As the telecommunications landscape evolves, the integration of Internet of Things (IoT) technology within Artificial Intelligence in Radio Access Networks (AI-RAN) is poised to redefine connectivity and operational efficiency.
Source: NVIDIA, 2025
With a projected market value of $6.18 billion by 2032, according to ABI Research, AI-RAN represents a significant opportunity for enhancing network performance through intelligent automation. However, the pathway to widespread adoption remains fraught with challenges.
Vendors are enthusiastic about AI-RAN, evidenced by the rapid growth of the AI-RAN Alliance from 11 to over 80 members in just a year. This surge underscores a collective recognition of AI's potential to optimise radio access networks. Yet, operators remain cautious, with only eight members currently engaging.
“The real growth in AI-RAN will only come when performance benchmarks are validated in the field,” states Samuel Bowling, research analyst at ABI Research. Operators are seeking tangible evidence of cost savings and enhanced performance before committing to these advanced technologies.
The IoT aspect of AI-RAN lies in its ability to support a multitude of connected devices seamlessly. As networks become increasingly dense with IoT devices—ranging from smart meters to autonomous vehicles—the demand for efficient data processing and low-latency communication grows. AI-RAN's potential to manage these connections intelligently could transform how operators handle the surge in IoT traffic.
Current pilot programs, like SoftBank’s AITRAS and NVIDIA’s initiatives in Shenzhen, highlight the capabilities of AI-RAN in real-world scenarios. However, these projects still require third-party validation to build operator confidence.
“Operators need more than hype; they need transparent, validated evidence that AI-RAN delivers real-world performance and long-term value,” emphasises Bowling. Without this validation, the shift from trial to commercial deployment remains elusive.
The debate surrounding compute architecture is also pivotal. GPU-based platforms may offer high performance for complex tasks such as massive MIMO and beamforming, yet concerns over energy consumption and vendor lock-in are significant. In contrast, CPU-based solutions provide a more energy-efficient route, aligning with operators’ needs for flexibility and cost-effectiveness.
To bridge the gap between vendor enthusiasm and operator skepticism, the AI-RAN Alliance must focus on facilitating real pilot deployments. Standardised benchmarks comparing GPU, CPU, and custom silicon solutions will be essential for proving the viability of AI-RAN as a backbone for IoT-driven networks.
Demonstrating effectiveness across various environments—urban, rural, and remote—will be crucial for building trust and advancing toward full-scale implementation by 2030.
