Accelerating AI-RAN from concept to commercialization
As the AI supercycle unfolds, it is creating a major transformation in how the telecommunication industry can capture value from networks. We are moving from providing broadband data plans for predictable, largely human-driven use cases toward supporting AI workloads that depend on reliable token delivery and increasingly deterministic connectivity.
This is why we at Nokia, together with our industry partners, are accelerating the development of AI-RAN from vision to commercialization. It’s clear to us that AI-RAN represents something bigger than a typical generational shift in networks.
At a recent industry discussion we hosted, I was joined by leaders from Elisa and NVIDIA to explore what AI-RAN really means in practice and why collaboration across the ecosystem is essential to make AI-RAN a commercial reality. If you didn’t have a chance to catch our live session, you can access the event recording here.
These are the three key points that came up in our discussion:
- AI-accelerated computing capacity at the base station sites redefines the role of RAN
- AI-RAN is both an efficiency engine and a new revenue platform
- AI-RAN cannot be delivered by one player alone
Let me now go into some more detail on each of these.
AI is redefining the role of RAN
The beauty of AI-RAN is that it truly redefines what a RAN site can be. When we introduce AI-accelerated computing into the base stations, they can become a distributed platform for AI infrastructure while also enabling a significant leap in RAN efficiency. Telecommunication providers are uniquely positioned to capture this shift into AI-driven value chains because they already own a highly distributed footprint of RAN sites.
Imagine the emerging distributed AI grid at the network edge:
- It allows telecommunication providers to leverage a programmable, software-defined RAN architecture to scale innovation at software speed.
- It introduces AI processing at scale closer to where data is generated and consumed.
- It supports real-time enterprise applications and physical AI, which require ultra-low deterministic latency and tightly controlled jitter.
AI-RAN unlocks new efficiencies and revenue streams
Already today, telecommunication vendors have introduced AI-driven enhancements that bring concrete business outcomes in customer networks. The challenge is how to scale the efficiencies across the entire RAN footprint so that the network can dynamically adapt to the new types of traffic profiles that AI brings and, eventually, deliver deterministic connectivity for physical AI. With AI-RAN, we can make all that happen while also enabling telecommunication providers to generate more revenue with the new AI infrastructure.
- The immediate value is a more efficient use of RAN assets, both during peak traffic and when the network has spare capacity. With AI-accelerated computing in the base station and AI-driven algorithms for many functions such as channel estimation, carrier aggregation, multi-user MIMO and other deep receiver / transmitter blocks, we can deliver a significantly higher gain on spectral efficiency.
The future value is in the new monetization streams. While AI model training remains largely centralized in data centers, inference must become increasingly distributed. According to McKinsey (February 2026), demand for AI inference workloads is growing significantly faster than demand for model training, at a growth rate of around 35% CAGR. But the reality is that centralized AI inference alone cannot meet this demand. It’s the network edge, anchored in AI-RAN, where this capacity demand will increasingly land.
AI-RAN requires ecosystem collaboration
AI-RAN marks the moment when networks shift from being carriers of traffic to dynamic platforms for distributed intelligence. At Nokia, we know that one player cannot deliver the needed AI infrastructure alone. That is why we are actively engaging with an ecosystem of leading technology companies.
A great example is our collaboration with Elisa and NVIDIA:
- Nokia brings RAN expertise and software-defined network capabilities.
- NVIDIA contributes AI-accelerated computing platforms with innovation driven by a large customer base beyond telecommunication networks.
- Elisa provides real-world experience and insights from live commercial networks.
Together, we are moving fast to validate AI-RAN in Elisa’s live network. We have already seen concrete results in a joint proof-of-concept, which confirms that we are on the right path to accelerate the journey to distributed AI intelligence and new AI-driven ecosystems at the network edge.
To learn more about how we are taking AI-RAN from concept to commercialization, watch the recording of our event here.
