaS is in early stages, these predictions should be treated with caution.

① Conservative scenario: Operators as foundational channel distributors and long-tail bundlers, with 15% market share—TaaS annual revenue of 585 billion RMB and gross profit of 58.5 billion RMB by 2030.

② Baseline scenario: Operators as backbone dispatch centers of national computing power network, with 30% share—TaaS annual revenue of 1170 billion RMB and gross profit of 117 billion RMB by 2030.

③ Optimistic scenario: “Computing-electric synergy” builds high barriers, operators dominate half of TaaS ecosystem, with 50% share—TaaS annual revenue of 1950 billion RMB and gross profit of 195 billion RMB by 2030.

Conclusion:

For telecom operators, moving to TaaS is not just “politically correct” aligned with the computing power network policy, but also the only strategic choice as their main business hits bottlenecks. Without moving up the value chain, operators will become mere “digital plumbers” in the AI era.

The core barriers of operator TaaS are “network-computing fusion + government/enterprise trust,” not model capability. This means operators should position themselves as “intelligent service integrators in the AI era,” aggregating quality AI capabilities society-wide through a “self-developed + agency” model supermarket, marketing “security, proximity, cost optimization” as differentiated selling points.

On the other hand, the Token economy’s commercial closed loop has taken shape, but profit models are diverging. The split—“Tokens with high capabilities make money, low-capability Tokens lose money”—is emerging. Operators should focus on segments where they hold comparative advantage—government/enterprise Token scenarios, computing power scheduling services, security/compliance premiums—rather than competing head-on with algorithm companies in the red ocean of generic models.

From a secondary market perspective, operators must still digest tax policy changes and the pain of traditional business decline in the short term, but in the medium term, new TaaS business will be key to whether operator valuations can shift from “utility” to “technology service.”

Risk Disclaimer and Exemption ClauseThe market carries risks; invest with caution. This article does not constitute personal investment advice nor does it consider individual users’ special investment goals, financial status, or needs. Users should decide whether any opinions, viewpoints, or conclusions herein are suitable for their specific situation. Investing based on this is at your own risk.