Modern telecommunications networks are growing increasingly complex. Managing these vast systems with traditional manual methods is no longer practical or cost-effective. To maintain high network reliability and service quality, telecom companies must adopt artificial intelligence. However, running AI models requires a continuous stream of clean, real-time data. Unfortunately, most telecom operators find their data locked within fragmented, isolated databases.
To address this problem, Nokia and Databricks completed a joint proof of concept. They demonstrated a unified, cloud-agnostic data platform designed to support AI-driven autonomous networks. The successful trial proved that operators can scale real-time analytics to tier-1 levels without facing vendor lock-in. By creating a flexible, modern data architecture, this collaboration provides the foundation for the next generation of self-healing telecommunications networks.
The Legacy Hurdle: The Costs of Siloed Telecom Data
Telecom providers manage vast infrastructures with millions of connected devices. These systems generate massive amounts of telemetry data daily, including packet logs, signal strength readings, call records, and billing transactions. However, this information is rarely centralized. Instead, it is scattered across hundreds of separate Operations Support Systems (OSS) and Business Support Systems (BSS). These platforms are often built by different vendors, using distinct data architectures and formats with minimal tolerance for error.
Data silos are an expensive operational burden for the global telecom industry. Research indicates that data fragmentation and poor integration cost large organizations an average of $12.9 million annually. Additionally, around 83% of executives report that internal silos restrict operational efficiency and slow down digital transformation efforts. When network teams operate in silos, they struggle to gain a unified view of their network’s health.
For a telecom operator aiming to use AI, data fragmentation is a massive obstacle. AI models cannot make accurate decisions if they cannot access clean, unified datasets. When data is stuck in silos, engineers must spend weeks or months writing custom pipelines for each database. By the time they extract, clean, and process the data, network conditions have changed, and the opportunity to prevent a service disruption is lost. The joint proof of concept by Nokia and Databricks offers a way out of this trap.
Decoding the Nokia and Databricks Joint Architecture
To solve these problems, the engineering teams designed a system that lets operators run data pipelines on any underlying infrastructure. Whether an operator uses a private cloud, public cloud, or on-premises servers, the architecture keeps the data workflows consistent. The test case used simulated, real-time performance data to prove that engineers can write their data logic once and run it anywhere.
Seamless Cross-Platform Pipelines Without Code Rewrites
A primary success of the proof of concept is its complete cloud-agnostic capability. Telecom providers often face heavy vendor lock-in once they choose a cloud platform or a proprietary data warehouse. Migrating to another system typically requires rewriting thousands of lines of data processing code, which costs millions of dollars.
The Nokia and Databricks project proved that operators can deploy a data pipeline once and run it across different systems without code changes. During trials, the same data workflows ran successfully on two distinct environments: the Databricks platform and a completely open-source data stack. This dual compatibility means an operator can move workloads to balance costs, performance, or regional security policies without paying software developers to rewrite the pipelines from scratch.
Decoupling Logic with Python and Abstract Compilers
To achieve this flexibility, Nokia engineers created a unique abstraction layer. Instead of writing code tied to specific cloud platforms or proprietary SQL formats, the team designed the core data transformation rules using vendor-neutral Python. This approach completely separated the business rules from the execution software.
When an operator deploys a workflow, a custom compiler automatically translates the abstract Python code into the native format of the chosen target platform. If the workflow runs on the Databricks platform, the compiler translates the logic into Delta Live Tables. If the pipeline runs on the open-source stack, the compiler converts the logic into Flink SQL. This automation saves hundreds of engineering hours and removes the manual coding errors that often happen during platform migrations.
Generative AI and Intelligent Data Fabric Agents
The proof of concept also incorporated generative AI capabilities to simplify how operators access data. The system features an intelligent data fabric agent that processes natural language requests to create and deploy new pipelines.
Typically, if a network operations team wants to track a new performance metric, they must submit a ticket to the data engineering department. An engineer then spends days writing database queries and designing a pipeline. With the generative AI capabilities validated in this test, a network manager can type a simple request in plain English, such as asking for a real-time monitor of dropped calls in a specific neighborhood. The intelligent agent automatically writes the pipeline code, requests human approval, and deploys it immediately, shrinking deployment times from weeks to minutes.
Driving the Autonomous Telecom Network Movement
The telecommunications industry is under intense pressure to automate its operations. Global mobile data traffic is rising fast, yet average revenues per user remain largely flat. To maintain profit margins, operators must find ways to lower their operating expenses while maintaining high network quality.
Real-Time Stream Ingestion via Apache Kafka and Flink
To demonstrate that their platform can handle real-world demands, Nokia and Databricks simulated data ingestion at a scale matching a tier-1 telecom operator. A tier-1 network handles millions of concurrent devices and thousands of base stations, producing massive, high-velocity streams of telemetry data.
The architecture relies on Apache Kafka to capture this fast-moving data, acting as a buffer to prevent downstream systems from being overwhelmed during peak traffic. Once ingested, Apache Flink processes the events in real time. Flink calculates key performance metrics, such as signal quality and packet loss, as the data flows through the system. This low-latency processing allows the platform’s AI agents to spot anomalies and adjust network configurations immediately, preventing service issues before customers notice them.
Navigating the Complex Path of Legacy Infrastructure Transition
While the simulated tests showed great promise, deploying such an advanced platform across real telecommunications networks is a complex task. Most telecom providers do not run on a single, modern software environment. Instead, they operate hybrid networks where legacy 3G and 4G hardware platforms run alongside modern 5G base stations.
Legacy systems do not natively support modern open-source streaming tools like Apache Kafka or Iceberg. Converting and normalizing legacy telemetry data requires additional middle layers, which can add complexity and introduce processing delays. Ensuring data consistency across different vendor ecosystems and maintaining strict data security compliance in a multi-cloud environment remain some of the biggest real-world challenges for operators during this transition.
Financial and Strategic Implications for the Tech Giants
The financial stakes in the autonomous networks market are high. As mobile operators seek greater efficiency, they are devoting significant portions of their capital budgets to automated technologies.
Market Scale and the High Stakes of Telecom Automation
Industry analysts estimate the global autonomous networks market is valued at approximately $7.0 billion and will reach $17.5 billion by 2029, growing at a compound annual rate of 20.1%. Leading operators are already committing major capital; for example, major telcos in India spent roughly $1.2 billion in 2025 on network automation and AI integrations to keep up with surging demand. TM Forum studies show that mature autonomous network operations can reduce operations and maintenance costs by up to 55%, while simultaneously boosting customer satisfaction by up to 71%.
This rapid market expansion makes the partnership highly strategic for both Nokia and Databricks. For the full year 2025, Nokia reported solid financial performance, with net sales reaching €19.889 billion (around $21.4 billion) and trailing twelve-month revenues of $23.42 billion. Nokia also secured over €2.4 billion in AI and cloud orders, proving that network automation is a major driver of its business.
Databricks, a privately held data and AI leader, is also experiencing massive growth. After securing a $4 billion Series L funding round that valued the company at $134 billion, Databricks reported an annual revenue run-rate of $5.4 billion, a 65% year-over-year increase. The company is currently targeting a new funding round that could value it at $175 billion ahead of a potential public offering. By collaborating with Nokia, Databricks positions its platform as a core component of global telecommunications infrastructure.
The Future Path of Telecom Automation
This proof of concept shows that the telecommunications industry can overcome its long-standing data fragmentation problems. By creating a unified, cloud-agnostic data platform, Nokia and Databricks have shown that operators can successfully deploy scalable, real-time AI analytics. This flexible approach protects operators from vendor lock-in while providing the essential data foundation needed to transition to fully autonomous, self-healing networks.
As networks continue to evolve, the ability to process data across cloud boundaries will become a standard operational requirement. The success of this joint trial offers a practical blueprint for global telecommunications operators to modernize their data practices and prepare their businesses for the age of automated operations.
