AI
Nokia’s AI-RAN Platform: Inside the NVIDIA-Powered Bet on Radio’s Future
Nokia just made a bold claim in the telecom world — that it has launched the industry’s first AI-native radio access network (RAN) platform. Built in partnership with NVIDIA, the launch is being framed as one of the biggest shifts in radio network architecture in decades. But the story behind the announcement is just as interesting as the technology itself: this is as much about Nokia’s business turnaround as it is about spectral efficiency gains.
Here’s a clear breakdown of what Nokia announced, how it stacks up against rival Ericsson’s competing approach, and why the stakes are so high for Nokia specifically.
What Is Nokia’s AI-RAN Platform?
Nokia’s new platform combines its own anyRAN software with NVIDIA’s Aerial system to bring AI-native processing directly into radio network infrastructure. The pitch to mobile operators is simple: squeeze significantly more capacity out of spectrum they already own, without needing to acquire additional airwaves.
Nokia says the platform has already demonstrated spectral efficiency gains of more than 20%, with ambitious internal targets of 50% by 2027 and more than 100% by 2028 — a point at which, by Nokia’s own projections, operators could effectively double the capacity of their existing spectrum.
It’s worth being precise here: the 20% figure is a claimed result, while the 50% and 100%+ targets are forward-looking goals, not achievements. Nokia’s own roadmap has pilot deployments starting toward the end of 2026, with broader commercial availability arriving in 2027.
How Operators Would Actually Buy It
Rather than requiring a wholesale hardware replacement, Nokia is positioning this as a software subscription. Operators get three deployment paths to choose from:
- A GPU-powered plug-in card that fits into existing AirScale radio sites
- A standalone, purpose-built AI-RAN node
- A cloud-server deployment delivered through partner infrastructure
This flexibility is a deliberate move to lower the barrier for operators who don’t want a full network overhaul just to access AI-driven efficiency gains.
Why This Matters So Much for Nokia
It would be easy to read this purely as a product launch, but it’s really a comeback story for one of Nokia’s most troubled business units. Mobile infrastructure has been the toughest challenge facing CEO Justin Hotard since he took the helm in 2025. At the company’s November capital markets day, he acknowledged that the mobile business wasn’t delivering acceptable returns, and it was subsequently folded into a newly created Mobile Infrastructure segment alongside a fresh round of cost cuts.
The NVIDIA partnership — announced in October 2025, backed by a $1 billion investment from NVIDIA for roughly a 3% equity stake — sits right at the heart of that recovery effort. By leaning on NVIDIA’s chips and CUDA software instead of continuing to develop custom silicon in-house, Nokia can redirect resources away from expensive hardware R&D and toward software, which is exactly the pivot Hotard has been pushing for.
Investors appear to be buying the story. Nokia’s stock has re-rated significantly through 2026 on the back of its AI and cloud momentum, and this AI-RAN launch landed just days ahead of its second-quarter earnings report. Industry analysts have projected the cumulative AI-RAN market opportunity could exceed $200 billion by 2030 — a number that helps explain why Nokia is moving so aggressively here.
Is Nokia’s Platform Really “The Industry’s First”?
This is where the story gets more nuanced. In June, Ericsson began commercially selling its own AI-in-RAN software subscription, claiming gains of up to 20% higher downlink throughput and up to 10% better spectral efficiency across more than 15 live deployments. Crucially, Ericsson’s approach runs on operators’ existing baseband silicon — no GPU required — and it’s already generating revenue in the market today.
So when Nokia claims to be “first,” it’s leaning on a narrower definition: the first GPU-accelerated AI-RAN platform specifically, which represents a fundamentally different architecture than layering AI software on top of existing baseband hardware. Both companies’ claims can technically be true at once — they’re just describing different things.
The deeper divide is architectural philosophy. Nokia has tied its radio roadmap tightly to NVIDIA’s hardware, and its own CTO has acknowledged that at least some of the platform’s Layer 1 software is bound to that underlying silicon. Ericsson deliberately took the opposite path, keeping its AI capabilities silicon-independent specifically to avoid that kind of hardware dependency.
Nokia does point to Marvell’s merchant silicon within its broader ecosystem and describes the new platform as Open RAN-compliant. But the actual performance numbers it’s promoting — those spectral efficiency gains — currently run specifically through NVIDIA’s stack, and no directly equivalent alternative exists elsewhere today. In other words, Nokia’s messaging leans on openness while its core engineering leans on NVIDIA dependency — both are true simultaneously.
The Bigger Picture
None of this necessarily makes Nokia’s strategy a mistake. Betting on the dominant AI chip supplier to solve a silicon problem Nokia had struggled to crack alone is a reasonable answer to a genuinely difficult business challenge. And shifting radio infrastructure toward a subscription model gives Nokia a shot at the kind of recurring revenue that traditional hardware refresh cycles never provided.
But it’s important to keep the claims in perspective. The platform isn’t yet commercially available, its headline efficiency numbers are projections still two years away, and a major competitor already beat it to market using a different technical approach. This is a comeback that’s underway — not one that’s finished. And for Nokia, that comeback now runs directly through NVIDIA.
Frequently Asked Questions
1. What is Nokia’s AI-RAN platform?
It’s a new radio access network platform that combines Nokia’s anyRAN software with NVIDIA’s Aerial system to bring AI-native, GPU-accelerated processing into mobile radio networks, aiming to boost spectral efficiency without requiring new spectrum.
2. What spectral efficiency gains is Nokia claiming?
Nokia says it has already achieved more than 20% spectral efficiency gains, with targets of 50% by 2027 and over 100% by 2028. The later figures are goals, not confirmed results.
3. When will Nokia’s AI-RAN platform be commercially available?
Nokia’s roadmap points to pilot deployments toward the end of 2026, with broader commercial availability expected in 2027.
4. How can operators deploy the platform?
Nokia is offering three options: a GPU-powered plug-in card for existing AirScale sites, a standalone AI-RAN node, or a cloud-server deployment delivered through partners — all sold as a software subscription rather than a hardware purchase.
5. What is Nokia’s partnership with NVIDIA?
Announced in October 2025, the partnership includes a $1 billion investment from NVIDIA for roughly a 3% stake in Nokia. It centers on using NVIDIA’s chips and CUDA software as the foundation for Nokia’s AI-native radio strategy.
6. Why is this launch important for Nokia as a company?
Mobile infrastructure has been Nokia’s weakest-performing business, and CEO Justin Hotard has been working to turn it around since 2025. The AI-RAN platform and NVIDIA partnership are central to that recovery strategy, shifting Nokia’s radio business toward software and recurring revenue.
7. Is Nokia’s platform really the industry’s “first” AI-RAN system?
Nokia’s claim refers specifically to being first with a GPU-accelerated AI-RAN platform. Ericsson already began commercially selling a different type of AI-in-RAN software in June that runs on existing baseband silicon without requiring GPUs.
8. How does Nokia’s approach differ from Ericsson’s?
Nokia’s platform depends on NVIDIA’s GPU hardware and software stack, while Ericsson deliberately built its AI-RAN features to be silicon-independent, avoiding reliance on any single hardware vendor.
9. How big is the AI-RAN market opportunity?
Industry analysts, including Omdia, have estimated the cumulative AI-RAN market opportunity could exceed $200 billion by 2030.
10. Is Nokia’s AI-RAN platform Open RAN-compliant?
Nokia describes the platform as Open RAN-compliant and points to merchant silicon from Marvell within its broader ecosystem, though its headline performance claims currently rely specifically on NVIDIA’s technology stack.
