Nokia
2026-07-15
Nokia and NVIDIA Launch First Commercial AI-RAN Platform, Replacing ASICs with GPUs
Summary
Key Takeaways
On July 15, 2026, Nokia and NVIDIA announced the first commercial AI-RAN platform, integrating Nokia's anyRAN software with NVIDIA's Aerial platform. The key innovation is replacing traditional ASICs with general-purpose GPUs in base stations. AI-driven radio optimization has achieved >20% spectral efficiency improvement, targeting 50% by 2027 and >100% by 2028. Trials begin end of 2026, commercial in 2027.
Nokia CEO Justin Hotard stated AI-RAN makes networks intelligent, extending AI to the physical world. NVIDIA CEO Jensen Huang called it a generational change for operators, laying foundation for 6G. The platform also enables base stations to act as AI compute providers during idle periods, as demonstrated with SoftBank. Nokia will offer a software subscription model.
This marks a shift from proprietary hardware to a software-defined, AI-native RAN architecture, though specific power and latency metrics remain undisclosed.
Why It Matters
Nokia's AI-RAN partnership with NVIDIA is a defensive move against Ericsson and Huawei, but it cedes control to NVIDIA's CUDA ecosystem. The platform locks operators into NVIDIA GPU supply chains and software stacks, reducing architectural flexibility. The announcement downplays the physical limitations of GPUs in base stations: higher power density compared to ASICs, and potential tail latency issues in real-time signal processing. The software subscription model may increase long-term TCO. Nokia risks becoming a hardware integrator for NVIDIA, losing its software differentiation.
PRO Decision
[Vendors] Ericsson, Huawei should emphasize ASIC efficiency and real-time performance advantages over GPUs. Develop chip-neutral AI-RAN solutions supporting multiple accelerators to counter NVIDIA lock-in. Publish independent benchmarks highlighting power consumption and tail latency trade-offs.
[Enterprises] Operators should conduct zero-trust audits demanding detailed power, latency, and reliability metrics. Assess CUDA lock-in risks, require open APIs and hardware portability. Include contractual guarantees for spectral efficiency gains with penalties for underperformance.
[Investors] Recognize Nokia's move as a defensive pivot amid RAN decline, but dependency on NVIDIA may compress margins. Monitor trial data for actual efficiency and power costs. NVIDIA gains a new telecom foothold, but Nokia's differentiation erodes.
Get 3-5 key AI infrastructure signals weekly →
💬 Comments (0)