Global Communications Network Restructuring Imminent: Consensus, Disagreements, and Hidden Battles Among Three Factions

In this game involving a trillion-dollar market, the attitude of operators will be crucial. They are eager to reduce costs, increase efficiency, and develop new business models, but also wary of the hardware vendors’ “pipeline” fate repeating itself.

Text | Researcher Zhou Yuan, Caijing

Editor | Xie Lirong

If artificial intelligence (AI) over the past two years has been reshaping the application layer across industries, starting from 2026, it is initiating a fundamental reconstruction of the communication networks at the very bottom of the industry.

At this year’s Mobile World Congress (MWC 2026), themed “The IQ Era,” a clear consensus has emerged — as AI becomes a basic resource like water and electricity, the communication channels that carry it must be redefined first.

On February 28, Nvidia founder and CEO Jensen Huang publicly stated: “AI is redefining the computing field, driving the largest infrastructure wave in human history, and the telecommunications industry will become the next main battlefield.”

This reconstruction is full of uncertainties: on one hand, the industry has reached an unprecedented technical consensus on the question of “where the network is headed” in response to the flood of data brought by AI agents; on the other hand, a “three-country” game led by Nvidia, Qualcomm, Huawei, and ZTE is quietly underway over the path of “how to reach,” with the outcome directly determining the distribution of value and influence in the global communications industry over the next decade.

Consensus: The “Upward Surge” Era is Coming

During the Chinese New Year in 2026, a Shanghai art museum launched an AI agent-guided tour service. Visitors could upload photos or videos of exhibits via their phones to receive real-time explanations and background information.

This AI agent service was immediately popular, serving over 3,000 people daily, but it also caused a shift in traffic structure: in traditional mobile networks, uplink traffic accounts for less than 10%, but in this AI scenario, uplink traffic surged to 63%, surpassing downlink for the first time.

This is a microcosm of the changing network demands in the AI era — AI is changing the “tidal direction” of traffic.

Traditional mobile networks are designed for “downlink” — whether streaming videos or browsing web pages, 90% of traffic flows from cloud to user. But the logic of AI agents is: terminal data collection, cloud inference, real-time feedback. This makes uplink traffic a bottleneck for the first time.

If the network shifts to center around AI agents, the logic is: AI proxies handle perception, interaction, decision-making, and full-process execution. This transformation pushes the network from “downlink dominance” toward “both uplink and downlink, with uplink prioritized.” The real-time data collection by AI agents, cloud inference, and instant feedback demand a “big bandwidth, low latency, high reliability, high stability” — the “one big and three high” core requirements.

Uplink capacity has been recognized as the key to “Network for AI,” a critical direction for network upgrades in the AI era.

During MWC 2026, GSMA (Global System for Mobile Communications Association), together with Huawei, Nokia, YuShu Technology, China Telecom, China Unicom, Turkcell, and global industry partners, launched a large-scale uplink network initiative for operators. With core indicators of ubiquitous 20 Mbps and peak 1 Gbps, they aim to promote GigaUplink (super large uplink), breaking through bottlenecks in mobile AI deployment and addressing challenges in data transmission, network collaboration, and large-scale AI development.

Currently, Huawei, Ericsson, ZTE, and Nokia have all launched new solutions around “big uplink.”

Huawei believes that the continuous enhancement of 5G-A capabilities depends on new spectrum introduction, with ultra-wide bandwidth U6GHz spectrum becoming key to unlocking network potential. At MWC 2026, Huawei announced a U6GHz full-scenario product series, covering macro, small, and microwave matrices, meeting core needs for large capacity, low latency, and high experience in AI applications, and providing systematic solutions for performance leap and smooth evolution toward 6G (more details in “Searching for the Key Variables of ‘5.5G+AI’, Huawei’s Bold U6GHz Spectrum Layout”).

Ericsson released ten AI-ready wireless products, focusing on improving uplink performance, and introduced neural network accelerators that deploy AI computing power directly into radio frequency units and computing nodes, turning base stations into edge AI execution platforms. On the software side, Ericsson launched several native AI functions, such as AI-native adaptive downlink and AI-optimized beamforming algorithms.

ZTE and China Unicom jointly launched the 5GAxI UniMAX ubiquitous deterministic connectivity solution, which ensures differentiated service experience while improving user experience consistency by over 20%, and spectrum efficiency by over 40%, effectively solving industry pain points like cliff-like boundary experience, interference, and high overhead in dense networks.

Nokia announced a new generation integrated wireless product supporting FDD tri-band fusion and high-power multi-band deployment, with new algorithms to enhance uplink performance and spectrum efficiency, plus deep sleep energy-saving tech, improving network capacity while optimizing operator investment returns and operational costs.

It should be noted that 5G networks can achieve uplink enhancement through localized technology, but capabilities are limited and hard to scale. Truly “big uplink” is a systemic core capability only available in the 5GA era, representing a key generational upgrade for AI and intelligent agents.

China is the fastest country in mobile network construction. According to the Ministry of Industry and Information Technology, by the end of 2025, China’s 5G user base exceeded 1.2 billion, with 4.838 million 5G base stations built. The commercial scale of 5G-A is also progressing steadily, covering over 300 cities.

Meanwhile, China ranks first globally in AI large models and intelligent terminals, which has motivated domestic operators to actively and swiftly build uplink capacity.

China Mobile, in partnership with Qualcomm, Huawei, ZTE, and others, released the “China Mobile 5G-A Super Uplink Action Plan,” proposing and establishing a “3+2+3” uplink enhancement technology system, which boosts uplink capacity through optimizing uplink slot ratios, flexible spectrum use, and terminal capabilities. This plan promotes systematic innovation and large-scale application of uplink enhancement technologies, providing clear guidance for end-to-end industry layout and long-term capacity building.

Currently, China Mobile has completed field validation of key technologies such as 4.9 GHz frame structure adjustment, SUL (Supplementary Uplink), and three-carrier uplink aggregation, with measured results showing peak uplink speeds of 750 Mbps and edge speeds of 20 Mbps, doubling and tripling previous levels.

At China Unicom’s MWC 2026 booth, a staff member told Caijing that during the Beijing Marathon in November 2025, China Unicom and Huawei deployed a 5GA large uplink network solution, achieving over 95% satisfaction for 20 Mbps uplink rate along the entire race track.

China Unicom also announced ongoing efforts to verify core wide uplink capabilities, building end-to-end solutions from terminals to networks and applications, and systematically testing the performance of 5G-A new technologies in complex scenarios to clarify future network evolution directions.

China Telecom has launched 5G-A×AI large uplink pilot projects in six provinces, moving from technical verification to large-scale deployment. During MWC 2026, China Telecom, with Huawei, released and demonstrated 5GA large uplink technology and commercial achievements under the theme “Intelligent Gathering for Large Uplinks.”

However, under clear trends, global telecom operators face multiple challenges, such as high network construction and maintenance costs, continuous traffic growth with declining prices, sluggish personal market growth, and short-term difficulty in scaling enterprise markets.

Therefore, while ensuring “Network for AI,” how operators seize AI opportunities—shifting from traffic monetization to experience and value monetization—becomes another key issue.

Magnus Ewerbring, CTO of Ericsson and head of Asia-Pacific, believes that in the next one or two years, deep integration of technology and application scenarios will significantly expand the breadth and depth of communication services. Differentiated traffic and customized services will help operators transform their business models and open new growth and profit spaces. Network slicing and Network API are crucial core technologies, with the latter enabling open network capabilities and supporting high-performance network demands.

Ericsson’s Network API is an open platform for 5G/5G-A network capabilities, encapsulating the underlying 5G network functions into standardized, callable APIs, allowing application developers, enterprises, and terminal manufacturers to directly invoke network capabilities without understanding complex communication tech, enabling “programmable networks and customizable services.”

Huawei’s Senior Vice President and President of ICT Sales and Services, Li Peng, offers three suggestions: first, evolve all services, terminals, and frequency bands toward 5G-A to solidify the ecosystem; second, enhance intelligence in B.O.M. (Business, Operations, Management) domains to support diverse operations; third, promote infrastructure intelligence to prepare for network architecture upgrades.

Disputes: The Battle of Three Camps Over 6G Roadmaps and Ecosystems

If “big uplink” is a consensus, then how to build an AI-native (AI-Native) network for 6G is the most intense point of disagreement. This is not only a technical debate but also a contest over ecosystem dominance and standard-setting.

Currently, the wireless industry is in a phase of deepening 5G-A and preparing for 6G. Yang Chaobin, CEO of Huawei ICT BG, revealed during MWC 2026 that the global 6G standardization work officially started in March 2026, with the first 3GPP standard version expected to freeze no earlier than March 2029.

Unlike 5G, 6G is not just about higher data rates. Liu Wu, engineer at China Information and Communication Technology Group’s Optical Communication Technology and Network National Key Laboratory, told domestic media that compared to previous generations, 6G’s significant advancement is “full coverage of air, land, sea, and space,” integrating ground, satellite, aerial, and marine communications, with deep fusion of communication, sensing, and AI.

“AI-native” is recognized as the core feature of 6G, but there is clear disagreement on how to build AI-native networks. The core of this dispute revolves around technical routes, ecosystem leadership, and standard-setting, forming a competitive landscape among three major camps, each shaping the future development and value distribution of the industry.

On the stage, three camps are clearly visible.

The disruptor is Nvidia-led AI-RAN alliance.

Nvidia aims to use its expertise in general-purpose GPUs and CUDA ecosystem to overthrow the base station, a fortress dominated by dedicated chips for decades. The AI-RAN alliance led by Nvidia advocates using GPUs to unify wireless access network computing power, enabling base stations to handle both communication and AI tasks. Nokia’s involvement and the planned commercial deployment in 2027 suggest this upheaval is imminent.

To understand Nvidia’s AI-RAN alliance, one must first grasp the challenges faced by Radio Access Networks (RAN): in mobile networks, core networks have become open through virtualization and cloudification, but base stations still rely heavily on proprietary chips, closed interfaces, and customized hardware. This closed nature stems from operators’ strict demands on cost, power consumption, and stability for massive base station deployments, but also hampers intelligent capabilities at the network edge.

Nvidia sees an opportunity, proposing to replace proprietary chips in base stations with general-purpose GPUs, supporting AI and communication needs via its proprietary AI Aerial platform and CUDA ecosystem. This implies a fundamental disruption of existing communication architecture, which is why Jensen Huang said “telecom is the next main battlefield.”

Nokia, SoftBank, T-Mobile, and other US, European, Japanese, and Korean companies have joined Nvidia’s AI-RAN alliance, but Chinese companies are absent. Nokia, however, is a key partner of Nvidia. During MWC 2026, Nokia showcased AI-RAN solutions developed with Nvidia and announced plans to start initial AI-RAN commercial trials in 2026, with full global deployment scheduled for 2027.

The reformers are Qualcomm-led end-to-end communication + AI camp.

Faced with Nvidia’s “dimensionality reduction” attack, Qualcomm and other traditional players, along with China’s three major operators and Xiaomi, have built a second line of defense. They insist on evolution within the 3GPP standard framework, advocating “dedicated communication SoC + edge-cloud collaboration.” This is not only a conservative technical route but also a defense of business models — if base stations become general-purpose GPU farms, Qualcomm’s traditional licensing and hardware sales models could be undermined.

The independent camp is represented by Huawei and ZTE, advocating for autonomous and controllable solutions.

As core players in China’s telecom industry, Huawei and ZTE have not joined the two major camps. Like Qualcomm, Huawei and ZTE believe that 6G and AI-RAN’s intelligence should not be tied to general-purpose GPU architectures. Both are exploring “native communication + deep AI integration” independent routes, becoming an indispensable “third pole” in the 6G race.

ZTE’s Senior Vice President Zhang Wanchun stated during MWC 2026: “AI native does not equal GPU architecture,” explicitly opposing binding base station AI capabilities to general-purpose GPUs. ZTE insists that the construction principle of AI-native infrastructure is heterogeneous computing (ASIC AI + xPU), as the structured nature of communication tasks makes some algorithms more efficient with traditional methods.

Huawei also clearly communicates this stance through its technical solutions. The company advocates embedding AI as an “endogenous gene” into network architecture, with its three-layer AI-centric network injecting AI capabilities at every level (network element, network, service), making base stations smarter (spectral efficiency), networks more autonomous (self-thinking), and services more valuable (multi-dimensional monetization). Huawei’s U6GHz full-scenario solution supports smooth transition from 5G-A to 6G, protecting long-term investments.

Major industry players sharing similar views include Intel. During MWC 2026, Intel’s EVP and Data Center Group GM Kevork Kechichian published an article emphasizing that the CPU-GPU dichotomy should be rejected. He argued that this does not align with infrastructure evolution or how operators build networks. Applying GPU-centric thinking to inference-intensive workloads increases costs, complexity, and creates operational silos. Intel advocates matching the right computing power to different AI workloads.

The covert war: 6G positioning and ecosystem battles

Behind the consensus and disagreements lies the early stage of the global 6G competition.

With 3GPP officially launching 6G standardization work in March 2026, the blueprint for future network architecture is being drafted. China holds a dominant position with 40.3% of core 6G patents.

Patent layout and standard-setting are always the core high ground in the race among companies and nations for 6G dominance. China has performed remarkably well; according to the “China Internet Development Report 2025,” by June 2025, China’s 6G core patent applications accounted for 40.3% of the global total, ranking first worldwide. Recently, the Ministry of Industry and Information Technology announced that China has successfully completed the first phase of 6G technical trials, with over 300 key technology reserves formed, and has fully launched the second phase.

But patent quantity does not equal industry dominance — the real contest is whether future networks will be defined by AI for communication, or by communication carrying AI.

Nvidia’s answer is the former, attempting to reconstruct the value chain through computing power; Qualcomm’s is the latter, trying to adapt AI to networks via standards; Huawei and ZTE advocate for a fusion of both, led by communication industry players.

In this trillion-dollar game, the attitude of operators will be decisive. They desire cost reduction, efficiency, and new business models, but also fear repeating the hardware vendor “pipeline” fate.

Consensus clarifies direction, but disagreements determine how far and how fast. For every player caught in this wave, team choices and strategies are just beginning.