How AI-driven autonomous networks are beginning to take shape

As capex tightens, commoditisation deepens, and hyperscalers chip away at traditional revenue streams, telcos are under mounting pressure to rethink their networks. Yet, as Prashant Ramesh Malkani, Country Head – Network Infrastructure APAC Market at Nokia India, tells Voice&data, this is also a moment of immense possibility.

There is untapped potential in edge computing, cloud and AI factories, FTTX, and fibre expansion through XGS-PON and 25G PON. Opportunities are emerging in quantum-resistant security and QKD, alongside architectural shifts such as disaggregation, open-source software (OSS) modernisation, IP softwarisation, and backend overhauls. The adoption of Full-Stack Disaggregated Networks (FSDN) is setting the stage for autonomous, AI-optimised networks.

In this in-depth conversation, Malkani shares how telcos can shift from connectivity providers to intelligent service platforms—leaning into dark fibre, edge-native capabilities, sustainable networks, and AI-driven orchestration—to stay relevant in the decade ahead.

As telco capex slides, commoditisation pervades, and revenues take a hit in the telecom industry, how do you view the telco market ahead? Could you share your perspective on segments such as Edge, Cloud, MVNOs, private networks, hyperscalers, and AI factories?

We are seeing declining revenues, rising costs, and increasing competition from alternative network providers. However, emerging segments like edge computing, cloud, and AI factories present massive opportunities for telcos to unlock new revenue streams. Telcos could deploy AI-enabled edge data centres at radio sites to host enterprise and AI workloads. They can also partner with cloud providers to offer Edge-as-a-Service, moving beyond traditional connectivity models.

Collaboration with hyperscalers opens up the possibility of delivering cloud-native networking solutions, while productising private cloud connectivity and hybrid cloud orchestration. Last but not least, telcos could expand their dark fibre footprint and build AI grids to power compute centres, enabling AI-driven network automation and offering predictive analytics as a service.

How much would networks change in areas such as disaggregation, OSS modernisation, IP layer softwareisation, new architectures, back-end overhaul, and digital twins?

Nokia anticipates significant transformation in network infrastructure across all these areas—disaggregation, OSS modernisation, IP softwarisation, new architectures, back-end overhaul, and digital twins. Telcos are moving away from monolithic, hardware-driven networks toward disaggregated, software-defined infrastructures. They are increasingly adopting cloud-native, microservices-based OSS platforms and leveraging AI/ML for predictive maintenance, fault resolution, and service orchestration.

The adoption of FSDN is accelerating automation, while cloud-native IP stacks and zero-touch provisioning are streamlining network operations. AI-driven autonomous networks are beginning to take shape, capable of real-time self-optimisation. In parallel, the full-scale adoption of digital twins is transforming network planning and real-time simulation, using AI-powered predictive analytics to prevent outages, optimise traffic flows, and enhance overall network performance.

What are telcos doing right—and not—when it comes to their networks?

As demand for both AI and non-AI-based high-speed services continues to rise, telcos are rightly focusing on expanding their fibre optic and IP transport networks, with interface speeds reaching up to 800 Gbps. As networks become more complex, telcos can benefit from Software-Defined Networking (SDN) to enhance flexibility and control. Upgrading Customer Premises Equipment (CPEs) has become crucial to support the growing need for higher speeds, lower latency, and improved Quality of Experience (QoE) for end-users. Telcos are also exploring AI and automation to improve network efficiency and streamline operations—this is a step in the right direction.

Expanding fibre networks through technologies like XGS-PON and 25G PON, along with Fixed Wireless Access (FWA), will address rising bandwidth demands.

Looking ahead, telcos have the opportunity to evolve from traditional connectivity providers into enablers of AI-driven services. This includes deploying AI for traffic optimisation, predictive analytics for proactive maintenance and fault detection, and building greater network reliability. Quantum-secure communications will be key to enhancing data security. At the same time, expanding fibre networks through technologies like XGS-PON and 25G PON, along with Fixed Wireless Access (FWA), will address rising bandwidth demands. The deployment of additional edge nodes will further support content delivery, IoT use cases, and latency-sensitive services, enabling a more efficient, scalable, and responsive network infrastructure.

Should telcos start looking at dark fibre and AI grids as the next big business opportunity? What is the right path to pursue that?

Telcos should expand dark fibre routes around AI data centres, partner with hyperscalers for dedicated fibre leasing, and enable fibre-sharing models to support AI-driven networks. An AI Grid refers to a distributed network of AI compute resources interconnected via high-speed fibre and edge networks. These grids distribute compute workloads across fibre-connected edge nodes to reduce latency and cost. By enabling such infrastructure, telcos can move up the value chain—monetising their networks by offering AI-powered services such as AI-as-a-Service (AIaaS), and positioning themselves as integral players in the AI compute ecosystem.

What has changed about network infrastructure for telcos today, with the advent of fibreisation, AI, Software-Defined Networking (SDN), mesh, and Fibre-to-the-X (FTTx)?

The telecommunications industry is undergoing a significant transformation, driven by new use cases that demand high bandwidth, low latency, and improved reliability. With the surge in AI applications, OTT content, 4K streaming, and gaming, telcos are being pushed to upgrade their network infrastructure, ranging from backhaul to Customer Premises Equipment (CPEs).

Fibreisation is at the core of this shift. The move toward fibre networks and the adoption of FTTx technologies allow for far greater speeds and more stable connections compared to legacy copper-based systems, such as DSL. Fibre is now commonly used to backhaul data from 5G cell towers, and with emerging applications demanding stringent KPIs, FTTx is evolving rapidly—from 2.5 Gbps to 10, 25, 50, and even 100 Gbps.

On the architecture side, the rise of decentralised data centres and mesh networking is boosting resilience and reachability. Mesh topologies allow data to flow through multiple paths, reducing single points of failure and enhancing network robustness. Meanwhile, Software-Defined Networking (SDN) is enabling telcos to decouple the control plane from the underlying hardware, making networks more programmable and agile. This results in simplified management, enhanced operational flexibility, and greater cost efficiency, which are crucial for meeting the demands of next-generation services.

How much is changing with optical chips and self-healing networks?

The network is undergoing a fundamental reinvention with the advent of optical chips, which are significantly improving efficiency by reducing power consumption and enabling higher-speed data transmission. This evolution is reshaping how we define, design, and deploy network infrastructure.

The introduction of self-healing networks is becoming increasingly critical, especially as the scale and complexity of transport networks—both optical and IP—continue to grow in the telco domain. 

At the same time, the introduction of self-healing networks is becoming increasingly critical, especially as the scale and complexity of transport networks—both optical and IP—continue to grow in the telco domain. These networks can automatically reroute traffic or adjust configurations in real time to bypass failed components, thereby minimising downtime and improving overall resilience.

The real force multiplier here is the integration of AI and Machine Learning (ML). These technologies are being applied across IP and optical layers for traffic analysis, fault detection, predictive maintenance, capacity planning, and real-time optimisation—ultimately enhancing both operational efficiency and user experience.

Does open source help or impede innovation in networks? Could you share some information about your role in NetOps and event-driven automation?

Open source plays a crucial role in driving innovation in networking by promoting collaboration, standardisation, and interoperability. It allows telecom and enterprise networks to adopt cutting-edge technologies without vendor lock-in, fostering rapid evolution. However, challenges such as integration complexities can sometimes impede progress. Nokia actively embraces NetOps (Network Operations Automation) and event-driven automation to enhance network efficiency, reduce operational costs, and improve service reliability.

Would networks become more critical, more vulnerable, or more utility-ish in the next two to three years? Would this differ for various industries?

In the coming years, networks will become more critical and vulnerable, but the impact will depend on the industry. The rise of AI-driven automation, 5G, cloudification, and quantum computing will increase networks’ dependence while exposing them to greater security risks. Networks will become increasingly critical for manufacturing, airports, railways, utilities, defence, and AI-powered businesses, requiring low-latency, high-reliability networks.

Due to low-latency use case requirements, workloads will shift from centralised data centres to edge networks, which will require real-time, always-on connectivity. Cyber threats, ransomware, and quantum computing will increase the vulnerability of networks, particularly in terms of encryption, authentication, and data integrity. However, this issue can be addressed in the long run by enhancing network security through the implementation of quantum-resistant cryptography and ultra-secure communication methods, such as Quantum Key Distribution (QKD).