Nokia’s new deal leans on software, not just refreshed radios
In sum – what we know:
- Software over hardware – The AirScale radio upgrade is evolutionary; the real change is AI embedded in how the network tunes, heals, and powers itself.
- From MoU to deployment – A March 2025 MWC memorandum covering AI mobile networks has now moved from proofs of concept to network-wide commercial rollout.
- Energy as the throughline – ReefShark silicon plus AI-driven power algorithms support Taiwan Mobile’s RE100 pledge to hit 100% renewable electricity by 2040 — and cut opex along the way.
Nokia and Taiwan Mobile have signed a new agreement that extends their long-running 5G partnership — and shifts its focus. Rather than another round of standard hardware upgrades, the deal centers on building what both companies call an “AI-native” mobile network, with artificial intelligence embedded across the network lifecycle, from operations and automation to energy management.
The deployment covers Nokia’s latest modular, high-capacity AirScale baseband and radio solutions, including advanced massive MIMO radios and remote radio heads. That builds on earlier deployments of the Habrok 32 and Osprey 32 massive MIMO radios introduced in prior deals, so the hardware story here is evolutionary. What separates this agreement from a traditional 5G expansion is the software layer — AI integrated directly into how the network runs, tunes itself, and manages power, rather than simply refreshed radios on existing towers.
For Nokia, the deal also reinforces a remarkably durable relationship. The Finnish vendor has supplied Taiwan Mobile’s networks from 2G through 5G, often as the sole equipment provider, and this agreement keeps that incumbency intact for the AI era.
AI software, automation, and resilience
The software side of the agreement is where most of the substance lives. Nokia is introducing its Predictive Hardware Analytics service, designed to detect and prevent hardware failures before they affect service, and expanding Taiwan Mobile’s existing MantaRay SON deployment with advanced AI capabilities. The self-organizing network handles automatic parameter tuning — power levels, handovers, load balancing — that would otherwise require manual intervention.
The resilience features follow the same pattern. Dynamic traffic steering reallocates capacity to maintain service continuity when loads shift or parts of the network go down, while self-healing functions detect faults and reconfigure systems autonomously. Whether these capabilities deliver at scale is the real question, of course. Nokia frames them as commercial-grade, but AI-driven network autonomy is still a young discipline across the industry.
That said, this isn’t a cold start. The two companies signed a memorandum of understanding at MWC in March 2025 covering AI mobile networks, and this agreement moves those proofs of concept into broad, network-wide commercial deployment. Part of that earlier work included the Nokia Assurance Center, which combines digital twins with generative AI for real-time anomaly detection — essentially simulating network components to spot problems before they surface for customers.
In a regional market where Ericsson, Huawei, and ZTE are all competing for carrier business, the move positions Taiwan Mobile as an early adopter of AI-driven telecom infrastructure. Whether rival Taiwanese operators pursue comparable transformations with other vendors remains to be seen, but being first out of the gate carries at least some differentiation value.
Infrastructure performance and new services
The next-generation baseband units at the heart of the deployment deliver higher capacity and — notably — better uplink performance. That uplink emphasis is deliberate. AI applications generate meaningful upstream traffic, a reversal of the download-heavy patterns mobile networks were historically built around, and Nokia is clearly designing for it.
Spectrum flexibility gets attention too. Dynamic spectrum sharing across 4G and 5G lets Taiwan Mobile use its spectrum assets where they’re needed rather than locking bands to a single generation, while carrier aggregation combines multiple frequency bands to boost data rates, throughput, and end-user speeds. Neither technology is new to this partnership, but both carry forward into the expanded network.
The more interesting service angles sit on top of Nokia’s previously deployed 5G standalone core. Network slicing enables tailored virtual networks for enterprise customers — a capability carriers have talked about monetizing for years, with mixed results so far. The agreement also adds support for RedCap devices, the reduced-capability tier of 5G aimed at lower-power, lower-cost IoT sensors and wearables that don’t need full 5G performance but benefit from the connectivity.
Compared to the LTE network it’s steadily replacing, the buildout continues a densification push that has already modernized thousands of LTE sites and deployed more than 3,000 5G cell sites nationwide.
Energy efficiency is key
Power consumption runs through nearly every element of this deal. The AirScale hardware relies on Nokia’s ReefShark system-on-chip technology, which the company markets as lowering power consumption per bit — a vendor claim, though one consistent with the broader industry trend of efficiency gains per hardware generation. Layered on top, AI-powered algorithms optimize power usage against real-time traffic patterns, proactively cutting consumption during low-traffic periods or in underutilized cells.
There’s a concrete corporate target behind this. Taiwan Mobile has committed to RE100, aiming to run on 100% renewable electricity by 2040, and a network that draws less power makes that target more achievable. The savings cut operating costs too, which is arguably the more immediate motivation.
The Taiwan deployment also mirrors what Nokia is doing elsewhere. The company has a similar AI-RAN collaboration underway with Telia Finland, testing AI-powered RAN use cases in European markets — a sign that the Taiwan Mobile deal is one piece of a broader commercialization strategy rather than a one-off.
Plenty remains undisclosed. Neither company has shared the financial terms — capex size, payment structure, or the revenue impact for Nokia. The implementation timeline and rollout phases are unannounced, as is any assessment of what automated network operations mean for the staff who currently run them. And the public agreement says nothing about AI governance, security, or how algorithmic bias in areas like traffic prioritization will be managed.
Table of Contents