Key Takeaways

  • The joint proof of concept validated a shared data architecture that runs across cloud and open-source environments without code changes.
  • Unified data models can support real-time AI agents in future autonomous networks.
  • Industry demand is rising as communications service providers move toward intent-based and autonomous operations.

The validation of a unified data platform by Nokia and Databricks supports the telecom sector's shift toward AI-driven autonomous networks. Operators have dealt with data fragmentation for years, and this complexity compounds as networks grow more distributed and multi-vendor.

Most large communication service providers operate with hundreds of operational and business support systems, each with its own format, lineage, constraints, and aging workflows. The engineering teams built a joint proof of concept around this challenge, demonstrating how a common data layer can help telecom operators apply AI across domains without rewriting pipelines for each cloud or on-premise environment. Carrier networks are already seeing increased real-time data movement as functions shift toward software-defined operations, driving the search for more adaptable data platforms.

The proof of concept, completed on 24 June 2026, focused on a real-time performance management use case, simulating workloads at the scale expected from a tier-1 operator. The project verified that identical data pipelines could run both on managed data platforms and an open-source stack built on Apache Flink, Kafka, and Iceberg. This cross-platform capability reduces the friction operators often encounter when running analytics across hybrid or multi-cloud estates. It aligns with the trend highlighted by the Gartner telecom operations forecast, which noted that more than 60% of communications service providers are expected to adopt intent-based or autonomous network platforms by 2027.

Engineering teams designed transformation logic using an abstract, platform-neutral expression in Python. By separating logic from connectors, these workflows could be redeployed across different environments without adjusting the core code. The teams validated a custom compiler that automatically adapts each workflow at deployment time. In practice, operators can take a single pipeline and deploy it as Delta Live Tables or as Flink SQL on an open-source platform. This functionality reflects a growing shift toward what telecom analysts and researchers describe as modular data fabrics.

The proof of concept also explored the AI-driven creation of data products. Through natural language prompts, an intelligent data fabric agent can generate and deploy new data products, ask for human approval when needed, and prepare pipelines automatically. In agentic network designs, other agents can request data products dynamically, creating a more fluid system of data exchange that will require new governance approaches as network functions become increasingly AI-mediated.

According to ComputerWeekly coverage of the Autonomous Networks Fabric in 2025, telecom operators have actively searched for ways to unify data management to deploy telco-trained AI models without rebuilding infrastructure. That earlier initiative framed data fabric as the backbone of autonomous network functions. The recent proof of concept extends that trajectory by demonstrating how those data foundations operate across diverse environments.

The collaboration also addresses broader industry economics. A 2024 WAN Forecast noted that autonomous networks could generate around 5% of telecom revenue in combined savings and revenue uplift. For a typical provider, that translates to roughly $800 million a year. Driven by these figures, telecom executives require operational models that help manage complexity, reduce manual interventions, and provide predictive insights.

Global AI traffic across wide-area networks is forecasted to reach 921 exabytes per month by 2034, rising at a 23% compound annual growth rate and approaching 30% of total WAN traffic. As traffic scales, operators face difficult trade-offs regarding what data to store, process in real time, or move across network boundaries. The data fabric capabilities highlighted in this proof of concept, including zero-copy sharing and query-time data products, directly address these workflow bottlenecks.

The ability to selectively feed upper temporal layers in the cloud for retrospective analysis gives AI agents more context when performing root-cause analysis. Operators often struggle with post-incident investigations because their data sets are scattered across operational silos. A unified platform streamlines these tasks, improving the throughput of incident-resolution agents trained to correlate logs, events, and topology data.

The networking vendor has been working with major cloud providers like AWS and Google Cloud to expand deployment pathways for autonomous networks. Parallel initiatives around data mesh, detailed in a 2023 white paper, indicate that distributed, domain-oriented data products will play a central role in how future network automation platforms operate.

Distributed data systems require strict guardrails to ensure reliability across clusters, particularly when real-time analytics feed into downstream automated actions. Telecom operators must manage these architectural trade-offs as they transition from proofs of concept to production deployments.

Nokia and Databricks plan to continue their joint development, aligning with the telecom sector's broader push toward network autonomy. The recent proof of concept demonstrates that cloud-agnostic design patterns and agent-ready data fabrics are gaining traction in historically complex and risk-averse operational environments.