AI-RAN Alliance
Vision and Mission White Paper
ai-ran.org
Contents
1. Introduction ............................................................................................................................ 3
1.1. Why we need AI-RAN Alliance ........................................................................................... 4
1.2. Founding Members................................................................................................................... 4
2. Vision and Mission ................................................................................................................ 5
3. Strategic Goals and Objectives .......................................................................................... 5
3.1. Development .............................................................................................................................. 5
3.2. Collaboration .............................................................................................................................. 6
3.3. Standards .................................................................................................................................... 7
4. Organizational Structure ..................................................................................................... 7
4.1. Membership ................................................................................................................................ 7
4.2. Governance ................................................................................................................................. 7
5. Working Groups and Roadmaps........................................................................................ 9
5.1. Working Group 1: AI-for-RAN .............................................................................................. 9
5.2. Working Group 2: AI-and-RAN......................................................................................... 10
5.3. Working Group 3: AI-on-RAN ........................................................................................... 11
5.4. Roadmaps ................................................................................................................................ 12
6. Conclusion ............................................................................................................................. 13
2
1. Introduction
The Radio Access Network (RAN) plays a pivotal role in our increasingly connected
world. Over the course of five generations of technological innovation, RAN has vastly
expanded the boundaries of connectivity, enabling the seamless communication
experiences we rely on today. Despite these advances, RAN's full potential remains
largely untapped. Now, with the power of Artificial Intelligence (AI), we stand on the
brink of a new era for RAN—one that is more adaptive, intelligent, performant, efficient,
and versatile.
Traditional RAN architectures provide a solid foundation, offering robust performance
that meets the needs of modern telecommunication. However, AI presents an
opportunity to elevate RAN further, enabling not only higher levels of efficiency and
automation but also potential for new business opportunities. Integration of AI with RAN
allows for RAN-related performance improvements, e.g., by intelligently optimizing the
use of radio resources, dynamically adapting to shifting traffic and spectrum conditions,
and improving the overall efficiency while delivering superior user experiences. It can
also allow for RAN infrastructure to be used for AI application workloads not necessarily
linked to RAN-specific operations. This potentially unlocks new business opportunities
for the CSPs and the infrastructure providers as it diversifies the ways they can monetize
the deployed networks and valuable spectrum investments.
The evolution of AI with RAN is closely tied to advances in compute architectures
ranging from edge processing for low-latency AI inferences to cloud-based resources
for AI model training, each playing a crucial role in the application of AI technologies
with RAN. Cloudification and programmability further enable this transformation,
allowing for higher levels of automation, more efficient operations and agility to create
new services.
Although early explorations of AI-RAN concepts and applications are promising,
collaborative effort across the ecosystem to validate them must accelerate. Thus, there
is a clear need for further development of AI-RAN on real-world scenarios to identify
where AI integration can be truly beneficial. Moreover, efforts to identify and address
challenges related to the introduction of AI-RAN should be undertaken by the telecom
industry to ensure the promised benefits are realizable.
3
1.1. Why we need AI-RAN Alliance
To address these objectives, the AI-RAN Alliance was formed, bringing together
industry leaders, research institutions, and academia to collaboratively explore and
develop solutions and accelerate the realization of an AI-native RAN. By pooling
expertise and resources, the alliance builds an ecosystem to establish a roadmap for
implementing AI-RAN at scale, focusing on practical use cases that maximize benefits.
This partnership seeks to create an AI-native RAN that not only meets the demands of
the present but also anticipates and adapts to the needs of the future, ensuring a more
connected and efficient world for all.
1.2. Founding Members
The AI-RAN Alliance was founded by a coalition of industry and academic leaders,
including network operators, equipment manufacturers, and research institutions. The
eleven founding members of the AI-RAN Alliance are:
• Arm
• DeepSig
• Ericsson
• Microsoft
• Nokia
• Northeastern University
• NVIDIA
• Samsung
• SoftBank
• T-Mobile
• The University of Tokyo
This diverse group of founding members, together with other members, share a vision
of an AI-native future for RAN, where networks are more intelligent, sustainable, and
adaptable. By collaborating on research and innovation, the alliance aims to accelerate
the development of AI based solutions that will shape next-generation connectivity,
unlocking new business models and transforming industries through intelligent, AI-
native RAN systems.
4
2. Vision and Mission
Our vision is to transform the telecommunications industry by integrating AI with RAN,
driving innovation, increasing efficiency, and unlocking new economic opportunities for
the telecom ecosystem – further growing the value of RAN by embracing AI across the
stack.
The AI-RAN Alliance brings together industry leaders and academic institutions to
advance mobile network performance and capabilities through state-of-the-art AI
innovation. Our mission is to maximize RAN asset utilization, generate new revenue
streams, and drive the telecom industry's seamless and profitable transition toward 6G.
3. Strategic Goals and Objectives
The objective of the AI-RAN Alliance is to explore various applications and coexistence
of AI with RAN, enhancing the performance of RAN and unlocking new capabilities and
business opportunities.
This section outlines the alliance’s strategic development objectives, describes how we
foster cross-industry partnerships, and explains how we address industry standards.
3.1. Development
The AI-RAN Alliance three key development areas are:
• AI-for-RAN: focusing on the integration of AI into RAN to significantly enhance
RAN performance, such as improving spectral and operational efficiency,
optimizing radio resource management, and enabling predictive maintenance.
• AI-and-RAN: seeks to create a shared infrastructure between RAN and AI
workloads, enabling concurrent resource utilization on this converged computer-
and-communications infrastructure. The goal is to increase the overall utilization
of such platform infrastructure supporting the RAN workloads and open new
monetization opportunities for running non-RAN AI applications on the same
infrastructure.
• AI-on-RAN: aims to enable new RAN services to enhance AI applications running
at the network edge, to be able to offer new consumer and enterprise services
and applications from the edge of the network.
5
AI-RAN promises to offer the telecommunications industry several benefits, including:
• Increased Efficiency: AI solutions to optimize network operations, leading to
higher spectral efficiency, reduced power consumption and new intelligent
spectrum coordination schemes.
• Cost Reduction: AI technologies help lower operational costs by automating
processes and improving resource management.
• Enhanced Network Performance: AI integration leads to more reliable and
higher-quality network performance, improving user experience.
• New Revenue Streams: AI-RAN capabilities create opportunities for new
services and applications, unlocking additional revenue sources.
• Future-proofing Infrastructure: Scaling up AI capabilities future-proves existing
infrastructure for 6G and beyond, protecting investments and facilitating smooth
transitions to next-generation technologies.
• Knowledge Sharing Among Ecosystem Members: AI-RAN Alliance fosters a
collaborative environment where knowledge sharing among operators, vendors,
academia and other stakeholders accelerates innovation and the development of
standardized solutions.
• Strategic Path Towards Next-Generation RAN: AI-RAN Alliance provides a
roadmap for evolving RAN architectures, ensuring a strategic transition to the
next generation of networks through advanced capabilities such as intent-based
networking and AI-native operations.
3.2. Collaboration
Integrating AI with RAN requires close collaboration among telecom vendors and
operators, AI and cloud technology providers, hardware and software specialists, and
research institutions. The AI-RAN Alliance is committed to building an effective
ecosystem for its members. This includes establishing cross-industry working groups to
facilitate collaboration and providing shared labs equipped with simulators and real-
time systems, access to comprehensive data sets, and robust test, benchmarking tools,
and AI-RAN proof of concepts. The Alliance will produce whitepapers, blueprints, and
evaluation and benchmarking results to further foster industry-wide collaboration.
6
3.3. Standards
The AI-RAN Alliance is not a standard development organization (SDO) and as an
independent body, the Alliance is not mandated to provide feedback to any SDOs.
However, while formal standardization is outside its scope, the AI-RAN Alliance’s
outputs are designed to facilitate AI integration into RAN systems, supporting industry
efforts where appropriate. Development of guidelines and reference materials, such as
blueprints, can potentially be used by standard-setting bodies like 3GPP and O-RAN
Alliance.
4. Organizational Structure
As highlighted above, the AI-RAN Alliance comprises of diverse members, united with
the objective of advancing the implementation of AI with RAN and leveraging AI as a
tool to substantially enhance RAN efficiencies and ecosystem benefits. The structure of
the alliance is designed to promote collaboration, ensure efficient decision-making, and
drive innovation in line with strategic goals.
4.1. Membership
The AI-RAN Alliance is comprised of Executive and General Members. Executive
Members form the core leadership of the alliance, seated on the Board of Directors, and
guide the strategic direction of the organization. General Members represent
organizations committed to advancing and contributing to the alliance’s objectives. All
members participate actively in working groups and the alliance’s technical initiatives.
Each member plays a distinct role in contributing to the advancement of AI-RAN
technologies, with specific responsibilities in research, development, testing, and
commercialization.
4.2. Governance
The AI-RAN Alliance is governed by its Board of Directors, which is composed of the
founding Executive Members. The Board is responsible for the strategic oversight of the
alliance, shaping its long-term vision, and approving key decisions such as the formation
of new working groups. All activities of the alliance are guided by a set of foundational
7
by-laws, and working procedures which define membership categories, governance
protocols, and operational procedures.
While the Board of Directors is responsible for overseeing the decision-making in the
alliance, the technical and marketing development is guided by the Technical Steering
Committee (TSC) and Marketing Steering Committee (MSC), respectively.
Operating within the strategic direction set by the Board of Directors, the TSC owns the
alliance’s technology roadmap and workplan, and is the principal forum for the
discussion and management of the work products. The TSC ensures that the technical
roadmap is aligned the objective of the alliance and is responsible for managing
resources for Working Groups (e.g., labs) and supporting the setup of infrastructure
needed for testing and implementation. In addition, the TSC is responsible for initiating,
reviewing, and responding to technical liaisons from other organizations, as well as
providing and reviewing technical content for the alliance’s marketing efforts, including
whitepapers, blueprints, and demos.
The MSC provides strategic direction and oversight for the organization’s marketing
initiatives. The MSC aims to enhance AI-RAN’s brand visibility, engage diverse audiences,
and support the organization’s mission to advance AI research, analysis, and maturation.
The committee guides and evaluates marketing strategies, campaigns, and activities to
ensure alignment with the AI-RAN Alliance’s goals.
The three Working Groups (WGs) are where the core of technical collaboration occurs,
each tasked with a specific focus area. Each of the WGs operates under its respective
charters, approved by the Board of Directors, that defines its scope, deliverables, and
objectives. The groups are led by Chairpersons, who are responsible for driving active
engagement and ensuring progress in-line with the goals established in their charters.
All WGs follow the Working Group Policies, which provide governance on group
formation, operations, and performance evaluations. These policies ensure that all work
within the groups is executed in alignment with the alliance’s broader goals.
8
5. Working Groups and Roadmaps
Each working group is responsible for work items that cover various focus areas,
collectively contributing to the alliance’s overall goals.
5.1. Working Group 1: AI-for-RAN
WG1 focuses on the application of AI for RAN systems to improve efficiency, capacity,
and performance metrics. The group’s objectives include conducting literature reviews,
defining use cases, and implementing proof-of-concept systems, with an emphasis on
leveraging and advancing AI/Machine Learning (AI/ML) efforts in industry initiatives,
potentially also standards.
The WG1 focus areas are:
• AI-native air interface and signal processing, where the use cases focus on
improving the signal processing chain by replacing or augmenting traditional
methods with AI/ML techniques for enhanced efficiency, reduced overhead, and
better system performance. Some specific examples cover AI for receiver
optimization, end-to-end learned air interface, Joint Communication and Sensing
(JCAS) / Integrated Sensing and Communication (ISAC), Distributed MIMO (D-MIMO),
Extra-Large Scale MIMO (XL-MIMO), etc.
• Positioning and beam management, where AI is used to improve the positioning
accuracy, improve beamforming agility and performance. Examples include SRS and
PRS based accurate positioning, CSI compression and prediction for beam
management, etc.
• Radio resource management and scheduling, leveraging AI to enhance the dynamic
allocation and utilization of radio and spectrum resources. This ensures the RAN can
efficiently adapt to fluctuating traffic patterns and varying network conditions,
ultimately contributing to optimized resource use and improved overall performance.
• Energy and spectrum efficiency, where the focus is on reducing the energy
consumption, thus the operational cost of RAN components, and optimizing spectrum
usage. Some specific use cases include cell energy saving and spectrum
sensing/sharing.
• Network optimization and anomaly detection, where AI is leveraged to detect and
predict network anomalies, as well as to reduce operational risks and enhance
9
network resilience. Specific examples include RAN digital twin and site-specific
optimization, anomaly detection, fault prediction and mitigation, air-interface
assurance, semantic communications, etc.
While AI has demonstrated promising performance across various domains, further
integration of AI with RAN presents new research and engineering challenges.
The requirements on RAN in terms of availability, reliability, security and privacy are very
high. RAN is also a highly distributed system with real-time functionalities operating
under resource and latency constraints, imposing further requirements on AI algorithms,
their implementation and their performance.
AI functionality in RAN is subject to challenging constraints. Solutions need to address
scalability, security, privacy, resource efficiency, robustness, real-time processing, and
more. The network usage will also evolve over time and data patterns and evolving
business and use-case needs. AI-native RAN includes automated monitoring and
management of ML models and pipelines to ensure they remain accurate and aligned
with this evolution and variation.
5.2. Working Group 2: AI-and-RAN
Concentrating on the co-location of non-RAN AI workloads with RAN infrastructure, this
group explores how both can coexist on shared platforms to maximize resource
utilization and infrastructure value. In this regard, the group aims to increase platform
utilization while creating monetization opportunities by running AI on shared
computational resources. The group’s activities include validating multi-tenancy systems
and creating resource-partitioning frameworks driven by AI, for example, in edge data
centers.
The focus areas are:
• Design architecture and components of multi-tenant system, including
configuration management, security and privacy ensuring tenant isolation,
automation, orchestration, and service management.
• Lifecycle management of AI and RAN workloads, where the management of AI and
RAN workloads is addressed, ensuring seamless integration and operation of these
workloads within the AI-RAN infrastructure.
10
• Validating the multitenancy scenarios, where the non-RAN AI and RAN workloads
can run side by side and maintain RAN service level agreements in all scenarios.
• Data center optimization for AI and RAN workloads, where data center resources
are optimized to maximize resource utilization, reduce operational costs, and
contribute to integration goals.
5.3. Working Group 3: AI-on-RAN
Dedicated to identifying key RAN requirements, for present and future systems, to
deliver and benchmark AI applications over RAN connectivity and infrastructure. The
group’s objectives include reviewing current AI-centric techniques, identify challenges,
define use cases, develop test plans, and provide system implementation blueprints.
A critical aspect of WG3’s work is the role of network-differentiated connectivity and
device capabilities which are essential to enhance the Quality of Experience (QoE) for
AI-powered applications.
The focus areas are:
• AI-based multimedia applications, where AI is employed to enhance multimedia
experiences such as video analytics, augmented/virtual reality (AR/VR), immersive
gaming, and next-generation multimedia applications like tactile Internet and
holographic communications.
• AI-based security and critical applications, where AI is leveraged to strengthen
security applications, such as intrusion detection, and support critical applications,
such as healthcare.
• AI-based automation and industrial applications, that deals with the application of
AI for automating industrial operations and managing autonomous vehicles, such as
unmanned aerial vehicles (UAVs), drones and automated guided vehicles (AGVs).
• GenAI/AI-based network services, that focuses on the use of AI and GenAI to
enhance network-based services such as customer service, user interactions, user
positioning, etc. These AI solutions improve communication efficiency and enable
seamless collaboration between users and applications.
• Efficient AI/ML model splitting, where an AI/ML model may be split between the
device and RAN depending on several different factors such as radio link and device
battery life conditions.
11
5.4. Roadmaps
Each working group has set its own roadmap to guide their work progress. The
roadmaps outline specific results and deliverables that will be shared with the
community.
The roadmaps focus on a systematic development and evaluation of AI techniques for
each of the focus areas outlined in section 5.1. They start surveying existing standards
and reviewing the latest AI/ML and RAN literature. The next step is to identify the key
challenges, opportunities, and the most relevant use cases to be prioritized in the
workstreams. The groups also establish performance benchmarks and design high-level
system components for each of the selected use cases. For the evaluation, labs are built
using appropriate tools as well as member capabilities, frameworks and platforms while
ensuring vendor and technology neutrality in core functions to ensure a multi-vendor
solution space and foster interoperability. Different implementations are evaluated
through a comparative analysis which should leverage common benchmarks and enable
fair comparison and highlight tradeoffs. Additionally, blueprints for system
implementation, reference interoperation and evaluation frameworks are developed, and
the need for standardization is assessed, including identifying relevant organizations for
collaboration.
The tasks above collectively provide a systematic approach to exploring AI techniques,
from initial research to potential standardization, fostering innovation, practical
implementation, and realistic evaluation and validation.
To foster collaboration, expand membership, and drive industry adoption of AI-RAN
solutions and blueprints, the marketing roadmap includes producing content such as
whitepapers and case studies on working group initiatives and building partnerships
with industry stakeholders. Additionally, the strategy involves participating in public
events, such as MWC, and organizing live demos to showcase AI-RAN innovations, and
joint PR to highlight member activities, while regularly providing updates from the
working groups to engage the community and highlight progress and opportunities.
12
6. Conclusion
As the telecom industry embarks upon the transformative journey towards the
integration of AI throughout the RAN and 6G, the AI-RAN Alliance offers a unique
opportunity for members to lead this technology shift. Top industry leaders, academic
institutions, and innovators collaborate to shape the future of RAN technologies and
applications through AI integration. The AI-RAN Alliance drives impactful innovations
for enhancing spectral efficiency, optimizing infrastructure, deploying cutting-edge AI
services at the network edge and much more.
We invite interested parties to consider joining the AI-RAN Alliance to:
• Access a collaborative ecosystem that accelerates research and development in AI-
RAN.
• Contribute to new and existing AI-RAN use cases that enhance network performance,
efficiency, sustainability, and functionality.
• Work alongside global leaders to pioneer the next generation of mobile networks.
• Test and validate AI/ML concepts and ideas in industry-led testbeds, with results
published in a peer-reviewed manner.
As we progress, we will soon share more detailed updates on each working group. Stay
tuned for further insights into how the AI-RAN Alliance is driving AI integration in RAN
use cases, unlocking new opportunities for innovation, value creation, and advancing the
ever-evolving intersection of AI and RAN capabilities.
13
