Shashi Bellamkonda  ·  March 1, 2026  ·  8 min read

AI Strategy  ·  Telecom Infrastructure  ·  5G  ·  Enterprise AI

I was not in Barcelona this week for Mobile World Congress 2026, but I have been following the announcements closely — and what Nokia revealed on March 1, the opening day of the congress, is worth explaining carefully for technology leaders. This is not a single vendor announcement. It is a picture of an emerging global operator coalition aligned around a specific vision of what the next generation of wireless networks is for.

Panel at Nokia's MWC 2026 session, Barcelona, March 2026. Named on the backdrop left to right: Justin Hotard, President and Chief Executive Officer, Nokia | Sami Komulainen, Chief Operating Officer and Executive Vice President Technology and Operations, Elisa | Vikram Sinha, President Director and Chief Executive Officer, Indosat Ooredoo Hutchison | Ronnie Vasishta, Senior Vice President Telecom, NVIDIA | Dr. John Saw, President of Technology and Chief Technology Officer, T-Mobile. A sixth participant — unlabelled on the backdrop — appears to be a moderator. I was covering this remotely.

The panel Nokia assembled on stage says more than any press release. Five organisations seated together across three continents: Nokia, Elisa from Finland, Indosat Ooredoo Hutchison from Indonesia, NVIDIA, and T-Mobile from the United States. A technology announcement with one carrier partner can be read as a bilateral arrangement. A panel spanning a Finnish operator, a major Southeast Asian carrier, and a US network leader is commercial momentum — not a single-market proof of concept.

What Nokia Announced at MWC 2026

Nokia's March 1 announcement detailed three categories of progress: functional tests completed with active operator partners, new traction engagements with additional operators, and live demonstrations on the show floor.

Operators With Functional Tests Complete

Nokia confirmed that its anyRAN software on the NVIDIA Graphics Processing Unit (GPU)-accelerated AI-Radio Access Network (AI-RAN) platform has completed successful functional tests with T-Mobile in the United States, Indosat Ooredoo Hutchison in Indonesia, and SoftBank Corp. in Japan.

The Indosat milestone is the most striking. Nokia and Indosat Ooredoo Hutchison achieved Southeast Asia's first AI-RAN-powered Layer 3 5G call — demonstrated live at MWC 2026 in Barcelona. The call used Indosat's open, cloud-native network combined with Nokia's AirScale remote radio heads and RAN software accelerated by NVIDIA GPUs. What this proves is that AI and RAN workloads can run simultaneously on shared GPU infrastructure in a live operator environment — not a lab. Operators do not need to build a separate AI compute layer alongside their radio network. The same GPU running the radio can also run AI inference tasks, which changes the cost model significantly.

SoftBank's demonstration extended this idea further. SoftBank showed how spare AI-RAN compute capacity — GPU cycles not currently used by the radio — can be identified by SoftBank's own AITRAS Orchestrator system and reassigned to run third-party AI tasks. A carrier could monetise unused compute in its base stations by offering AI inference capacity to enterprises. The Radio Access Network, traditionally a cost centre, starts generating revenue beyond connectivity.

Operators in Active AI-RAN Traction

Beyond the functional test partners, Nokia announced AI-RAN traction with BT Group in the United Kingdom, Elisa in Finland, NTT DOCOMO in Japan, and Vodafone Group across Europe and Africa. Elisa's Chief Operating Officer Sami Komulainen, who joined the panel at MWC 2026, described the direction clearly: AI-RAN is a key enabler not just for today's network performance improvements but for the agentic, robotic, and physical AI applications that will define the next decade of connectivity.

The Ecosystem Behind the Platform

Nokia also expanded its infrastructure partner ecosystem at MWC 2026. In addition to Dell Technologies, which provides PowerEdge servers for Nokia's AI-RAN solution, Nokia added Quanta Cloud Technology and SuperMicro as hardware partners, and Red Hat OpenShift as the cloud-native orchestration layer enabling operators to deploy AI and RAN workloads across hybrid cloud environments. Nokia's AI-RAN platform is not proprietary top-to-bottom — operators can choose from an open ecosystem of server and software partners.

Why This Matters: The Latency Problem of AI at the Edge

Today, when a user on a mobile device sends a query to a cloud AI system, the response travels through a wireless network designed for human communications — voice calls, messages, video. The latency tolerances of that network were engineered for human perception, not for machine-speed AI inference.

Physical AI changes that constraint entirely. An autonomous vehicle navigating a busy intersection cannot wait for a cloud round-trip on current network latency. A manufacturing robot receiving real-time instructions needs a connection that behaves more like memory than a telephone line. Augmented reality glasses overlaying AI-generated information on the physical world require a pipeline between device and cloud that essentially disappears.

Nokia's AI-RAN approach addresses this by making the Radio Access Network itself AI-native — not simply a carrier of AI traffic, but a compute platform that can run AI inference at the edge, adapt in real time, and optimise spectrum allocation using machine learning rather than static protocols. The Indosat and SoftBank demonstrations at MWC 2026 are the first live evidence that this is not theoretical.

How Nokia Compares to Ericsson and Samsung

MWC 2026 made clear that Nokia is not alone in pursuing AI-native networks. Ericsson and Samsung are both making serious investments in this direction, each with a different architectural bet. Which approach wins at commercial scale will shape carrier infrastructure decisions for the next decade.

Ericsson: The Custom Silicon Path

Ericsson is a founding member of the AI-RAN Alliance and a partner in the T-Mobile AI-RAN Innovation Center in Bellevue, Washington. It launched a new wave of AI-ready radios, antennas, and AI-RAN software at its pre-MWC 2026 event in London.

Where Nokia has chosen to build on NVIDIA's GPU platform, Ericsson is designing purpose-built ASICs and working with Intel-based silicon. Ericsson's argument is that purpose-built silicon delivers performance and energy efficiency that general-purpose GPUs cannot match for all use cases — and that silicon independence protects operators from vendor lock-in. Ericsson has demonstrated AI features at Layer 1 of the radio stack without GPU dependency, using Intel's advanced matrix extensions for AI inference directly on the processor.

This is a real engineering debate, not a marketing distinction. Both approaches are credible, and Ericsson's largest-installed-base position outside China gives its argument weight that Nokia cannot dismiss. Operators who have spent a decade building on Ericsson infrastructure will not switch platforms on the basis of a press release.

Samsung: The Flexible Multi-Compute Approach

Samsung Networks brings an end-to-end portfolio spanning chipsets, radios, and network core, along with deep experience in large-scale virtualized RAN deployments. At MWC 2026 in Barcelona, Samsung demonstrated AI beamforming — using AI algorithms to improve signal directionality and spectral efficiency — running on NVIDIA's Aerial platform, showing measurable throughput gains from existing spectrum.

Samsung's approach to compute is deliberately flexible, working with both NVIDIA CUDA platforms and Intel Xeon 6 processors. In December 2025, Samsung and KT Corporation validated AI-RAN on a live commercial network in South Korea — the first such live-network validation in that market. Samsung's position is that the total cost of ownership for GPU-based base stations is still being evaluated by operators, and it intends to serve them whichever way those economics resolve.

A Word on Huawei: The Market Reality

Any honest discussion of the Radio Access Network market has to acknowledge that Huawei and Ericsson together hold nearly two-thirds of global RAN market share, and that Huawei is the global market leader by revenue. Huawei is running its own parallel AI-native network strategy under the 5.5G and 5G-Advanced banner, with significant investments in AI-integrated radio solutions.

Huawei is excluded from US networks under federal restrictions, is not part of the Western AI-RAN Alliance ecosystem, and is not partnering with NVIDIA on the AI-RAN platform described in this post. The competitive dynamic between Western vendors building toward open, AI-native 6G and Huawei's proprietary approach is a meaningful geopolitical and infrastructure story — but a separate one from the operator coalition Nokia assembled in Barcelona this week.

What Technology Leaders Should Take From This

For Chief Information Officers and Chief Technology Officers, three things stand out from this week's announcements.

The operator geography tells a story. Nokia's MWC 2026 list spans the United States, Japan, Indonesia, Finland, and the United Kingdom. This is not a US 6G story. Operators in high-growth emerging markets are joining from the start — and that means the AI-RAN rollout will reach enterprise customers in those markets earlier than most people expect. For organisations operating across these regions, the relevant question is which carrier they use and which RAN vendor that carrier has aligned with. Those choices will determine what AI-capable edge infrastructure is available to enterprise applications over the next five years.

The architecture question is not academic. The three-way race between Nokia's GPU-native platform, Ericsson's custom silicon path, and Samsung's flexible multi-compute approach will shape what carriers can offer enterprises — in latency, throughput, and edge compute. Technology leaders should be asking their carrier account managers directly: which AI-RAN direction is your network taking, and what does that unlock for our applications?

The timeline is already in motion. Nokia and Indosat made a live 5G call on AI-RAN infrastructure at MWC 2026 this week — not in a lab, on a show floor in Barcelona. T-Mobile trials begin in 2026. Samsung validated AI-RAN on live networks in Korea in December 2025. The infrastructure that will carry physical AI — autonomous vehicles, factory robots, agricultural sensors, augmented reality — is being tested now, and the vendors building it are working to commercial schedules.

That shift from connecting devices to connecting intelligence is not a keynote theme. It is a design brief for the next decade of infrastructure — and the building has already started.

Sources
Nokia MWC 2026 announcement, March 1, 2026  ·  NVIDIA and Nokia strategic partnership press release, October 2025  ·  NVIDIA 6G Open Platform commitment, MWC 2026 Barcelona, March 1, 2026  ·  Samsung MWC 2026 AI-RAN demonstration, March 2026  ·  Samsung and KT Corporation AI-RAN commercial validation, December 2025  ·  Ericsson AI-RAN London event, February 2026  ·  T-Mobile AI-RAN Innovation Center announcement, September 2024  ·  Dell'Oro Group RAN market data, 2025  ·  Omdia RAN market tracker, 2025

About Shashi Bellamkonda

Loading bio...

LinkedIn Full Bio