AI Is Reshaping Wireless Communication Architecture and Ecosystem

With the global deployment of 5G-A networks and the accelerated standardization of 6G, the deep convergence of AI and wireless communications has become the most significant driver of innovation and the greatest disruptive force in the telecommunications sector. Extensive research within the telecom industry indicates that AI technologies are reshaping the underlying architecture and upper-layer application ecosystem of wireless communications by optimizing network performance, enhancing edge computing capabilities, and restructuring service models.

The telecom industry is systematically exploring AI-empowered wireless communication trends and cutting-edge technologies across multiple dimensions, including network performance enhancement, operational model transformation, AI-native architectures, and novel applications and business models.

AI Enables Intelligent Leaps in Performance, Operations and Management of Wireless Networks

Network optimization is one of the earliest and most fundamental applications of AI in telecommunications. Regarding transmission efficiency, research shows that AI, through improved channel measurement accuracy and signal demodulation optimization, can increase user-perceived speed of 5G-A networks by 30% and expand coverage radius by 15%.

In beam management, deep learning-based smart antenna systems can achieve up to 20% coverage expansion and up to 35% interference suppression, significantly improving network stability in densely populated urban environments. For example, millimeter-wave beam alignment technology supported by convolutional neural networks (CNNs) reduces beam search time to milliseconds, increasing edge user throughput by 40%.

3GPP Release 18 (Rel-18) defines the technical specifications for leveraging AI in scenarios such as channel state information (CSI) feedback, beam management, and positioning. AI-enhanced channel state feedback (CSF) can achieve a throughput gain of up to 95%. Furthermore, AI-driven distributed training architecture can reduce end-to-end latency for XR services to below 20 ms and energy consumption by 25%. ZTE, through cross-layer resource scheduling AI algorithms, achieves dynamic coordination of computing and communication resources, improving spectrum efficiency by 40%. Channel prediction models based on long short-term memory (LSTM) replace traditional linear estimation, improving CSI feedback accuracy by 18 dB. 

In network operations, generative AI (GenAI) can predict and manage network load, optimize routing, and proactively address potential problems before potential failures impact service. This predictive capability not only improves network efficiency and reliability but also reduces operating costs. In customer service, GenAI is transforming how telecom companies interact with customers. Chatbots and virtual assistants powered by GenAI handle a wide range of customer queries, providing personalized and efficient service around the clock. These AI-driven systems learn from each interaction, continuously improving their ability to resolve customer issues.

Predictive maintenance, network optimization, anomaly detection, and automated troubleshooting using AI and machine learning can enable dynamic resource allocation, traffic load prediction, and congestion management. AI can automate customer service functions, from virtual assistants and chatbots handling queries to intelligent automation of service setup and management, enhancing user experience. ZTE is reshaping the operational paradigm through intent-based networking and large language models (LLMs), achieving a high-level L4 autonomous network and empowering efficient network operation and maintenance.

AI Is Driving a Paradigm Shift in Wireless Network Architecture

Radio access networks (RANs) are transitioning from traditional communication-centric infrastructure to converged computing and communication platforms. Simultaneously, driven by GenAI, AI and RAN are rapidly converging. Consequently, new architectures such as AI-RAN, digital twin RAN (DT-RAN), AI-agent networks, AI-native networks, and autonomous networks (AN) are emerging.

AI-RAN integrates RAN AI workloads on the same infrastructure, leveraging machine learning, automation, and real-time data processing to improve RAN efficiency and adaptability, enabling smarter resource allocation, intelligent interference management, optimized network slicing, automated network operation, AI edge computing, predictive maintenance, and failure detection. As a leader in intelligent RAN architecture for heterogeneous communication and dedicated networks, ZTE's AIR RAN solution sets a new benchmark for wireless network experience, energy efficiency, and maintenance efficiency.

DT-RAN creates an accurate, real-time model of the mobile network using enhanced data and models. The DT model reflects reality more accurately than a simplified aggregated model, enabling the simulation, evaluation, and optimization of physical network entities through synchronized digital replicas. For example, LLMs like ChannelGPT utilize multimodal data from wireless channels and the corresponding physical environment, along with their sensing capabilities, to simultaneously generate multi-scenario channel parameters, relevant map information, and wireless knowledge, as required for each task, based on finely tuned large models. In addition, supported by online multi-dimensional channel and environmental information, network entities will make accurate, real-time decisions for each wireless system layer.

The AI-native design philosophy introduced in 3GPP Rel-18 is driving the evolution of wireless networks from "function-oriented" to "cognitive-driven". Furthermore, agentic AI is a revolutionary approach that embeds intelligent, autonomous AI agents into field operations. The next-generation advanced wireless network architecture, agentic AI RAN, integrates agentic AI to enhance field operational capabilities. Agentic AI consists of a group of specialized agents that act as subject-matter experts (SMEs), each focusing on specific areas such as network troubleshooting, decision support, and workflow optimization. Leveraging this structured, multi-agent AI system, organizations can extend their expertise, reduce problem-solving time, and improve operational efficiency. With agentic AI, network troubleshooting moves away from manual trial and shifts to a data-driven, AI-guided process. When technicians encounter problems, AI agents collaborate to provide precise, real-time suggestions. Agentic AI no longer simply reacts to service outages; it evolves towards proactive and predictive network maintenance.

Whether it's AI-RAN, DT-RAN, agentic AI RAN, or AI-native RAN, the essence lies in a paradigm shift from traditional systems to autonomous, intelligent, and continuously self-optimizing networks.

AI Powers New Services, Applications, Terminals, and Security in Wireless Networks

AI is reshaping wireless communication networks beyond connectivity, supporting a series of novel applications such as GenAI and robotics, and extending AI-driven traffic to wireless networks. ChatGPT is expanding the consumer base for content traffic, shifting from human users to machines, and has the potential to become the next killer application in the telecommunications industry. Furthermore, GenAI, with its ability to efficiently learn complex data distributions, generate synthetic data, and present the original data in various forms, has become a powerful AI paradigm, diversifying existing services and applications. GenAI technologies, especially those supporting LLMs and creative content generation, are beginning to play a key role in the transformation of telecommunications services.

In addition, GenAI is facilitating the creation of new services and content, enabling telecommunications companies to offer unique value-added services. From generating personalized content recommendations to creating digitally interactive virtual environments, GenAI is continuously expanding the range of services. For example, KT M Mobile has launched a new AI-powered eSIM activation service for its South Korean customers, marking a significant advancement in automated mobile service configuration. The new system uses advanced AI algorithms to automate eSIM activation, reducing the need for manual processing and potential errors. Meanwhile, ZTE's AIR Core, based on cloud-native and AI-native technologies, intelligently generates new services through experience orchestration, helping operators create new revenue streams and reshape network value.

Furthermore, the integration of AI into mobile phones is creating new revenue models beyond traditional smartphone sales. AI capabilities enable new business models, including subscription services for advanced AI features, personalized advertising, and enhanced application capabilities that leverage the processors on network devices. These innovations promise to diversify revenue streams, drive deeper customer engagement, and deliver continuously improving, personalized user experiences.

While AI will support entirely new applications, ranging  from autonomous systems to immersive virtual experiences, it will also face more complex threats. AI-centric network architectures may introduce new attack vectors, necessitating robust security measures to mitigate potential threats. AI is revolutionizing cybersecurity by enhancing endpoint visibility, improving anti-malware defenses, and simplifying firewall management. Through behavioral analysis, AI systems can detect anomalous activity, thereby more effectively mitigating potential security threats. This proactive security approach highlights AI's ability to adapt to and respond to evolving cyber threats, ensuring robust network protection. AI can play a role in all aspects of cybersecurity: access control, anti-malware control, firewalls, behavioral analysis, application security, and more.

The convergence of AI and wireless communication is evolving from "tool empowerment" to "architectural restructuring", and from "cost reduction and efficiency improvement" to "value creation". In the future, the standardization of 6G's AI-native architecture, the engineering of new protocols and algorithms such as model context protocols (MCPs), and the establishment of cross-domain federated learning frameworks will be key to breakthroughs in higher-order autonomous networks. With the advancement of 3GPP Rel-20 and the introduction of disruptive technologies such as hybrid quantum cloud, wireless networks will gradually achieve a leap from "intelligent assistance" to "cognitive autonomy", ultimately building a ubiquitous, inclusive, and self-evolving new paradigm for wireless communication.