Recently, the interim reports of the three major telecom operators — China Mobile, China Unicom, and China Telecom — were released one after another.
It was a tale of two extremes.
On one hand, legacy businesses are retreating, with revenue from mobile voice, fixed-line telephony, and mobile messaging falling sharply.
On the other hand, new businesses are surging against the trend, with computing power service revenue growing at double-digit rates.
With traditional business growth hitting a ceiling, computing power has become the operators' new growth engine.
What advantages do operators have in selling computing power, and what challenges do they face?
According to the latest data from the Ministry of Industry and Information Technology on the communications industry's performance in the first half of 2026, China's telecom business revenue totaled 887.3 billion yuan in the period, down 2.1% year on year.
The growth rate turned negative, hitting a historic low for telecom revenue growth in recent years.
By segment, mobile voice call duration fell 6.2% year over year, fixed-line outgoing call duration dropped 20.4%, mobile SMS volume declined 8%, and SMS-related revenue plunged 12.9%.
Cumulative growth rates for telecom service revenue and total telecom business volume.
Voice and messaging services, once the mainstay of carrier revenue, have now passed their prime.
This is consistent with our own consumption experience.
Most people now rarely send SMS, make as few mobile calls as possible, and instead use WeChat for messages or voice calls.
It is hardly surprising that carriers' voice and SMS services are declining.
The question is, shouldn't data traffic be growing?
Indeed, in the first half of this year, cumulative mobile internet traffic reached 219.7 billion GB, up 17.7% year over year.
Curiously, data-related revenue at China's three major carriers is declining.
China Mobile's communication service revenue, which includes data, fell 5.7% from a year earlier to 350.4 billion yuan.
China Telecom's mobile service revenue dropped 1.1% to 105.44 billion yuan.
China Unicom's broadband and mobile data service revenue fell 4.05% to 73.93 billion yuan.
Data services are in the awkward position of rising usage but falling revenue.
Telecom is a capital-intensive industry, requiring massive annual spending on base station upgrades, network enhancements, and data center maintenance, keeping fixed costs high.
Meanwhile, competition has pushed down data tariffs, eroding average revenue per user.
Additionally, the value-added tax rate on mobile data and other services will increase to 9% from 6% effective Jan. 1, 2026, cutting into carriers' revenue and profit.
In fact, the mobile user market is already saturated.
Data show that China's total mobile phone subscribers have reached 1.847 billion, and mobile internet users total 1.638 billion.
Both far exceed the country's 1.4 billion population, leaving little room for large-scale new growth.
5G mobile phone subscriber numbers.
After voice and SMS, the data business is also running out of steam.
Following that logic, carriers' computing power business is a bright spot.
In the first half of this year, China Mobile's computing power service revenue reached 52.9 billion yuan, up 14% year-on-year.
Computing power and intelligent services together accounted for 22.6% of main business revenue, up 2.2 percentage points year-on-year, making them the primary growth curve.
China Telecom posted RMB 31.1 billion in intelligent business revenue, up 7.1% year-on-year, representing 12.8% of total revenue.
China Unicom's computing power revenue rose 13% to RMB 41.9 billion, positioning it as the primary engine driving overall growth.
Legacy businesses are declining, while new businesses are taking off.
Telecom operators are quietly shifting from old to new growth drivers.
The structural tilt in capital expenditure underscores this trend.
China Mobile's RMB 40 billion data center in Hohhot covers 1,402 mu (about 934 acres), with a planned gross floor area of roughly 1.42 million square meters.
Construction started in 2012 and operations began in 2016. With total computing power of 20,100 petaflops, it is China Mobile's largest, most technologically advanced and best-supported core node.
In the first half, the three carriers' total capex reached RMB 117.5 billion, with computing-network investment of about RMB 40 billion — 34% of the total — up more than 80% year-on-year.
The share of computing-related investment at the three major carriers.
For carriers, selling computing power is not a choice but a matter of survival.
Selling Computing Power: Can Telecom Operators Win?
Many believe that with nationwide networks, data centers and infrastructure, carriers will inevitably replicate their dominance from the traffic era.
In fact, the computing market is far more competitive than the traffic market.
Start with market share.
China's public cloud IaaS market totaled $15.97 billion in the second half of 2025, according to IDC's tracker.
Alibaba led with a 27.9% market share, holding a clear advantage in scale.
China Telecom and China Mobile together hold a 22.2% cloud share—just 80% of Alibaba's.
Huawei follows closely with a 9.5% market share.
China's public cloud IaaS market has a distinct competitive structure.
From another angle, China Telecom and China Mobile together hold a 22.2% share of the public cloud IaaS market; including China Unicom, the combined share is about 30%.
In other words, the three telecom operators account for only 30% of the market.
The dominant force in China's computing power market is internet cloud providers such as Alibaba, Huawei and Tencent, not telecom operators.
How did internet cloud providers achieve their leading position?
Besides first-mover advantage, internet cloud providers' core advantage is their full industry chain.
Alibaba Cloud, for example, has become the only public cloud vendor in China with full-stack self-developed capabilities in GPU, CPU, storage and network chips.
At the chip and computing infrastructure layer, Alibaba Cloud, through its semiconductor arm T-Head, has developed a full line of self-designed chips covering AI training/inference, general computing, networking, and storage.
At the model layer, Alibaba Cloud has gained command of the "operating system" of the AI era through its self-developed large models. Its latest flagship model, Qwen3.7-Max, has achieved top rankings on several authoritative leaderboards.
At the inference and application layer, Alibaba Cloud converts computing power and model capabilities into ready-to-use products and services.
In May 2026, Alibaba Cloud announced the full opening of its Bailian platform, partnering with Moonshot AI, MiniMax, Zhipu AI and others to offer one-stop services with a single entry point and multiple model options.
This full-industry-chain model has allowed Alibaba Cloud to move away from the low-end approach of simply selling computing hardware. It now earns high-value-added profits from software, platforms, ecosystems and solutions, with strong user stickiness.
Meanwhile, Huawei is deepening its industry solutions, Baidu is focusing on an AI-native ecosystem, and Tencent is rooted in various scenarios, keeping the computing power market highly competitive.
Telecom operators, by contrast, remain mostly at the basic IaaS level in their computing businesses, with core revenue still coming from GPU leasing, data center hosting and bandwidth rental.
In essence, selling computing power is no different from selling traffic or broadband in the past.
The bigger concern: in selling data traffic, carriers are the only game in town, able to profit effortlessly.
In selling computing power, carriers are just one of many players, easily replaceable.
If they cling to the package-deal mindset of the traffic era, carriers are doomed to become mere pipe operators.
Making chips and developing large models are not carriers' strengths, nor can they close that gap quickly.
How Will Telecom Operators Tackle the Computing Power Battle?
Consider carriers' three trump cards.
The first is the network.
Every deployment, transmission and scheduling of computing power depends on network support.
Carriers control national backbone networks, edge data centers and network nodes across urban and rural areas, giving them the ability to integrate computing and networking.
Operators can dynamically allocate computing power across regions and deploy edge computing locally, making them ideal for low-latency, high-stability applications — a core capability that internet cloud service providers cannot replace.
China Mobile has 13 intelligent computing centers nationwide, and through its 'Tianqiong' computing-network brain, it manages and schedules 147.7 EFLOPS of computing power across the network.
China Mobile's 'computing-network brain'
The second card is data.
No matter how powerful the large model, without industry data it is a castle in the air.
Operators hold communications network data including user behavior, network traffic, and location information.
In addition, operators cover government and enterprise, industrial, and urban-rural lower-tier markets, with access to massive amounts of real network, terminal, and scenario data.
This data is unavailable to internet cloud service providers and serves as 'exclusive fuel' for training industry models.
The third card is trust.
Computing power and data are core assets of the digital economy.
Key sectors including government, finance, and defense demand the highest levels of data security, independent controllability, compliance, and stability.
Compared with internet cloud providers, telecom operators are regarded as the "national team," giving them distinct advantages in security and compliance.
China Telecom data show it has developed more than 420 industry AI agents, 102 industry skills, and 82 industry digital employees for the government and enterprise market, serving 39,000 customers and reaching 96% penetration among central state-owned enterprises.
The operators' advantages are clear.
Network, data, and trust are the operators' core cards — and they are things internet cloud providers cannot buy with money.
With these three cards, operators are assured a place in the computing power market.
Government and enterprise projects form the bedrock of carriers' computing power business.
The problem is that the government-enterprise market has a ceiling.
To scale up, carriers must shift from selling computing power to selling solutions, and compete across the broader market.
Selling computing power means 'I give you the computing power, you use it yourself.'
Selling solutions means 'You have this business pain point, I help you solve it.'
The former is wholesale/retail, earning a margin on resources; the latter is consulting, earning a premium on expertise.
Selling solutions is precisely carriers' weakness.
Carriers are used to selling standardized products: plans are standardized, pricing is standardized, even customer service scripts are standardized.
Selling solutions requires customization capabilities, industry understanding, and rapid response.
China Unicom's digital-intelligence enabling solution for the transportation industry.
An AI solution for a smart city may span government, transportation, energy, and healthcare.
Operators need to understand each industry's business processes, data standards, and compliance requirements, and also be able to quickly allocate computing resources, mobilize model capabilities, and integrate ecosystem partners for one-stop delivery.
Customer needs vary widely; every project requires customization, with long delivery cycles, uncertain profits, and high talent demands.
Simply put, selling solutions is a tough job.
Unlike selling data traffic, which is easy money, it requires diving into each industry and grinding through it.
Hard as it is, it must be done.
If they fail to act, telecom operators will be left defending their core business while internet cloud providers capture the biggest slice of the computing market.
Operators must step out of their comfort zone and venture into the complex, industry-specific markets.
This is the toughest challenge yet — and the last opportunity.