Nokia's AI-RAN Rollout Has a Calendar Problem, Not a Demand Problem

Nokia has spent this week stacking up announcements that all point in the same direction: its radio software is being wired into AI-capable hardware platforms from multiple partners. What none of them do is put a number on the revenue. That gap is precisely what tomorrow's quarterly report will be measured against.

Bank of America confirmed its buy rating on the Finnish equipment maker on Monday with a price target of 16 euros, according to media reports. The brokerage expects AI-related orders exceeding EUR 2 billion — but flags production capacity as the constraint on growth. That caveat matters: demand alone doesn't answer how quickly Nokia can convert it into booked business. The order expectation is the analyst's projection, not a figure Nokia has reported.

Nokia has scheduled the release of its third-quarter and nine-month 2026 results for 22 October 2026, with an analyst call the same day. Beyond the headline numbers, commentary on the balance between incoming demand and deliverable supply will carry the most weight.

One platform, two workloads

The centrepiece of the week's news is a collaboration with Supermicro to run radio processing and AI applications on shared infrastructure, spanning everything from distributed mobile sites to central network nodes. The partners are pairing Nokia's radio network software with the NVIDIA Aerial RAN Computer and Supermicro servers, allowing conventional mobile network tasks and AI workloads to run side by side. Target use cases include speech and image processing, AI for physical applications, and autonomous network operations.

The commercial logic extends beyond AI-assisted optimisation of the radio itself: Nokia's radio technology becomes part of a platform that can also host computing workloads. No customer contracts or revenue expectations tied to the cooperation were disclosed.

Availability arrives in stages. Supermicro servers for NVIDIA ARC are already generally available. The ARC MGX variant with Vera processors is slated for later in 2026. For ARC-Pro, customer testing is also planned for later in 2026, with general availability in the first half of 2027. Shipping hardware is not the same as widespread commercial deployment by operators — the announced tests and rollout plans sit at different points along that path.

Lab gains, field questions

Separately, Deutsche Telekom, Nokia and NVIDIA reported progress on AI-based radio signal processing. In laboratories in Bonn and Oulu, improved uplink channel estimation delivered typical throughput gains of 10–15%, with larger improvements under more complex radio conditions. The results provide technical evidence for the approach but do not substitute for testing on a live network. A field trial on commercial systems is intended to validate performance under real-world radio conditions, leaving the transferability of the lab gains to everyday operations as a key checkpoint.

Elisa has already put shared use to the test: in a field trial, voice, mobile data and video analytics ran simultaneously on an AI-RAN platform. Distributed deployments and commercial services are expected in 2027. BT plans a live network test in Birmingham the same year. In Japan, a field trial with SoftBank is planned for 2027, deploying Nokia's anyRAN software on SoftBank's AITRAS platform with NVIDIA computing infrastructure under commercial network and traffic conditions.

Automation and a divestment on the side

Nokia and NVIDIA are also working on a sovereign platform for agentic AI, integrating Nemotron models and NVIDIA AI Enterprise. The platform is intended to enable controlled automated actions across radio, core, transport and cloud networks.

Adding to the mix, Nokia agreed roughly a week ago to sell its Enterprise Campus Edge business to AURELIUS. The deal is expected to close in the first half of 2027, subject to customary closing conditions and regulatory approvals. No transaction value was disclosed.

For investors, the through-line is a widening set of technical possibilities without a quantified financial contribution. Operator orders and commercial rollouts will be the yardstick for economic significance. Tomorrow's report has to hold several layers apart at once: the current business, expected AI orders, and longer-dated projects. The question is not simply whether Nokia can capture AI demand — it is how fast capacity can turn that demand into revenue.