AI-RAN delivers measurable operational ROI for telecommunications operators by moving beyond theoretical promises and providing real energy and spectral efficiency improvements. As revenue growth plateaus and network complexity increases, operators are prioritizing solutions that drive cost savings and automation rather than purely experimental features.
This practical focus aligns closely with Open RAN (O-RAN) principles of openness and disaggregation. Industry research supports this convergence: a Heavy Reading survey found that 30 percent of operators expect AI-RAN and O-RAN to be “very closely linked” within five years, and among operators planning O-RAN deployments by 2025, that figure rises to 47 percent.
For telecom leaders, the central concern is immediate value generation rather than debating long-term architectural philosophies.
Energy optimisation presents the clearest business case for AI-RAN today. Radio access networks are major power consumers, and reducing energy use without degrading user experience is a primary operational challenge.
Vendors and trials demonstrate significant gains. Ericsson reports power savings in the range of 10 to 15 percent across multiple trials and similar improvements in live networks. In a collaboration with Bell Canada using an “AI native link adaptation” capability, trials showed up to a 20 percent increase in downlink throughput and a 10 percent improvement in spectral efficiency versus baseline configurations. For operators, squeezing more capacity from existing spectrum directly improves capital efficiency and delays costly spectrum or hardware upgrades.
Infrastructure strategy: integration over isolation
Creating a dedicated, standalone infrastructure solely for AI workloads is generally a strategic mistake. Most architects advocate running AI workloads on the same cloud-native platforms operators already use for microservices and network functions to avoid fragmentation and simplify operations.
Fatih Nar, Distinguished Chief Architect in the CTO Office at Red Hat, cautions against creating “another platform yet for AI only.” Instead, he recommends enhancing existing platforms with AI capabilities so workloads can move fluidly between on-premises and hybrid cloud environments without creating silos.
This principle applies to hardware choices as well. Kanika Atri, Senior Director of Telecoms Marketing at NVIDIA, addresses lingering concerns about the cost and power consumption of accelerated computing. She argues the industry must discard the blanket assumption that GPUs are too costly or power-hungry for edge deployments.
“We would not be doing this if we cannot meet the baseline stringent requirements for telcos radio rollouts,” Atri says.
Solutions must fit within the site’s cost and power envelope. Atri points to platforms such as ArcPro, which operate in a compact form factor under 300 watts while supporting 5G feature sets, as examples that accelerated computing can meet telco constraints when designed appropriately.
Toward intent-based automation
AI-RAN also enables a move toward intent-based automation—allowing operators to specify what the network should achieve rather than the low-level configuration steps required to reach that outcome.
In practical terms, intent-based systems can rapidly reconfigure the network for high-capacity events, like a major concert, by automatically provisioning high-throughput slices and reallocating resources. This reduces manual engineering overhead and speeds response times for temporary or rapidly changing demands.
However, expectations about full autonomy should be tempered. The move to highly autonomous operations will be gradual. Nar notes that even large cloud providers with vast resources have not attained complete autonomy, and complex access networks—especially radio networks—pose additional challenges.
“You think more complicated access networks such as radio… [will] fully be autonomous? Not likely,” he warns, underscoring that humans will remain essential to handle physical factors such as weather and interference.
The human capital challenge of implementing AI-RAN
Successful AI-RAN adoption demands organizational adjustments. A common mistake is isolating AI and data science teams away from network engineering teams, creating disconnects between model development and operational realities.
“Please do not segregate your AI and data scientists from the domain expert,” Nar advises. He recommends avoiding a detached “Centre of Excellence” and instead fostering a “Community of Practice” that embeds AI expertise within radio access network teams.
Domain knowledge is essential because telco-focused AI models differ substantially from general Large Language Models. Effective models require an understanding of the “language of wireless environments,” including radio frequency propagation, interference patterns, and other physical-layer phenomena that generalist AI practitioners may not fully grasp.
Rather than seeking wholesale infrastructure replacement, operators should pursue pragmatic, incremental steps to realize AI-RAN benefits. Recommended actions include:
- Leverage O-RAN principles: Even partial alignment with O-RAN’s openness and disaggregation supports data portability and easier integration of AI capabilities.
- Start with energy: Prioritize pilots focused on energy management and sleep modes, where ROI is measurable and verifiable.
- Unify platforms: Avoid separate AI silos; run AI workloads alongside existing network functions on a common cloud-native platform to reduce operational friction.
- Integrate teams: Embed data scientists within network engineering units so models reflect the realities of physical networks and operational constraints.
As the industry prepares for 6G, the foundational work undertaken now—establishing robust data pipelines, refining models, and proving commercial viability—will determine which operators successfully transition to highly programmable, high-performance networks.
See also: 6G networks will host AI agents to automate enterprise workflows
Want to learn more about AI and big data from industry leaders? Consider attending industry events such as AI & Big Data Expo, which convenes experts and practitioners to discuss AI, data strategy, and real-world deployments across multiple locations. These events provide practical sessions on deploying AI at the edge, managing data pipelines, and aligning AI initiatives with operational objectives.
Telecoms coverage is produced by TechForge Media. Explore upcoming enterprise technology events and webinars to deepen understanding of AI, cloud-native operations, and the evolving role of AI in telecommunications.