AI-RAN Convergence: From Separate Tracks to an Intelligence Fabric
- August 30, 2026
- 7 mins
- Technology
- 5g 6g ai ran cloud native network automation open ran telecom
AI-RAN Convergence: From Separate Tracks to an Intelligence Fabric
The radio access network (RAN) is entering another major period of change, but this time the transition is not being driven by a single new architecture or technology. Open RAN, cloud-native infrastructure and artificial intelligence (AI) have largely developed along separate tracks. Increasingly, those tracks are converging around a shared, although varied, vision of an AI-RAN that’s more programmable, adaptive and capable of continuously optimizing itself around network, business and customer outcomes.
This convergence is the subject of a new report, “Rethinking the RAN,” which examines how real-world operator strategies are beginning to move these ideas from architecture diagrams into operating models. The report looks closely at TELUS’ brownfield Open RAN transformation, including how the operator has aligned infrastructure refresh, virtualization, interoperability and automation. It also examines T-Mobile US’ approach to AI-driven network operations, where traditional network KPIs are increasingly being connected with application performance, customer experience and progressively more autonomous decision-making.
The longer-term implications extend well beyond 5G. As AI moves deeper into the radio stack and RAN and AI workloads begin sharing distributed infrastructure, orchestration, data architecture and lifecycle management become increasingly important. The report argues that this points toward a future RAN that functions less like static communications infrastructure and more like an intelligence fabric, dynamically coordinating connectivity, compute, data and policy.
Exactly how operators get there, and where the near-term business cases justify investment, remains the critical question.
The Evolution of Open RAN and AI-RAN
The journey toward AI-RAN has been incremental, building on years of work in network virtualization, software-defined networking, and the gradual opening of RAN interfaces. Open RAN, which began as a movement to disaggregate hardware and software, created the foundation for more flexible and intelligent networks. Cloud-native principles, such as microservices and containerization, brought the scalability and agility of IT to telecom. AI, initially applied to network operations and optimization, has now begun to permeate the RAN itself.
Early AI use cases in telecom focused on areas like predictive maintenance, traffic forecasting, and customer experience management. These were largely centralized, using data collected from the network to drive insights and actions. However, the latency and bandwidth requirements of real-time RAN optimization demanded that AI move closer to the edge, into the RAN itself. This is where AI-RAN comes in, integrating AI capabilities directly into the radio access network to enable autonomous, self-optimizing operations.
The convergence of these tracks is not accidental. It is driven by the need for greater efficiency, performance, and agility in the face of exploding data traffic and the promise of new revenue streams from enterprise and industrial applications. Operators are realizing that a static, hardware-centric RAN cannot meet these demands. They need a network that can adapt in real-time, allocate resources dynamically, and learn from its own operations.
Operator Strategies: TELUS and T-Mobile US
TELUS’ brownfield Open RAN transformation is a case study in how an existing network can evolve toward AI-RAN. The Canadian operator has been a strong advocate for Open RAN, and its approach involves a careful alignment of infrastructure refresh, virtualization, interoperability, and automation. By virtualizing network functions and deploying them on commercial off-the-shelf hardware, TELUS has created a more flexible and programmable RAN. This, in turn, enables the introduction of AI-powered optimization tools that can continuously adjust parameters to improve performance and efficiency.
T-Mobile US, on the other hand, has focused on AI-driven network operations. The operator has been integrating AI across its network to connect traditional network KPIs with application performance and customer experience. This involves collecting data from the RAN, core, and applications, and using AI to gain insights and automate decision-making. T-Mobile’s approach is more focused on the operational side, but it is gradually moving toward more autonomous network management, where AI makes decisions in real-time to optimize the network.
These two strategies illustrate different paths to AI-RAN. TELUS is building the infrastructure foundation, while T-Mobile is focusing on the intelligence layer. Both are essential, and the report suggests that successful AI-RAN deployments will require a combination of both.
The Role of AI in the RAN
AI in the RAN can be applied in several areas: radio resource management, beamforming, interference mitigation, energy saving, and predictive maintenance. The goal is to make the RAN self-optimizing, reducing the need for manual intervention and enabling faster response to changing conditions. AI can also enable new capabilities, such as network slicing, where resources are dynamically allocated to different services with varying requirements.
One of the key challenges is integrating AI with the RAN’s real-time constraints. The RAN operates on millisecond timescales, and AI algorithms must be able to process data and make decisions within these tight windows. This requires specialized hardware, such as GPUs and AI accelerators, and software architectures that can support low-latency inference. The rise of edge computing and the availability of powerful AI chips are making this feasible.
The Convergence of RAN and AI Workloads
As AI becomes more embedded in the RAN, the distinction between RAN workloads and AI workloads begins to blur. In the future, the same infrastructure may host both RAN functions and AI applications, sharing compute, storage, and network resources. This convergence has implications for orchestration, data architecture, and lifecycle management.
Orchestration platforms will need to manage mixed workloads, ensuring that RAN functions meet their service level agreements while also providing resources for AI applications. Data architecture must support the flow of data between the RAN and AI systems, with appropriate data governance and privacy controls. Lifecycle management becomes more complex, as software updates and scaling need to be coordinated across both domains.
The report argues that this points toward a future RAN that functions less like static communications infrastructure and more like an intelligence fabric. This fabric would dynamically coordinate connectivity, compute, data, and policy, enabling new services and business models.
The Business Case for AI-RAN
Despite the technical promise, the business case for AI-RAN remains a critical question. Operators need to justify investments in new infrastructure and software, and they need to see clear returns in terms of cost savings, revenue growth, or competitive advantage. The report suggests that near-term business cases are emerging in areas like network optimization, energy savings, and improved customer experience.
For example, AI-driven energy management can reduce power consumption, which is a significant operational cost. Predictive maintenance can reduce downtime and repair costs. Improved network performance can lead to higher customer satisfaction and reduced churn. Additionally, AI-RAN can enable new services, such as ultra-reliable low-latency communications for industrial automation, which could open new revenue streams.
However, these benefits are not automatic. Operators need to carefully plan their AI-RAN deployments, considering the total cost of ownership, the availability of skills, and the integration with existing systems. The report emphasizes that AI-RAN is not a one-size-fits-all proposition; each operator will need to find its own path based on its specific circumstances.
The Path to AI-Native 6G
The evolution toward AI-RAN is also setting the stage for 6G, which is expected to be AI-native from the start. While 5G was designed with some flexibility, 6G is being designed with AI as a core component. This means that AI will be embedded in the network architecture, enabling new capabilities such as sensing, positioning, and the integration of non-terrestrial networks.
The report argues that the lessons learned from AI-RAN deployments in 5G will be invaluable for 6G. Operators that have experience with AI-powered RAN will be better positioned to adopt 6G when it arrives. Moreover, the convergence of RAN and AI workloads will likely continue, with 6G networks being designed from the ground up to support both.
Conclusion: The Intelligence Fabric
The convergence of Open RAN, cloud-native infrastructure, and AI is transforming the RAN from static communications infrastructure into an intelligence fabric. This fabric will be more programmable, adaptive, and capable of optimizing itself around network, business, and customer outcomes. Operators like TELUS and T-Mobile are leading the way, but the journey is complex and requires careful planning.
The report “Rethinking the RAN” provides a comprehensive analysis of this convergence, offering insights into operator strategies, technical challenges, and commercial outlook. As the industry moves toward AI-native 6G, the lessons learned from AI-RAN will be crucial. The future RAN will be an intelligence fabric, and the time to start building it is now.