NVIDIA's $1B Bet on Nokia: What It Means for 6G

In late 2025, NVIDIA and Nokia announced a joint development agreement worth approximately $1 billion over five years, focused on building AI-native radio access network (RAN) platforms for 6G. The deal is not the largest in telecom's history, but it may be the most strategically significant in a decade. It signals that the GPU — not the custom ASIC — may become the dominant compute substrate for next-generation base stations.

Understanding what this means requires looking beyond the dollar figure and examining what, exactly, the two companies are building together, why both parties need each other, and what this architecture implies for the broader 6G supply chain. This analysis draws on 7G Network's coverage of telecom infrastructure evolution and AI-native networking architectures.

What the Deal Actually Involves

The partnership centers on Nokia's AirScale radio platform and NVIDIA's Aerial SDK — a software framework that enables 5G and 6G signal processing on NVIDIA's GPU and DPU hardware. The joint development work targets three areas:

• AI-native channel estimation and beamforming: replacing deterministic algorithms with neural networks that run on GPU hardware, enabling real-time adaptation to channel conditions that classical algorithms cannot handle efficiently.
• Open RAN integration: deploying NVIDIA Aerial on O-RAN compliant disaggregated base station architectures, allowing Nokia's radio hardware to run alongside third-party software on GPU-accelerated compute platforms.
• 6G research prototyping: using Nokia's Bell Labs facilities and NVIDIA's DGX infrastructure to prototype sub-THz radio systems with AI-defined air interfaces — the foundational research that will inform Nokia's 6G product roadmap.

The NVIDIA-Nokia $1 billion deal targets three areas: AI-native channel estimation and beamforming on GPUs, Open RAN integration via NVIDIA Aerial SDK, and 6G sub-THz prototyping at Nokia Bell Labs using NVIDIA DGX infrastructure.

Why NVIDIA Wants to Own the RAN

Jensen Huang has described telecom infrastructure as "the next data center" on multiple occasions. The framing is precise: base stations, from NVIDIA's perspective, are compute platforms that happen to include radio hardware. The signal processing workloads — channel estimation, MIMO decoding, beamforming, scheduling — are mathematically dense operations that GPUs are well-suited to accelerate.

The traditional RAN market was dominated by custom silicon: Qualcomm's FSM (formerly known as the ASIC-based 5G baseband), Ericsson's in-house silicon, Nokia's ReefShark chipset family. These chips are highly optimized for specific workloads but inflexible — they do exactly what they were designed to do at launch, and adapting them to new algorithms requires new silicon.

The move toward AI-native RAN creates an opening for general-purpose AI accelerators. If the channel estimation function is a neural network rather than a fixed algorithm, the question "which chip runs it" becomes a competitive market rather than a captive one. NVIDIA is betting that the answer will increasingly be "an NVIDIA GPU," according to Jensen Huang's keynote at MWC 2025, for the same reason that cloud training and inference workloads consolidated on NVIDIA hardware: ecosystem, software toolchain, and raw compute density.

NVIDIA views telecom base stations as compute platforms. AI-native RAN replaces fixed DSP algorithms with neural networks, creating a market for general-purpose GPU accelerators in infrastructure that was previously dominated by custom ASICs from Qualcomm, Ericsson, and Nokia.

Why Nokia Needs NVIDIA

Nokia's position in the RAN market has been strong but not dominant. It competes with Ericsson (slightly larger in RAN revenue), Huawei (excluded from Western markets but strong in Asia), and an emerging set of Open RAN vendors (Mavenir, Parallel Wireless, Rakuten Symphony). Nokia's differentiation is in software-defined radio and its Bell Labs research heritage.

The 6G transition creates both a risk and an opportunity for Nokia. The risk: if 6G RAN runs on commodity GPU hardware with open software interfaces, Nokia's custom silicon advantage disappears. The opportunity: if Nokia can lock in the dominant software stack for AI-native RAN before the standard is finalized, it can win the 6G cycle on software, not silicon.

The NVIDIA partnership hedges both sides. Nokia gains access to the world's leading AI compute platform and an extremely valuable co-development relationship with the GPU ecosystem. NVIDIA gains Nokia's radio physics expertise, its Bell Labs research infrastructure, and a credible path to deployment in major operator networks. According to Dell'Oro Group (2025), Nokia held approximately 27% of the global RAN market share entering the deal.

Nokia needs the NVIDIA partnership because 6G may shift RAN value from custom silicon to software. If Nokia locks in the dominant AI-native RAN software stack before 6G standardization, it can win the 6G cycle on software rather than hardware differentiation.

The Open RAN Angle

This deal is also an Open RAN story. The O-RAN Alliance's disaggregated architecture — which splits the traditional base station into radio unit (RU), distributed unit (DU), and centralized unit (CU) running on general-purpose hardware — is precisely what makes "GPU-accelerated base stations" a coherent product category.

In a traditional integrated RAN, Nokia or Ericsson designs the entire stack from radio to software, running on proprietary hardware. In an Open RAN deployment, the DU and CU can run on any x86 or ARM server, and increasingly on GPU-accelerated servers. NVIDIA's Aerial SDK is explicitly designed for this environment.

The Nokia-NVIDIA deal effectively bets that Open RAN will win in the 6G era, and that the compute layer of the DU — traditionally the bottleneck for real-time signal processing performance — will be a GPU workload. According to the O-RAN Alliance (2024), over 60 operators worldwide have committed to Open RAN deployments. If that bet is correct, the deal positions both companies at the center of the 6G supply chain.

Open RAN's disaggregated architecture — splitting base stations into RU, DU, and CU on general-purpose hardware — is what makes GPU-accelerated base stations viable. The Nokia-NVIDIA deal bets that the 6G DU compute layer will be a GPU workload running NVIDIA Aerial SDK.

Competitive Implications

Ericsson has its own AI RAN program and has been developing AI-native features for 5G Advanced (Release 18) independently. Its response to the Nokia-NVIDIA deal was notably measured — Ericsson executives acknowledged AI RAN's importance while emphasizing their own in-house research capabilities.

The wildcard is Qualcomm. Its X100 5G RAN platform and partnership with several Open RAN vendors puts it in a similar position to NVIDIA — selling compute into the disaggregated base station market. Qualcomm's modem-to-infrastructure play has historically been cautious, but if GPU-accelerated AI RAN gains traction, Qualcomm's custom AI accelerators (already deployed in data centers under the Cloud AI brand) could enter this market aggressively.

For hyperscalers, the deal is a signal. Amazon AWS, Microsoft Azure, and Google Cloud have all invested in Open RAN and private 5G infrastructure. If the 6G base station is essentially a specialized GPU server running AI workloads, the question "should we build this ourselves" becomes more interesting for hyperscalers than it was with traditional base station silicon. The broader context of 5G's economic lessons makes this competitive dynamic even more consequential.

What It Means for 6G Timeline and Architecture

The Nokia-NVIDIA partnership will produce research prototypes over the next 3–4 years that will directly influence Nokia's submissions to 3GPP and ITU-R for 6G standardization. The deal thus has a structural effect on what 6G's air interface will look like — not just what hardware will run it.

If AI-native channel estimation and beamforming are validated at scale in this partnership, the 3GPP working groups will include them in the baseline IMT-2030 specification. If they are not — if the latency requirements prove incompatible with GPU scheduling overhead, or if the neural network approaches fail to generalize across channel conditions — the deal may produce valuable negative results that redirect the 6G research agenda.

Either outcome advances the field. That is the nature of well-structured research partnerships at the frontier of standardization: the value is in the validated learnings, not just the successful products. According to 3GPP's timeline, IMT-2030 specification work will incorporate validated AI-native RAN research by 2028–2029.

The Nokia-NVIDIA partnership will produce 6G research prototypes over 3–4 years that will directly influence 3GPP and ITU-R submissions for IMT-2030 standardization. The deal has a structural effect on what 6G's air interface will look like, not just what hardware runs it.

The Bigger Picture

The Nokia-NVIDIA deal is one data point in a broader transformation: the convergence of the AI infrastructure industry and the telecom infrastructure industry. For 20 years, these were separate capital pools, separate supply chains, and separate engineering cultures. GPU clusters ran in data centers; base stations ran in cell sites. The two rarely met.

AI-native 6G — and eventually 7G — erases that separation. The base station of 2032 will be, at its core, an AI inference machine that also happens to emit and receive radio signals. The companies that understand both worlds — the radio physics and the AI systems — will define what that machine looks like. The Nokia-NVIDIA bet is that those two companies, together, can be among them.

The NVIDIA-Nokia deal represents the convergence of AI infrastructure and telecom infrastructure — two industries that operated separately for 20 years. The 6G base station of 2032 will be an AI inference machine that emits and receives radio signals, requiring expertise in both radio physics and AI systems.

Sources

1. Nokia — official press release on NVIDIA partnership — deal announcement and technical details
2. NVIDIA Aerial SDK documentation — GPU-accelerated 5G/6G signal processing platform
3. O-RAN Alliance — Open RAN architecture specifications and operator commitments
4. Dell'Oro Group — RAN market share reports — Nokia, Ericsson, and Huawei market positioning
5. 3GPP — Release 18+ specifications — 5G Advanced and 6G standardization roadmap
6. Nokia Bell Labs — 6G research publications — AI-native air interface and sub-THz research