By Daniel Okafor, Telecom Gazette

Nokia has told operators that long-term 6G cannot be built without Nvidia. The Finnish vendor's position frames Nvidia's GPU and AI platforms as core infrastructure for the IMT-2030 generation.

This is a sharp break from prior 6G commentary. A radio access network supplier has publicly named a single silicon partner as essential. Nokia executives argued that AI-native processing has moved from optional to mandatory for next-generation radio systems.

The context is the AI-RAN push that defined 2024 and 2025 industry events. Nvidia has spent two years positioning its Aerial software stack for vRAN and 5G/6G workloads. Operators and vendors have run GPU-based baseband trials across multiple markets in that window.

What does Nokia's claim change for practitioners?

RAN architects face a harder procurement question now. Any 6G planning cycle must include Nvidia's roadmap in its dependency map. Tracking CUDA software releases and GPU supply cycles becomes part of standard RAN planning work.

The cost envelope for 6G looks heavier as a result. GPU-based baseband carries a different capex profile versus traditional dedicated silicon. Greenfield 6G builds will absorb that delta first.

Brownfield operators face a longer migration. Network teams should also expect toolchain overlap. Nvidia's CUDA ecosystem covers data center, edge and RAN. That gives operators one AI compute fabric, but it concentrates lock-in with one supplier.

When does this matter on the calendar?

ITU-R has set 2030 as the IMT-2030 commercial availability horizon. The framework recommendation is expected later this decade. 3GPP Release 21 is expected to open formal 6G study work in 2025-2026. Vendor proposals will land inside that release window.

What is at stake for the vendor community?

Nokia's public stake in Nvidia creates a dual dependency. Any Nvidia supply constraint, export-control action or pricing shift now flows directly into Nokia's 6G delivery promise. Operators will price that risk into contract negotiations and indemnity clauses through 2026-2027.

The competing view comes from peers in the RAN market. Ericsson continues to position general-purpose server silicon and programmable hardware as sufficient for next-generation workloads. Huawei and Samsung signaled AI ambitions without naming a single chip vendor as essential.

That divergence gives operators choice architecture but no clean answer. Trial RFPs through 2026 and 2027 will likely require explicit AI-acceleration plans. Vendors without a credible GPU or AI-silicon story will need a written alternative to satisfy procurement offices.

What is the standards-side wildcard?

3GPP Release 21 study items have not yet frozen AI and machine-learning interface boundaries for the radio interface. A binding decision on AI-native radio processing could tighten or loosen the Nvidia dependency. That depends on whether distributed edge inference claims end up in the release.

Geopolitics sit in the background of this story. Export controls on advanced GPUs shape operator access to the latest silicon in regulated markets. Any change in those controls would reset the AI-RAN cost calculus across regions.

Operators should expect the AI-RAN debate to harden into procurement language. Look for that shift over the next 24 months. The Nokia-Nvidia pairing will draw close scrutiny from rival vendors and from operator procurement offices through the long IMT-2030 cycle.