Nokia accelerates AI for networking drive

With artificial intelligence (AI) firmly established in its product offerings, global comms tech provider Nokia has elevated its position in AI-ready networking, expanding its relationship with Amazon Web Services (AWS) to add its Autonomous Network Fabric to the hyperscaler. It also completed a joint proof of concept (PoC) with Databricks, demonstrating a unified, substrate-agnostic data platform designed to support AI-driven autonomous networks.

With availability expected later in 2026, the collaboration adds Nokia’s Autonomous Network Fabric to AWS with the core intention of addressing the challenge of networks historically relying on siloed operational and business support systems.

The Autonomous Network Fabric’s core capabilities include : unified data management across domains, agentic AI for service operations and optimisation, and digital twin simulations for proactive impact assessment.

For AWS, this offers network observability, analytics, security and automation together through unified data management, agentic AI for service operations and optimisation, digital twin simulations and intent-based networking.

Businesses will now have access to advanced AI and cloud services required for Level 4 autonomy. This builds on a set of existing digital operations applications from Nokia – covering orchestration, assurance and unified inventory – already on the platform.

The broad Nokia portfolio will now deliver intent-based service orchestration across multi-domain, multi-supplier networks, and provides 360-degree observability with AI-powered anomaly detection, root cause analysis and closed-loop resolution. Importantly, it looks to offer a single source for network topology and resources.

Running on AWS, Nokia said its solutions will gain elastic scalability, global availability and broad model choices through cloud, AI and machine learning (ML) services – including Amazon Bedrock and Amazon SageMaker – enabling operators to innovate faster while reducing infrastructure costs.

The net result is said to be intent-based networking that translates business goals into automated closed-loop actions. Nokia is also engineering an optimised cloud footprint designed to that minimise compute and storage requirements versus traditional on-premise deployments.

Meanwhile, the PoC with data and AI company Databricks aimed to show how network providers can simplify fragmented data environments and deploy real-time analytics at scale, enabling faster decision-making, improved network performance and more efficient operations.

The PoC looked to address a long-standing industry challenge: dealing with various siloed operational and business support systems, each with its own data architecture, making it difficult to apply AI consistently across domains. Nokia stressed that to truly harness AI and multi-agent systems, operators needed a common data platform that can run seamlessly across different cloud environments or on-premise infrastructure, without the need to rewrite code.

The POC showed the ability to develop a joint architecture that handles the massive scale and real-time ingestion speeds required to feed network data to AI agents for automated, cross-domain decision-making.

Engineering teams from both countries focused on a real-time performance management use case, simulating analytics ingestion with an intent to scale quickly to match tier-1 operator scale in the cloud.

Their work is said to have delivered several key technical breakthroughs designed to simplify how telecom operators build and run data-driven services across different environments cross-platform data pipelines, without coding complexity; supplier-neutral data logic design; automated deployment across environments; AI-powered creation of new data products; and a data fabric built for the agentic world.

Data pipelines were created once and deployed across different platforms without modification. In trials, the same data workflows ran on both Databricks and an open source stack based on Apache Flink, Kafka and Iceberg, supporting real-time streaming, batch processing and query-time data products.

To avoid lock-in to any single platform, Nokia engineers developed transformation logic using an abstract, platform-independent expression in Python. By separating the core logic from platform-specific connectors, the same data workflows were reused across multiple environments.

The teams also validated a custom compiler that automatically adapted workflows at deployment. Based on the target environment, it translated the abstract logic into native formats and added the platform-specific connectors, eliminating manual rework and accelerating time to deployment.

Using simple natural language prompts, an intelligent data fabric agent generated new data products, requested human validation and deploy the pipeline automatically, resulting in faster innovation with less manual effort. In the agentic world, the same mechanism can be used by other agents to create dynamic data products on demand by communicating agentic with the data fabric agent.