From traffic to intelligence: How does Huawei 5G-A open up the AI era?
At the beginning of 2025, when Deepseek emerged out of nowhere and flapped the wings of intelligence, attracting numerous industries to connect and adapt, the wind of intelligence began to spread more rapidly to more corners of the physical world.
As large models move from the cloud to terminals, traditional hardware represented by mobile phones, cars, and PCs are all trying to seize this wave of intelligence to find new growth points in their respective fields.
When the scenarios where AI penetrates undergo qualitative changes, more and more fields need to break down information silos and promote the advancement of intelligent experiences through data sharing and innovative interactions. In this process, how ICT enterprises that provide the core technology foundation can initiate the transformation of communication technologies in the AI era has become an issue that cannot be ignored.
2024 was a crucial node in the wave of intelligence.
On the one hand, it was the first year of 5G-A commercialization. In June of that year, the first version of the 5G-A R18 standard was officially frozen. Meanwhile, according to the "Mobile Economy Report 2025" released by GSMA, by the end of 2024, the number of 5G connections globally exceeded 2 billion.
On the other hand, it was also the first year of AI entering terminals. When large models were installed in smartphones, AI began to spread to more terminal hardware. From mobile phones and PCs to wearable devices, leading manufacturers have been considering launching more advanced AI terminals. It is estimated that by the end of 2024, the global shipment volume of AI terminals reached 170 million units.
If 5G-A broadens the boundaries of communication performance and challenges the physical limits of network speed, then the in-depth integration of 5G-A and AI technology is driving intelligent terminals into a new era of "independent thinking", thus giving rise to a new paradigm of "intelligent interconnection of all things".
Huawei Enterprise Business Booth
A question worth answering is: When 5G-A meets AI, how will Huawei, which made an appearance at MWC 2025, drive the digital world into the "Mobile AI Era"?
Boosting Intelligence with Networks: Building the Technological Foundation for the Mobile AI Era
The transformation of communication technologies is constantly reshaping the production and interaction of human information.
In 1991, the GSM system and mobile phones opened the era of mobile voice, making it possible to communicate anytime and anywhere. In 2007, the development of 3G technology launched the era of mobile Internet, making global information accessible at one's fingertips.
Currently, AI technology is developing at a speed beyond everyone's imagination, with a surge in AI applications and innovations. What kind of communication network do we actually need?
At MWC 2025, Yang Chaobin, a director of Huawei and CEO of the ICT BG, believed that humans are producing tokens on an unprecedented large scale.
Yang Chaobin, a director of Huawei and CEO of the ICT BG, delivering a keynote speech
In the past eight months, the daily average token usage increased by 33 times, among which the paid tokens increased by 15 times. The emergence of Deepseek has reduced the price of a single token by 97%.
He predicted that by 2030, the daily average traffic driven by token consumption across the entire network in China will reach 5 million TB, more than 3.5 times the current daily average total traffic of the mobile Internet (900,000 TB).
Tokens, as the "workload unit" for large models to process text, are linked to the computing power cost of AI. Their rapid growth is driven on the one hand by the implementation of technologies such as AIGC, and on the other hand by the diversified development of AI terminals such as smartphones, smart cars, and embodied intelligence.
As human-computer interaction evolves from traditional touch to full-modal interaction, the open source of large models has given rise to a large number of intelligent applications. When intelligence will be as ubiquitous as air in the future, as Yang Chaobin, a director of Huawei and CEO of the ICT BG, said, "The network architecture technology and capabilities urgently need to be innovated to meet the needs of the AI era."
Firstly, sub-millisecond latency is required.
Fields such as autonomous driving, remote surgery, and industrial robots require real-time responsive AI capabilities. Take autonomous driving for example. Autonomous vehicles need to process sensor data and make decisions within milliseconds. Excessive latency may lead to accidents.
Currently commercially available 5G-A can offload data processing to nodes closer to users through Multi-access Edge Computing (MEC), reducing the data transmission distance. Compared with the 4G network with longer latency, 5G-A can better meet the real-time requirements of AI.
Now, with the increasing popularity of smart cars and the maturing of the Internet of Vehicles, Huawei, in collaboration with partners such as Shanghai Mobile, has built the world's first 5G-A Internet of Vehicles full-factor demonstration route.
Launch ceremony of the "5G A² Demonstration City"
Secondly, higher bandwidth and uplink speed are needed.
The data demand in the AI era is growing exponentially. The International Data Corporation (IDC) once predicted that the global demand for AI training data will reach 200 ZB in 2025, equivalent to 20 times the current global data stock.
For example, a smart city may generate several terabytes of surveillance data per second, which needs to be quickly uploaded to the cloud for processing. 5G-A can increase the uplink rate to the gigabit level (above 1 Gbps) through multi-band aggregation, millimeter-wave band expansion, and super uplink technology, thus meeting the large-scale data exchange requirements of applications such as high-definition video analysis, multi-modal interaction, and large model training.
An example that has already become a reality is that during the Hangzhou Asian Games, Huawei helped Zhejiang Mobile build 13 5G-A stations in the Asian Games Village, achieving the world's first deployment of a 10-gigabit network. This ultra-high bandwidth network speed not only provided new viewing experiences such as naked-eye 3D and 8K ultra-high definition for the audience of the Hangzhou Asian Games but also promoted Hangzhou to become the "Fastest City".
Next, it is necessary to connect a massive number of devices.
In the era of the Internet of Everything, the number of Internet of Things devices is growing explosively. How to connect millions of devices simultaneously and achieve stable communication has become a test for ICT manufacturers. Compared with 4G, which only supports about 100,000 devices per square kilometer and is prone to network congestion, 5G-A can achieve a connection density of millions by optimizing connection efficiency and reducing device energy consumption.
This connection density can greatly improve the deployment efficiency, especially for some smart homes or smart factories. For example, in the automated factory of Great Wall Precision in Baoding, Hebei, Huawei helped China Unicom and Great Wall Precision build the world's first wireless flexible production line, which was the first application of 5G-A in high-end industrial control.
Industrial robots receive timely instructions through the high-speed wireless data transmission of the 5G-A indoor base station to carry out work
Using 5G-A, it only takes milliseconds for these industrial devices to receive instructions from the control end, and they can operate stably at high speed without disconnection. Compared with traditional production lines, without the constraints of industrial cables, the production line equipment can also cooperate efficiently, and an innovative model of 5G + Industrial Internet has taken shape.
From lower latency, to higher bandwidth, and then to more connections, when the scenarios in the AI era undergo qualitative changes and the data demand grows exponentially, the 5G network also faces higher requirements. This is exactly the value and mission of 5G-A, as an evolved and enhanced version of 5G, with its four major characteristics of 10-gigabit downlink, gigabit uplink, tens of billions of connections, and inherent intelligence.
From the first commercial use of 5G in June 2019 to the official commercialization of 5G-A in June 2024, as the 5G network has been continuously extending from infrastructure coverage to high-value scenarios such as consumption, industry, transportation, medical care, and government affairs, more than 10 operators globally have deployed 5G-A on a large scale, and eight of them have launched commercial packages. At MWC 2025, operators including China Unicom, China Mobile, and China Telecom also announced large-scale commercialization plans.
It is understood that China Unicom announced that it will fully launch the large-scale commercial use of 5G-A in the main urban areas of 39 cities. China Mobile was the first to announce a commercialization plan for 300 cities in 2024 and launched the first 5G-A package in Shanghai Mobile in June 2024, targeting key scenarios such as business travel, gaming, and live streaming. China Telecom also promoted the implementation of a dual 10-gigabit city (10-gigabit optical network and 10-gigabit mobile network) in Shanghai and announced a joint innovation plan at this MWC.
As Yang Chaobin, a director of Huawei and CEO of the ICT BG, said, the fission of application scenarios and the evolution of technology in the AI era have brought unprecedented development opportunities to the ICT industry and also put forward new requirements for network infrastructure. It is urgent to achieve all-round breakthroughs in dimensions such as bandwidth, latency, coverage, and operation and maintenance.
Empowering Networks with Intelligence: Driving the Leap of Network Digital and Intelligent Productivity
As the fundamental technology for network automation, AI, while deeply integrating with 5G-A, has also injected new impetus into business experience and network operation and maintenance.
As early as MWC Shanghai last year, Yang Chaobin, a director of Huawei and CEO of the ICT BG, predicted that by 2030, AI will be applied to tens of thousands of industry-specific scenarios, the number of AI assistant users will reach one billion, and the number of users of digital live streaming, smart homes, embodied intelligence, etc. will all reach hundreds of millions.
At MWC Barcelona this year, he proposed that AI is reshaping our social landscape from three dimensions: Firstly, it empowers individuals to achieve differentiated business experiences; secondly, it drives organizations to evolve towards intelligent collaboration; thirdly, it promotes society to upgrade towards AI universalization.
Firstly, humans are the yardstick of everything in the AI era.
At the individual level, Yang Chaobin believes that the experience requirements of individual users are constantly differentiating. Individuals in different industries such as food delivery workers, live streamers, and gamers have different requirements for network uplink, downlink, and latency.
While AI drives the leap in traffic, AI-driven uplink has also become a common requirement for every user. "The needs of individual users have gradually shifted from a single traffic experience to a personalized experience, which requires the network to have differentiated capabilities to meet the different needs of users," Yang Chaobin said at MWC.
Secondly, the organizational collaboration mode has evolved from human-to-human collaboration and human-to-intelligence collaboration to intelligence-to-intelligence collaboration. When an organizational network needs to connect a larger number of devices, with higher interaction frequency and faster response speed, the configuration of network resources also needs to break through the traditional paradigm and shift towards supporting dynamic configuration of connections, providing ultra-high bandwidth with equal emphasis on uplink and downlink, high reliability, and extremely low latency in milliseconds.
Finally, at the social level, when L3/L4 intelligent cars will become the mainstream in 2030, AI will fully empower urban services and even play a major role in social health care. "AI applications will be ubiquitous, empowering the whole society to transform from digitalization to intelligence," Yang Chaobin believes that the universalization of social AI will stimulate full-scenario coverage and massive demand for connections. It is estimated that by 2030, the global demand for connections will reach 200 billion.
From individuals to organizations and then to society, as the wave of intelligence accelerates, at the Product and Solution Launch Event during MWC 2025 Barcelona, Huawei launched an AI-Centric Network solution.
In Yang Chaobin's view, Huawei's AI-Centric Network builds a solid ICT network foundation by enhancing network capabilities to achieve full-domain connectivity and improving the intelligence level to create a new paradigm of operation and maintenance oriented towards applications, thereby accelerating business and business model reshaping and seizing new opportunities in the intelligent era.
How exactly does Huawei, which advocates "empowering networks with intelligence", drive the leap of network digital and intelligent productivity?
Firstly, through full-domain network connectivity, it enhances the network's capabilities of ubiquitous ultra-wideband, determinism, and intelligent scheduling through AICore and 5.5G series solutions. When the network can intelligently orchestrate resources such as network routes and bandwidth, it can provide deterministic SLA guarantees for intelligent applications.
Secondly, with an operation and maintenance architecture oriented towards applications, every application in the intelligent world can obtain optimal guarantees.
As reducing OPEX has become an important challenge for operators, creating a new operation and maintenance model has become the key to breakthrough. Through instant and multi-dimensional experience guarantees, personalized operations can be achieved; through 7*24-hour optimization based on traffic prediction, network optimization at the application level can be realized; through Agent closed-loop self-healing and Copilot-assisted on-site operations, predictive and proactive maintenance can be achieved.
Finally, with an AI-centric network, it helps operators leap from the single model of "traffic management" to the multi-dimensional model of "differentiated experience management".
In the Mobile AI era, the new network foundation has also created a new fertile ground for innovation for operators. With the explosion