China Telecom's AI Economy

At the beginning of 2026, China Telecom Group Co., Ltd. (hereinafter referred to as “China Telecom”) put this AI move on the table.

First, the strategic tone was set. At China Telecom’s 2026 annual work conference, after using “cloud transformation and data transfer” for 7 years, the company added two new characters for the first time—“intelligence and benefit.” The strategic keywords were officially confirmed as: cloud transformation and data transfer, intelligence and benefit.

Second, personnel deployment. On January 13, 2026, Tang Ke was appointed as director of China Telecom Group and deputy secretary of the Party’s leadership group. Previously, during his tenure as vice president, he led an upgrade of the AI + terminal channel ecosystem. This included building two major ten-thousand-card intelligent computing clusters in Shanghai and Beijing, and creating an underlying platform for Xingchen large models.

The real deciding move came from Zhang Yunming, Vice Minister of the Ministry of Industry and Information Technology, in his remarks at China Telecom’s annual meeting: “to take the lead and charge at the front.” That is, in building infrastructure, he should take the lead; in advancing integrated development of the virtual and the real, he should charge at the front. This is a task assigned by higher authorities, and it is also a signed commitment of military orders.

Some people compare AI to water, saying it will ultimately become basic infrastructure like hydropower. Over the past 30 years, what China Telecom has been best at is precisely getting water to flow from the source into every household: laying fiber-optic cables is like installing water pipes; building ten-thousand-card intelligent computing clusters and setting up large-model bases are, in essence, repairing a larger water network.

01

Top-level focus

At the start of 2026, China Telecom’s moves on the AI chessboard were made by management personally stepping in.

First, the chairman focused on driving institutional innovation, optimizing the system and mechanism environment for AI development. The most noteworthy action by China Telecom’s chairman, Ke Ruowen, this year was to secure an institutional environment that matches the AI strategy. During the Two Sessions, he stated plainly that the traditional assessment mechanism oriented toward short-term scale conflicts fundamentally with the development laws of new technologies like AI—high investment, long cycle, and uncertainty. Enterprises need to shift from being guided by short-term scale to being guided by long-term value, and he called for pooling social capital and making good use of market forces to provide institutional safeguards for companies to plan future industries.

Shortly after the Two Sessions concluded, he presided over an expanded meeting of China Telecom’s Party leadership group, and clarified that the goal was to become a technology-leading enterprise, accelerating the construction of a nationwide integrated computing power network.

Second, the general manager focused on industry implementation: building platforms, building ecosystems, and expanding scenarios.

China Telecom’s general manager, Liu Guiqing, appeared more often on the front line of industry, shouldering the external execution role of the AI strategy. In January, he represented China Telecom in signing a strategic cooperation agreement with Seres. Together, they unveiled and inaugurated an innovation industry integration center, directly integrating AI capabilities into the key vertical field of intelligent vehicles. In March, at the Global Cloud Network and Broadband Industry Conference, he proposed that artificial intelligence is entering a new stage of “agentic artificial intelligence,” and advocated strengthening technological innovation and deepening industry collaboration through international cooperation, embedding China Telecom’s AI layout into the global industrial discourse system.

At the same time, he made frequent visits to regional subsidiaries and professional companies in multiple places including Inner Mongolia, Heilongjiang, Guangxi, and Shanghai, giving guidance on business development, technological innovation, and talent team building. At China Telecom Party School’s training class for young cadres, he put forward requirements for young officials: “strengthen the will to take responsibility, enhance the ability to take responsibility, and fulfill the responsibilities of taking responsibility,” emphasizing that they must hone their real skills and establish a correct view of political achievements. These directions are clear: the AI strategy must be implemented at the grassroots level, and first of all, it must have people who can shoulder the tasks.

Third, management signed agreements intensively, securing territory early. At the start of 2026, within a single month, Ke Ruowen held talks and signed agreements with the top leaders of Party and government in Qinghai and Shanxi one after another. The content was highly consistent: artificial intelligence + industrial applications, green computing power, and digital and intelligent transformation. These two cooperations will help secure priority opportunities for AI projects during the “15th Five-Year Plan” period in these two provinces.

02

Revitalization

Since the beginning of 2026, China Telecom has been exerting efforts in four aspects at the same time.

First, organizational activation—letting people who understand the business go into battle. At the beginning of 2026, China Telecom Group completed an important personnel arrangement: Tang Ke was appointed as director and deputy secretary of the Party leadership group. But the action with stronger signals of real combat was the assignment of city-level “number one” leaders to professional subsidiaries on secondment. In March 2026, Du Zhe, Party Secretary and general manager of Yunnan Lincang Telecom, was appointed Deputy General Manager of China Telecom Yikang Technology through a secondment arrangement. Public information shows that Yikang Technology is an important force in telecom medical informatization. Du Zhe has long worked in cities at the grassroots level, deeply involved in improving the smart service capability construction for the Lincang People’s Hospital, and has hands-on experience in renovating frontline hospitals.

Second, business activation—turning relationships into orders. In the first three months of 2026, China Telecom densely rolled out benchmark projects, all centered on “AI + existing systems.” For example, on March 1, Beijing Telecom jointly released the country’s first large model for public hospital operation management with Beijing Anzhen Hospital. It connected three major core systems: the hospital information system, the medical consumables supply chain system, and the operations data center, enabling free data circulation across end-to-end business processes and intelligent matching. As a result, the generation speed of an operations management report was compressed from several days to 5 minutes, and the management response speed improved by 95%. The response time for abnormal early warnings of key indicators did not exceed 5 minutes, and the efficiency of accurate decision-making improved by 60%.

Third, infrastructure activation—repairing the water pipes to be wider and steadier.

Training AI large models is, in essence, the same as using water: it requires a continuous flow of computing power water. On one side, it is to thicken the pipes. On February 24, 2026, the second phase of China Telecom’s Guangdong-Hong Kong-Macao Greater Bay Area data center started construction. Total investment is RMB 5.97 billion, and the long-term plan is RMB 18 billion. It will build ten-thousand-card-level intelligent computing clusters to provide solid support for AI large model training and inference.

On the other side, it is to interconnect the water pipes. In March, China Telecom completed a key technical validation: using a computing-power network with a 50-millisecond level, it connected computing power from 1,200 kilometers away, integrating dispersed data centers, backbone networks, power, and land resources into a unified intelligent computing network. This allows computing power to flow like water—from any source—to where it is needed.

Fourth, capital activation—so that computing power can grow good crops.

The water pipes are fixed, and there is water too. The key is that the water has to be poured into the fields, and only then can the crops grow.

On February 28, 2026, China Telecom led the investment in Mianbi Intelligence for several hundred million yuan. Mianbi Intelligence does edge-side large models. If we say China Telecom’s computing power is water and the computing power network is water pipes, then Mianbi’s model is the nozzle. With water and a nozzle, it can truly be poured into the fields, growing the “crop” of AI applications.

Another field is Tianyi Shilian. This company focuses on video surveillance and visual AI, with clear scenarios and a market big enough. At the start of 2026, it introduced six strategic investors: the Chengtong mixed-ownership reform fund, Hangzhou Data Group, Muxi Co., Ltd., ICBC Investment, Hikvision, and Hillhouse Capital. With capital, technology, and scenarios, the “crops” to be planted in this field next are AI applications worth hundreds of billions, such as urban governance and smart security.

03

Core

While organization and personnel, business scenarios, infrastructure, and capital coordination all pushed forward along four paths at the same time, their common capability base was Tianyi Cloud.

Why Tianyi Cloud? Because China Telecom’s most core assets are still fragmented: more than 400 million users, ground networks of companies across 31 provinces (autonomous regions and municipalities), and more than 200,000 employees. Provincial companies fight their own battles in their respective places, and frontline personnel are scattered on every street and in every alley. If these touchpoints cannot be unified and dispatched, they are only a scattered mess. The role of Tianyi Cloud is to truly string these fragments together.

Tianyi Cloud is like an intelligent water supply network covering the whole country: Xingchen large models run on the main computing power backbone pipelines, and more than 80 industry large models are connected to the network ends. More importantly, on this platform, more than 200,000 frontline employees can deliver AI services to customers without programming—just by clicking, tapping, and dragging. When AI competition moves from vying over the scale of model parameters to competing on the speed of deploying across thousands of industries, the advantages of this system become fully apparent. For AI home features, if no one teaches elderly people to operate them, they are just decorations. For factory quality inspection systems, if no one comes to debug them, they become scrap metal. China Telecom has a team that faces customers directly every day; this advantage of on-the-ground delivery networks is hard for internet giants to replicate.

Tianyi Cloud is precisely the main hub that unifies the dispatching of this service network.

In the first quarter of 2026, Tianyi Cloud was also completing a key transition: upgrading from a cloud service provider to an infrastructure operator for the AI era. Using the logic of repairing water pipes, this transition is clearly divided into three steps:

In January, it showed off its muscles to prove whose pipelines are the most solid. Tianyi Cloud appeared at the global open-source top conference Linux Plumbers Conference, showcasing the core cloud-native technology eBPF. This technology solves the challenge of monitoring large-scale computing power clusters: previously, monitoring 50,000 kernel functions required deployment time on the order of tens of minutes; now it can be done in 1 second, and memory consumption is also greatly reduced. This is like an open declaration to the tech community: in building the computing power pipeline network, Tianyi Cloud has the hardest technical foundation.

In February, it went after users and let them use the water first. No matter how strong the technology is, if it cannot be deployed, it’s useless. Tianyi Cloud’s approach is very straightforward: AI cloud computers come with large models ready to use at startup, and the Xirang platform gives a free 25 million Tokens. The logic behind it is clear: model capabilities are rapidly converging. Whoever gets users using the models first—and helps them form usage habits—will be the one to seize the users.

In March, it set the rules, connecting the pipeline network, terminals, and on-site services end-to-end. At MWC 2026, Tianyi Cloud revealed its complete playbook: Xirang 2.0 can dispatch 87 EFLOPS of computing power (with sufficiently wide pipelines), supporting voice models for 50 dialects, visual models with 600 million daily call volume, and specialized models covering more than 80 industries (corresponding to intelligent terminals for various scenarios). More importantly, frontline employees are fully empowered through a low-code platform: without programming, they can go onsite to complete deployments (forming a scaled on-site service team). In the same period, it jointly initiated a computing power interconnection and interoperability plan with the China Academy of Information and Communications Technology, and worked with Higuang to launch domestically produced confidential computing cloud hosts. The pipeline network must be both well-connected in every direction and airtight without leaks.

By this point, Tianyi Cloud’s role is already clear: the dispatching control center of the computing power pipeline network (Xirang 2.0), the distribution platform for intelligent water faucets (80+ industry models), the empowerment system for the on-site installation teams (200,000 frontline personnel), and the security mechanism for water safety (confidential computing). With four roles merged into one, Tianyi Cloud is the infrastructure operator in the AI era—not just repairing water pipes, but operating the entire intelligent water supply system.

04

Breakthrough

Betting on AI, and the results are evident.

Financial reports show that in the first three quarters of 2025, China Telecom’s industry digitalization business revenue reached 1055.49 billion yuan, up 5.8% year over year. Tianyi Cloud revenue continues to grow rapidly. Computing power business is also expanding. Some institutions predict that China Telecom’s 2026 computing power revenue is expected to be 80 billion yuan, with a gross margin of about 45%. This is real, tangible market demand.

But during the betting process, two thresholds must be faced.

The first threshold: When will the government and enterprise money be recovered?

Public information shows that in the first 10 months of 2025, 18 provincial companies of China Telecom had information technology service projects with negative revenue growth. For example, the decline for Inner Mongolia Telecom exceeded 40%. At the same time, the project revenue collection progress for Inner Mongolia Telecom was also below 80%. A decline in project revenue collection may be due to factors such as increased difficulty in obtaining new projects or longer payment recovery cycles.

Data also confirms this. Financial reports show that by the end of 2025, China Telecom’s accounts receivable turnover rate had fallen to 9.07 times per year, while the median over the previous 8 years was 17.71 times per year. This means that the same amount of money, which used to be turned over 17 times in a year, can now only be turned over 9 times.

Behind this is local governments’ fiscal pressure. The question is: if customers do not have money for a long time, how far can the path of advancing funds to sustain government-enterprise revenues go? In 2026, China Telecom’s government and enterprise system must answer one question: can it create products that customers truly are willing to pay for—rather than just buy on credit?

The second threshold: In the individual consumer market, is there still an opportunity?

Financial reports show that in the first three quarters of 2025, China Telecom’s mobile users had a net increase of 14.9 million, but monthly revenue per user declined. The broadband market is similar: users are still growing, but revenue contributed per user is decreasing.

Against this backdrop, AI terminals are being highly anticipated. Zhang Junqing, Vice President of China Telecom’s marketing department, disclosed that the goal for co-creation of terminals across all categories in 2026 has been set: breaking 15 million AI terminals, over 80 million smart home terminals, and over 65 million industry digitalization terminals.

With the targets set, actions have also followed. In March, Zhejiang Telecom partnered with Huawei to launch two home AI products: Tianyi Smart Box and Tianyi Smart Screen. Tianyi Smart Screen focuses on AI interaction: watch live streams, make calls, check surveillance, ask for recipes, and control home appliances—done with one sentence. Tianyi Smart Box focuses on the content ecosystem: it integrates China Blue Telecom TV and provides one-stop access to platforms such as iQIYI, Youku, Tencent, Mango TV, and Bilibili.

In the same month, Hubei Telecom signed a strategic cooperation agreement with Xiaomi. Both sides will increase terminal subsidy efforts and jointly build customized AI mobile phone terminals.

But the problem is that in directions like AI phones and smart homes, China Mobile is also doing it, and Huawei and Xiaomi are also doing it. In a competitive landscape with many similar products, how will China Telecom build differentiated advantages to attract users?

Massive information and precise interpretation—available in the Sina Finance app