Telecom & Connectivity | 4 min read

Nokia and NVIDIA Deploy AI-RAN Across Nearly 20 Live Operator Networks

Nearly 20 global telecom operators are now running AI natively inside their 5G networks, completing the shift from AI-assisted to AI-native infrastructure.

Hector Herrera
Hector Herrera
A network operations center featuring field, phone, related to Nokia and a chip manufacturer Deploy AI-RAN Across Nearly 20
Why this matters Nearly 20 global telecom operators are now running AI natively inside their 5G networks, completing the shift from AI-assisted to AI-native infrastructure.

Nokia and NVIDIA Deploy AI-RAN Across Nearly 20 Live Operator Networks

By Hector Herrera | November 17, 2026

Nearly 20 global telecom operators are now running artificial intelligence natively inside their 5G networks — not as a management overlay, but as the infrastructure itself. Nokia and NVIDIA announced November 16 that operators including T-Mobile, Deutsche Telekom, BT, Elisa, and Orange have announced or completed AI-RAN deployments built on Nokia's anyRAN software and NVIDIA's Aerial RAN Computer. T-Mobile completed the world's first AI-RAN call on a live U.S. commercial network; Elisa completed Europe's first field trial running voice calls, mobile data, and vision AI simultaneously on the same 5G spectrum. The milestone marks a structural shift in how mobile networks are built and operated.

The difference between AI-assisted and AI-native matters more than the terminology suggests. Most operators today use AI as a tool — it helps engineers understand traffic patterns, predict failures, and tune configuration. The engineers still push the buttons. AI-RAN eliminates that human step. The network itself detects conditions, models the optimal response, and executes in milliseconds, without a ticket being opened.

What AI-RAN Actually Is

Traditional radio access networks (RAN) — the hardware and software that connects your phone to the mobile network — run on fixed rules. When traffic spikes, those rules cannot adapt fast enough. Engineers manually adjust configuration. Spectrum gets allocated conservatively to avoid congestion, leaving capacity unused.

Nokia's anyRAN software replaces that fixed-rule logic with continuously learning AI models. NVIDIA's Aerial RAN Computer (ARC) provides the GPU acceleration to run AI inference fast enough to keep pace with live network traffic. Together, they embed intelligence at every node of the network rather than routing all decisions to a central brain.

The result: a network that dynamically reallocates spectrum, adjusts transmission power, and routes traffic in real time — continuously, without manual intervention.

The Operators and the Milestones

T-Mobile holds the most significant milestone in this round: the first AI-RAN voice call completed on a live U.S. commercial network, carrying actual subscriber traffic. Lab demonstrations and controlled field trials are common; live commercial deployments with paying subscribers are not.

Elisa, the Finnish operator, completed Europe's first AI-RAN field trial that simultaneously ran three distinct workloads on a single 5G network: voice calls, mobile broadband data, and computer vision inference. That last capability — edge AI running directly on the cellular network's infrastructure — is the one manufacturers and logistics operators have been anticipating. It means running AI analysis of factory floors, loading docks, or traffic intersections without routing video streams to a distant data center.

Other operators that have announced or completed deployments include BT (UK), Deutsche Telekom (Germany), and Orange (France), along with roughly 14 others not individually named in the announcement.

What This Unlocks for Enterprises

The Elisa multi-workload trial points to the most commercially interesting application of AI-RAN: network-native edge AI.

Today, if a factory wants to run computer vision for quality control, it needs dedicated edge computing hardware, its own network connection, and significant IT infrastructure to manage. On an AI-RAN network, the same 5G spectrum used for employee phones can dynamically allocate capacity to run vision inference workloads when traffic allows.

For enterprise customers:

  • Latency reduction — AI inference at the network edge is faster than routing data to a centralized cloud
  • Data sovereignty — video and sensor streams never leave the local facility
  • Operational simplicity — no separate edge computing stack to provision and maintain

For operators:

  • Energy savings — AI-optimized power output reduces energy consumption compared to broadcasting at maximum strength
  • Better spectrum utilization — dynamic allocation captures capacity that static rules leave unused
  • New revenue lines — edge AI-as-a-service on existing infrastructure, without new capital spending

The Competitive Landscape

Nokia and NVIDIA are not running this race alone. Ericsson has been developing its own AI-native RAN architecture, and Samsung has partnerships with AWS and Google on network AI. But the Nokia-NVIDIA partnership has now demonstrated the largest deployment footprint by number of live commercial operators.

The timing also intersects with a broader consolidation across AI. OpenAI simultaneously folded its separate robotics team into core research this week, arguing that general-purpose AI models can now handle specialized hardware domains without domain-specific training. The implication for telecom is similar: foundation model capabilities are absorbing what were once considered distinct engineering disciplines. The competitive advantage in 5G is shifting from spectrum holdings and tower density toward AI model quality and inference speed.

What to Watch

The next milestones are commercial service launches — operators charging enterprises for edge AI capabilities running on AI-RAN infrastructure. T-Mobile and Elisa are the most likely first movers. Watch for enterprise pricing announcements and pilot agreements in Q1 2027, and for the remaining 15-plus operators in this cohort to accelerate deployments before the first-mover window closes.

Key Takeaways

  • ✓ For enterprise customers:
  • ✓ Operational simplicity
  • ✓ Better spectrum utilization

Did this help you understand AI better?

Your feedback helps us write more useful content.

Hector Herrera

Written by

Hector Herrera

Hector Herrera is an AI systems architect in Houston and founder of Hex AI Systems. He designs and runs AI systems in production and writes daily about how AI is reshaping business, government and everyday life. 20+ years building for the web. Houston, TX.

More from Hector →

Get tomorrow's AI briefing

Join readers who start their day with NexChron. Free, daily, no spam.

More from NexChron