Arm Enters vRAN Fray with Kenyi to Challenge Intel in Open RAN and AI-Native Deployments
Introduction
As operators accelerate Open RAN and AI-native RAN initiatives, the underlying compute architecture for virtualized RAN (vRAN) is emerging as a critical battleground. A recent report from Light Reading highlights how Arm, in collaboration with startup Kenyi, is positioning its CPU architecture to deliver a credible alternative to Intel’s long-dominant x86 platform in vRAN deployments. This development arrives amid diverging strategies between major vendors like Ericsson and Nokia on AI-RAN paths, potentially reshaping hardware choices for disaggregated, multi-vendor networks.
The Kenyi-Arm Collaboration Details
Kenyi, founded by a former Qualcomm executive, is working to enable Arm-based processors to run virtualized RAN software stacks with minimal porting effort. According to the report, Ericsson and Nokia have already demonstrated the ability to port RAN software functions from Intel processors to Arm-based CPUs after relatively small modifications. This capability is significant because it lowers the barrier for operators seeking to diversify beyond Intel’s near-monopoly in vRAN, which stood at approximately 99% share in earlier deployments.
The effort focuses on delivering performance parity or better for baseband processing tasks, including those increasingly augmented by AI models for real-time optimization. Arm’s architecture offers advantages in power efficiency, which aligns with operator goals for energy reduction in AI-driven workloads such as predictive traffic management and fault prediction.
Relevance to Open RAN Disaggregation
Open RAN’s emphasis on disaggregation extends to the compute layer, where vRAN on merchant silicon like Arm CPUs supports greater vendor interoperability. Traditional purpose-built hardware is giving way to software running on commercial off-the-shelf (COTS) servers. Arm’s entry could accelerate this by providing an additional CPU option alongside x86 and emerging GPU-accelerated solutions.
This matters for AI-RAN because many intelligent controllers and dApps require flexible, high-performance compute that can handle both legacy 5G processing and emerging AI inference at the edge. Operators deploying multi-vendor Open RAN setups could benefit from Arm’s ecosystem for consistent performance across distributed units (DUs) and centralized units (CUs).
Context of Vendor Divergence on AI-RAN
The Light Reading piece situates the Arm development against the backdrop of Ericsson and Nokia pursuing distinct approaches to AI integration in RAN. Ericsson has emphasized telco-grade AI models running in basebands and radios, with features like AI-native schedulers and beamforming. Nokia, meanwhile, has leaned into GPU-accelerated platforms in partnership with NVIDIA for spectral efficiency gains.
Arm’s vRAN push provides a neutral compute foundation that could support either vendor’s AI-RAN software without locking operators into a single silicon vendor. This flexibility is particularly valuable in Open RAN environments where RIC (RAN Intelligent Controller) applications demand low-latency AI processing alongside traditional protocol stacks.
Implications for Operators and the Supply Chain
For operators, broader CPU options in vRAN reduce supply-chain risks and enable better matching of hardware to workload—Arm for efficiency-sensitive sites, GPUs for heavy AI inference. The report notes that Arm has historically lacked a direct vRAN equivalent to Intel’s offerings, making this Kenyi-enabled development a potential inflection point.
Qualitatively, early porting successes suggest that software maturity on Arm is advancing rapidly, though full production deployments will depend on ecosystem validation, performance benchmarks in live networks, and integration with Open RAN interfaces like O1, A1, and E2.
Challenges and Path Forward
Key hurdles include achieving consistent performance across diverse RAN workloads, ensuring compatibility with existing O-RAN Alliance specifications, and scaling the Kenyi-Arm solution beyond trials. Power and thermal characteristics in dense cell sites will also require rigorous testing.
Nevertheless, this development underscores the maturing infrastructure layer supporting AI-native Open RAN. As operators evaluate paths from disaggregated foundations to intelligent, automated networks, Arm-based vRAN represents another pragmatic option in the toolkit.