NNokia Autonomous Network Fabric

On June 19, 2026, NTT DOCOMO switched on Nokia’s MantaRay AutoPilot on a public cloud, cutting a network optimization cycle that had run once a day to as little as 15 minutes. DOCOMO calls it the world’s first commercial mobile network optimization of its kind on a public cloud. It is the clearest named result so far for the agentic AI that Nokia has built around its Autonomous Network Fabric, the AI layer it launched in June 2025 to help operators run networks with fewer manual steps.

The Fabric and its agents are best for large mobile, fixed and converged operators that want AI to triage alarms, find root causes and tune radio networks across equipment from several vendors. This is a broad enterprise portfolio priced by quote: the agents live across the Fabric, the Autonomous Networks Suite, MantaRay, Network Services Platform and fixed and optical products, each with its own name. We rate it 7.7/10 on our scale, with pricing transparency and public reviews as the areas to build. We have not used it; this review is based on Nokia’s and DOCOMO’s announcements and trade coverage.

Facts Pulled From Nokia’s Scattered Announcements

Nokia’s agentic AI story is spread across several announcements. These are the facts we confirmed on the source pages.

The DOCOMO case is the number to anchor on, because a named operator published it and it describes a running system rather than a target. Going from one optimization pass a day to one every 15 minutes means the network can respond to traffic shifts within the same hour rather than the next day.

The 60% to 80% productivity figure is Nokia’s own and is not tied to a named operator, so treat it as a ceiling to test. The Infinera purchase matters because it brought optical networking into Nokia, which is why the agentic work now reaches optical and IP domains as well as radio.

Data, Models and Agents in Three Layers

Nokia’s approach has three layers: data, models and agents. The Autonomous Network Fabric is the base layer, and the agents sit in products across Nokia’s portfolio.

The Data and Model Layer

The Fabric builds a library of correlated data products from across network domains on a data mesh, with low-code tools for creating new ones. Telco-trained LLMs, large action models and ML models work on that data. A knowledge engine explains how data was interpreted, how an issue was analyzed and why an action is recommended, and observability tracks data and AI use end to end, including chain of custody.

The Agent Library and MantaRay AutoPilot

• Agent Library: prebuilt agents for identifying zero-day attacks, reasoning through complex anomalies, triaging events for root cause, troubleshooting service order failures and coordinating several agents, with built-in observability and governance.
• MantaRay AutoPilot: the operator sets quality targets as intents; AutoPilot analyzes base station performance, decides parameter changes and timing, and sends directives to MantaRay SON, which reconfigures the base stations.
• MantaRay SMO: Open RAN-aligned service management with Non-Real-Time RIC functions and AI rApps that manage radio networks, detect anomalies and support dynamic network slicing.
• IP troubleshooting agent: in Nokia’s Network Services Platform, an agent reasons over live network context within set policies to find root causes, cut alert noise and suggest fixes.
• Fixed and optical: agentic features in Altiplano, Corteca and Broadband Easy for broadband, and in WaveSuite for detecting optical equipment failures before they hit service.
• Microsoft Fabric agents: agents for VoNR assurance, RAN diagnostics, 4G and 5G performance analysis and predictive maintenance on data shared through OneLake.

The Google Cloud side of the partnership lists planned capabilities that show where Nokia is taking the Fabric: real-time traffic monitoring, anomaly detection, zero-touch remediation, and elastic scale-out with disaster recovery. Nokia and Google Cloud said in 2025 they were working with a major European operator, which they did not name. Kal De, who leads product and engineering for Nokia’s cloud and network services, summed up the design choice as: “There is no good AI without good data.” That is why so much of the Fabric is about data products and lineage rather than the agents themselves.

Operators Set Intents and Approve Actions

Operators set the intents and policies agents work within. In the Microsoft solution, operational actions need manual approval, and in other areas agents recommend fixes for engineers to apply. Nokia says operators can adopt automation at their own pace, which in practice means deciding domain by domain where agents may act alone.

SaaS on Google Cloud, On Premises or Hybrid

The Fabric can run as SaaS on Google Cloud, on premises with Google Distributed Cloud or in hybrid setups, and the latest Autonomous Networks Suite adds on-premises options. The Microsoft Fabric pairing supports hybrid, private cloud and on-premises deployment for data sovereignty, with prebuilt connectors in place of multi-week data ingestion projects.

Negotiated Contracts by Domain and Module

Nokia does not publish prices for the Fabric, the Agent Library or MantaRay. Contracts are negotiated with operators and depend on which domains, modules and deployment model are in scope.

Turning Nokia’s Results Into Operating Effort

With no published prices, these scenarios show how Nokia’s stated results translate into operating effort. The team sizes and volumes are our own illustrative assumptions.

• A network operations team: a team spending 1,000 hours a month on event triage and troubleshooting would recover 600 to 800 hours a month if it matched Nokia’s typical 60% to 80% productivity figure.
• Radio optimization: moving from one optimization pass a day to one every 15 minutes, as at DOCOMO, means up to 96 passes a day instead of one.
• A broadband helpdesk: at 10,000 support contacts a month, first-contact resolution above 50% means more than 5,000 contacts closed without escalation, and halving return visits cuts 400 truck rolls to 200.

Costs to plan for beyond Nokia’s license: cloud consumption on Google Cloud or Microsoft Fabric, systems integration across non-Nokia equipment, and staff time to define intents, policies and approval rules.

Nokia Autonomous Network Fabric Compared With Amdocs aOS, Netcracker Agentic AI and Blue Planet

Nokia approaches agentic AI from the network side, while BSS-focused vendors start from customer care and billing.

The split matters when you build a shortlist. Network equipment vendors such as Nokia see live radio, IP and optical telemetry directly and can act on network elements, while BSS vendors see orders, customers and bills. Many operators will end up with agents from both sides, which makes shared data and clear handoffs between agents part of the design work from the start.

If radio and transport operations are the bottleneck, Nokia is the closest fit. For order failures and billing, look at Totogi BSS Magic or Amdocs. More options are in our AI agents for telecom guide.

Multi-Vendor RAN Operators Chasing Higher Autonomy

• Mobile operators with multi-vendor RAN: MantaRay SMO supports traditional RAN from several vendors as well as Open RAN.
• Operators chasing higher autonomy levels: DOCOMO is using AutoPilot on its path to TM Forum Autonomous Networks Level 4.
• Network operations centers handling heavy alarm volume: Agent Library agents triage events for root cause and troubleshoot service order failures.
• Fixed broadband providers: Altiplano, Corteca and Broadband Easy add agents aimed at helpdesk resolution and fewer field visits.
• Operators with data sovereignty rules: on-premises and hybrid options keep data in country.

Small ISPs and Billing-Focused Teams

• Small ISPs and enterprises: the portfolio is sold to operators on negotiated contracts.
• Teams focused on billing and customer care: Nokia’s agents target network operations; BSS-focused tools fit better.
• Buyers who need a public price: pricing comes through negotiation.
• Operators wanting one simple product: capabilities are spread across many Nokia products.
• Teams wanting independent reviews first: review-site ratings for these products have not appeared yet.

Agentic Features Spread Across Seven Products

• Product scope: agentic features sit in the Fabric, Suite, MantaRay, Network Services Platform, Altiplano, Corteca and WaveSuite. Get a written list of what each quote covers.
• Vendor figures: the 60% to 80% productivity range and fixed-network results are Nokia’s own.
• Approval design: some agents need manual approval for operational actions; decide where that adds safety and where it slows response.
• Cloud dependencies: Google Cloud and Microsoft Fabric options add their own bills and contracts.
• Platform fit: multi-vendor support helps, but deeper features run on Nokia platforms.

The Expanded Nokia and Microsoft Partnership

• On September 18, 2026, Nokia and Microsoft expanded their partnership so operators can pair Nokia Data Suite with Microsoft Fabric and run network agents.
• On June 23, 2026, at DTW in Copenhagen, Nokia launched the Autonomous Networks Agent Library and upgraded MantaRay SMO, Network Services Platform and WaveSuite with agentic AI.
• On June 22, 2026, NTT DOCOMO announced its MantaRay AutoPilot deployment, live from June 19 on a public cloud.
• On June 18, 2025, Nokia launched the Autonomous Network Fabric with an expanded Google Cloud partnership.

One of the Most Complete Network Operations Portfolios

For a large operator that wants AI agents inside network operations rather than customer care, Nokia’s portfolio is one of the most complete on offer, covering radio, IP, optical and fixed networks with governance and explainability built in. It scores 7.7/10 on our scale. Autonomy and integration score well; pricing transparency and public reviews are the areas where Nokia could add more.

DOCOMO’s move from daily to 15-minute optimization shows the approach working in a live network. Before signing, map which Nokia products each use case needs, confirm where agents may act without approval and test the productivity claims on one domain first.

Fabric Versus MantaRay and Where It Can Run

How Much Does Nokia Autonomous Network Fabric Cost?

Nokia does not publish prices. The Fabric, Agent Library and MantaRay products are sold on negotiated contracts, and cost depends on the network domains, modules and deployment model in scope.

What Is the Difference Between the Fabric and MantaRay?

The Autonomous Network Fabric is the data, model and governance layer. MantaRay is Nokia’s radio network management family, including SMO, SON and AutoPilot, which runs intent-based optimization on top of SON.

Can It Run on Premises?

Yes. Nokia offers the Fabric as SaaS on Google Cloud, on premises with Google Distributed Cloud or in hybrid setups, and the latest Autonomous Networks Suite adds on-premises options.

Does It Work With Microsoft Fabric?

Since September 2026, Nokia Data Suite connects to Microsoft Fabric’s OneLake with prebuilt connectors, and operators can run agents for VoNR assurance, RAN diagnostics and predictive maintenance on that data.

Where Do Nokia’s Network Agents Focus?

They focus on network operations rather than billing or customer care. Some operational actions need manual approval, and the agents work from operator-defined intents and policies. They are built for operators rather than as a self-serve product for small providers.