Mobile AI explosion imminent? China Telecom ramps up 5G-A x AI high-uplink network construction
```
Author | Huang Yu
In the era of mobile internet, ordinary users' perception of network experience upgrades has long been based on “downlink capability”—web pages open faster, videos are clearer, downloads are quicker. This logic is now being rewritten by AI.
Mobile AI, rapidly coming into effect, is accelerating the growth of uplink traffic and simultaneously raising demanding requirements for network latency and uplink/downlink speeds. As operators propose token-based business strategies, they are also starting to invest heavily in uplink network construction.
During the 2026 MWC Shanghai, Wall Street Insights learned that so far, Shanghai Telecom, in collaboration with Huawei, has built 3,450 3CC sites, achieving continuous deployment of a large uplink network. Utilizing F+T 5G-A multi-carrier aggregation and uplink technologies, actual peak uplink speed reached 1Gbps, and peak downlink speed nearly 4Gbps.
When "large uplink" has shifted from an option to a necessity, Shanghai Telecom was the first to begin building its large uplink network foundation. In March 2026, it announced the completion of the nation’s first 5G-A x AI commercial large uplink network.
It is reported that currently, Shanghai Telecom has achieved ubiquitous 20Mbps continuous uplink coverage across ten scenarios, such as popular commercial areas and pedestrian streets in the city's core.
By the end of 2026, Shanghai Telecom is expected to achieve ubiquitous 20Mbps continuous uplink coverage in central urban areas.
This is not a traditional network speed upgrade—it’s more of a directional shift in the fundamental capabilities of mobile communications.
In March 2026, GSMA, together with leading operators and global industry partners, released the “Large Uplink Network Initiative,” establishing GigaUplink as the priority direction for network evolution and proposing core capabilities such as “ubiquitous 20Mbps and peak 1Gbps.”
China Telecom's Chief Scientist Bi Qi also pointed out at the 2026 MWC conference that the comprehensive outbreak of Mobile AI is pushing networks from “downlink priority” to a new stage of “balanced uplink and downlink, with latency as a decisive experience indicator.”
Behind this change, the era of Mobile AI is placing entirely different demands on networks.
Previously, phones mainly received information, but AI terminals are now constantly sending data to the cloud.
Whether it's AI glasses recognizing the world in real time, phones with multimodal assistants continuously understanding the environment, or robots working steadily, they share a common feature—constantly uploading images, voice, video, and environmental information, then returning results immediately after model inference.
This means AI terminals need to continuously occupy uplink bandwidth and place higher demands on latency and stability.
Wider coverage by the large uplink network will provide a more reliable foundation for Mobile AI services such as real-time multimodal interaction, embodied intelligence, and AIGC, accelerating the large-scale implementation of Mobile AI.
Take the increasingly popular AI-assisted exhibition viewing as an example. While visiting a museum, AI applications continuously collect and upload video to the cloud for recognition, then feedback results to the user for real-time guidance and interactive experiences. However, just having devices equipped with high-quality AI applications is not enough; excellent uplink network quality is also essential.
Zhao Dong, Chief Marketing Officer of Huawei Wireless Network Business, predicts that as more people use Doubao to view exhibitions, they'll soon realize Doubao responds slowly—not due to Doubao itself, but because the network can't keep up.
As multiple users go online simultaneously, the system requires the network to provide deterministic uplink capability; otherwise, delays, lag, or even recognition interruptions will occur.
Wall Street Insights has learned that Shanghai Telecom has deployed the 5G-A x AI large uplink network in the West Bund Museum of Xuhui, Shanghai, and users who have activated related service packages will notice faster Doubao response when using it for exhibition viewing.
This means the key to future AI experiences may require not only strong model capabilities, but also network synergy capabilities beyond the models themselves.
From AI phones to AI glasses and embodied intelligent devices, a large number of terminals inherently depend on wide-area network operation. Without stable uplink, terminal capabilities are weakened, and cloud models can hardly realize their value.
Of course, network infrastructure alone is just a prerequisite. Only with consumer recognition and the formation of consumption behaviors can the 5G-A x AI large uplink commercial network truly achieve the leap from technical implementation to large-scale commercialization.
For years, operators have tried to create new value by upgrading networks, but most of the time user perception has not been obvious. The AI era gives networks a new opportunity to return to the center of industry, but whether this can truly become new commercial returns ultimately depends on whether applications experience explosive growth.
Risk Warning and DisclaimerThe market carries risks; investments require caution. This article does not constitute individual investment advice, nor does it take into account unique investment objectives, financial situations, or needs of specific users. Users should consider whether any opinions, views, or conclusions in this article suit their particular conditions. Investment based on this is at your own risk. ```