Using the AI-RAN Alliance framework as its foundation and leveraging our RAN Automation report, this new report sizes and forecasts the AI-driven RAN hardware, software and services revenue market. Defined as AI-RAN, it includes GPU-based RAN (gRAN) and its orchestrator; the O-RAN Alliance's near-Real Time RAN Intelligent Controller (RIC) xApps, non-Real Time RIC rApps, and service management orchestration (SMO); and the Third Generation Partnership Project's (3GPP) self-organizing networks (SON). Lastly, this report analyzes the ongoing evolution of AI-powered SON to RIC and SMO to AI-RAN.
Multitenant RAN and AI services enhance capital investment efficiency
and revenue difersification
Source: Softbank Corp. Research Institute of Advanced Technology, December 2024.
TABLE OF CONTENTS
ABSTRACT
KEY TAKEAWAYS
AI IS ALREADY IN THE RAN AND READY TO TAKE IT OVER, GEAR UP FOR AI-RAN!
- THE ABILITY TO RUN AND MANAGE RAN AND AI WORKLOADS CONCURRENTLY DEFINES AI-RAN
- NVIDIA's AI-RAN transforms traditional RAN from a cost center to a profit center
- WHILE SOFTBANK PLANS TO DEPLOY AI-RAN IN 2026, MICROSOFT PROPOSES A DISTRIBUTED AI PLATFORM FOR 6G RAN
- Whatever Microsoft and SoftBank do with AI in the RAN, it needs to be smoothly integrated in existing networks
- AI-RAN BRINGS GPU TO THE RAN
- That sparked an agitated GPU vs. CPU debate!
- AITRAS IS THE NAME OF SOFTBANK'S AI-RAN THAT ALSO INCLUDES FUJITSU'S RU AND VRAN SOFTWARE
- A GPU-based AI-RAN is an expensive proposition but SoftBank does not think so
- Microsoft's and Softbank's AI-RAN research led to the launch of the AI ran alliance
- While AI for RAN is here, SoftBank validated AI and RAN but AI on RAN is far fetched
- WHILE GPU-BASED RAN IS INCIPIENT, THE AI FOR RAN MARKET IS ALIVE AND KICKING
- AI for RAN will dominate the AI-RAN market throughout the forecast period
- We count on Softbank to introduce its AITRAS in its RAN footprint in 2026
- We expect Japan and South Korea to be early adopters of GRAN as a path to 6G
- In the U.S., only T-Mobile is on board so far but not in a rush
- CHINA IS SHOOTING FOR L4 AUTONOMOUS NETWORKS AND HAS ITS OWN AI-RAN APPROACH AND ECOSYSTEM
- In fact, with DeepSeek, China has a cheaper alternative to offer, which is going to change the world's AI order
- DeepSeek R1's support for Huawei's Ascend platform poses a direct challenge to Nvidia and therefore, to the AI-RAN Alliance
- AI-RAN WOULD HAVE NEVER BEEN POSSIBLE WITHOUT THE EVOLUTION FROM D-RAN TO C-RAN, VRAN AND OPEN RAN
- Embedded in D-RAN, D-SON paved the way to AI for RAN
- AI for RAN started with the introduction of 3GPP SON algorithms in 2008...
- Leveraging SON and MDT, 3GPP AI RAN Rel. 18 is the first release of 5G-Advanced that introduces AI and ML to the RAN
- 3GPP Release 19 will continue to embed AI/ML to make 5G-Advanced more intelligent
- From ML to LLMs for RAN optimization
- Generative AI is a quantum leap in the realization of autonomous RAN
- Aira Technologies adopted GenAI to augment its existing software, accelerate RAN automation, and deliver its autonomous RAN vision
THE ECOSYSTEM IS AT "AI FOR RAN," TIME TO GEAR UP FOR AI-RAN!
- LEADING RAN VENDORS STAY BUSY WITH AI FOR RAN WHILE CONDUCTING AI-RAN R&D
- Ericsson is incorporating AI in its RAN portfolio across L1, L2, L3
- Nokia proposes a neural transmitter, demapper and receiver
- Samsung is conducting research on AI-RAN, looking at how to integrate AI into RAN
- AS POINTED EARLIER, HUAWEI AND ZTE ARE FOCUSED ON AUTONOMOUS NETWORK, WHICH INCLUDES GENAI, LLM AND AN AIRAN COMPONENT
- Huawei's IntelligentRAN uses Generative AI, we don't know if it is DeepSeek
- ZTE provides its uSmartNets platform that includes an AI for RAN component