Nokia Refines AI-RAN Messaging: Merchant Silicon Confidence and Software-First Evolution Signal Maturing Open RAN Strategy

Executive Summary

On July 17, 2026, Dell’Oro Group published an in-depth analyst note examining Nokia’s latest AI-RAN positioning. The update reflects refined messaging rather than a wholesale strategy pivot, with Nokia demonstrating greater conviction in its merchant-silicon and GPU-centric approach. For Open RAN operators evaluating migration paths from disaggregated vRAN to AI-native architectures, the details matter: simplified deployment profiles, emphasis on software-driven innovation, and early quantification of spectral-efficiency gains that could influence RIC and xApp/rApp roadmaps. This article unpacks the technical and architectural implications.

Refined Migration Story: From Architecture Taxonomy to Performance Profiles

Nokia’s prior framework (shared extensively through 2025) delineated purpose-built D-RAN, D-RAN vRAN, and C-RAN vRAN options in its collaboration with NVIDIA. The July 2026 messaging collapses these into three flexible profiles:

• Installed-base leverage (brownfield optimization of existing Nokia hardware with AI overlays)
• High-capacity AI-RAN (dense urban or high-traffic cells emphasizing GPU-accelerated PHY and MAC layers)
• Cloud-native AI-RAN (hyperscale or edge-cloud deployments leveraging O-RAN interfaces for multi-vendor RIC integration)

This shift prioritizes operator pain points—capacity headroom, total cost of ownership, and incremental AI insertion—over rigid architectural labels. In O-RAN terms, the profiles map cleanly onto near-RT RIC (for real-time control loops) and non-RT RIC (for policy and model training) without mandating a single fronthaul or midhaul topology. Operators running O-RAN Alliance-compliant E2 interfaces can therefore layer Nokia’s AI capabilities atop existing O-DU/O-CU instances while preserving multi-vendor RU choices.

Growing Conviction in Merchant Silicon and GPU-Based Baseband

Perhaps the most technically salient takeaway is Nokia’s explicit endorsement of merchant silicon (primarily NVIDIA GPUs) as the primary vehicle for future software innovation. While the vendor continues to support purpose-built ASICs for legacy installed bases, management now indicates that the majority of new feature development—PHY acceleration, AI inference engines, and advanced beamforming algorithms—will target the GPU track.

From an architecture perspective this aligns with the broader AI-RAN movement: GPU-accelerated OCUDU stacks (as demonstrated by multiple ecosystem players) enable unified compute for both RAN workloads and AI workloads. Nokia’s stance reduces hedging around 6G R&D custom-silicon paths and positions the platform as inherently software-defined. For Open RAN integrators this is consequential; GPU-based O-DUs facilitate tighter integration with the AI-RAN Alliance reference architectures and allow third-party dApps to execute on the same silicon fabric as the baseband.

AI-on-RAN Beyond Optimization: Sensing, Positioning, and Third-Party Applications

Nokia is expanding the value narrative from “AI-for-RAN” (energy optimization, interference management, mobility robustness) to “AI-on-RAN.” The platform is positioned to host sensing, positioning/location services, and third-party software applications. In protocol terms this implies exposing additional APIs—potentially extending O-RAN’s E2 and A1 interfaces or leveraging new service models within the non-RT RIC—for external ML models to consume raw or processed radio telemetry.

For example, an operator could deploy a third-party xApp that fuses RAN channel-state information with environmental sensors to deliver sub-meter positioning accuracy, all while running on the same GPU cluster handling L1/L2 processing. This multi-tenant model mirrors cloud-native design principles already familiar to Open RAN operators and lowers the barrier for ecosystem monetization beyond traditional RAN vendors.

The 2× Spectral Efficiency Claim: Context and Caveats for Operators

Nokia highlights up to 2× spectral-efficiency improvements in targeted scenarios. While the headline number is aggressive relative to other vendor disclosures, Dell’Oro cautions that direct cross-vendor comparisons remain difficult without standardized benchmarks. Spectral efficiency is highly scenario-dependent—urban macro vs. indoor small cell, TDD vs. FDD, single-user vs. MU-MIMO loads.

Engineers evaluating the claim should request:

• Baseline configuration (existing Nokia 5G RAN vs. AI-enhanced)
• Reference channel models and traffic mixes
• Measurement methodology (e.g., Ookla-style drive tests or controlled lab setups)

Until such data matures, the figure serves as a directional signal that AI-native techniques (learned beamforming, predictive scheduling, joint PHY-MAC optimization) can deliver material gains beyond incremental 3GPP Release 18/19 features. Open RAN operators with mature RIC deployments are well-placed to validate these gains through controlled rApp trials.

Alignment with Broader Market Forecasts

Dell’Oro’s note explicitly ties Nokia’s trajectory to its own $35 B cumulative AI-RAN revenue projection through 2030 and an upward revision of GPU-RAN outlook exceeding$1 B by the same horizon. Nokia’s emphasis on AI-for-RAN as the primary near-term value driver, coupled with merchant-silicon confidence, reinforces the view that adoption will accelerate as operators prepare software-centric, AI-native platforms for 6G.

Strategic Context: RAN Market Share Recovery

Nokia has lost approximately 10 percentage points of RAN market share over the past decade. AI-RAN is framed internally as more than a technology roadmap—it is a strategic lever to regain scale in a concentrated market. By delivering measurable performance differentiation through software and flexible deployment, Nokia aims to strengthen its position with both greenfield Open RAN operators and brownfield customers seeking incremental intelligence.

Implications for Open RAN Operators and the Ecosystem

1. Migration Planning: Operators can now map existing O-RAN topologies to Nokia’s three profiles with less architectural friction.
2. RIC and App Strategy: Expect richer xApp/rApp portfolios that exploit GPU headroom for non-traditional RAN functions (sensing, positioning).
3. Vendor Selection Criteria: GPU compatibility, E2/A1 extensibility, and third-party application support become higher-priority evaluation factors.
4. Benchmarking Needs: The industry would benefit from standardized AI-RAN efficiency metrics—perhaps coordinated through the O-RAN Alliance or AI-RAN Alliance—to enable apples-to-apples comparisons.

Conclusion

Nokia’s July 2026 update is best read as a confidence signal rather than a disruptive reveal. By simplifying deployment narratives, doubling down on merchant silicon, and articulating concrete AI-on-RAN use cases, the vendor is executing against a multi-year plan while acknowledging operator realities. For technical audiences building or operating Open RAN networks, the message is clear: AI-native capabilities are transitioning from proof-of-concept to production-grade software platforms, and the architectural choices made today will determine how easily those capabilities integrate with existing disaggregated stacks.

Sources

• https://www.delloro.com/nokia-ai-ran-confidence-is-growing/
• https://www.delloro.com/news/ai-ran-to-reach-35-b-over-next-five-years/