The shift from towercos to fibrecos is redefining how fibre is monetised and managed. However, with only 38 per cent of towers currently fiberised, there is still significant ground to cover. In a panel discussion moderated by Vinish Bawa, Partner and Leader, Telecom, PwC, Chandan Kumar, Head of India Business Centre for Optical Transport, Nokia, and Deepak Sanghi, Executive Vice President (EVP) and Head – IP and Transport Network, Bharti Airtel, shared their perspectives on OFC demand from new and emerging segments. Excerpts from the discussion…
What trends are you observing in optical and transport networks to meet the growing 5G demand? What is Nokia’s perspective?
Chandan Kumar
In the early stages of digitisation, developments occurred in silos. Today, we are witnessing a convergence of the physical, digital and human worlds, ushering in a “hyper-digital” era. This future depends on robust networks, underpinned by high quality OFC infrastructure. These emerging applications demand not just capacity, but ultra-low latency to support real-time processing.
A second layer of convergence is taking place between cloud and data centre networks. True convergence means compute and connectivity working in tandem and OFC.
Artificial intelligence (AI) is adding another dimension to this evolution. With shorter innovation cycles, optical networking now sees chipset upgrades nearly every two years.
From a product standpoint, our focus is on optimising existing fibre assets. We are enhancing capacity per lambda and expanding into the C+L band to deliver more throughput without additional fibre deployment.
How is Airtel handling the diverse requirements of the rural, urban and enterprise sectors from an IP and optical network perspective?
Deepak Sanghi
Our network priorities are clear, and the scale we are operating is vastly different from a decade ago. Earlier, plesiochronous digital hierarchy and synchronous digital
hierarchy networks offered limited capacities, barely supporting 100 Mbps. Each telecom tower requires 1 Gbps or more, with total internet traffic in tens of Tbps, a hundredfold increase from just a few hundred Gbps 10 years ago.
However, with chipsets reaching 5nm and 3nm, performance gains are narrowing. We can no longer rely solely on laying more fibre; instead, we must deploy additional fibre pairs and densify existing corridors. Fibre has become fundamental across both urban and rural areas. While rural connectivity once relied on microwave, today’s 4G and 5G sites demand fibre-grade speeds. To address this, we are modernising core networks and aggressively expanding rural fibre using high-capacity channels (C-band, L-band and 400-600 Gbps).
The importance of fibre extends beyond telecom. Data centres, enterprises, satellite systems and private networks all depend on it. Even satellite constellations like Starlink require ground-based fibre backbones. Legacy copper is being replaced, and fibre is now the baseline for home broadband, with fibre-to-the-home (FTTH) and its backhaul growing rapidly.
What key trends is Nokia observing across Industry 4.0, Industry 5.0, private networks and the broader B2B segment?
Chandan Kumar
In the context of 5G and emerging use cases such as AI, power efficiency is a top priority for vendors. Another critical factor for 5G and Industry 5.0 is real-time data processing, which demands low-latency fibre.
Traditional fibre has limitations. That is why we are seeing an increased focus on advanced fibre types such as hollow-core and multi-core fibre. Previously used in short-reach or subsea applications, multi-core fibre is now being tested for metro and long-haul networks. In one of our trials, we tested a hollow-core fibre spool. The lab results showed very low loss. Latency performance was excellent, ideal for real-time 5G applications in healthcare and security.
The next challenge is compatibility. We are working with fibre manufacturers to create adapters that enable the integration of hollow-core and multi-core fibres with existing single-mode infrastructure. While hollow-core fibre is still early in its life cycle, commercial pilots have begun. The progress seems promising though full-scale adoption will take time. New protocols, practices and workforce training are needed.
How is Airtel approaching fibre investments? How is the company prioritising fibre deployment and network planning?
Deepak Sanghi
We monitor utilisation, run trend analysis and apply predictive analytics across every part of our network. Predicting B2C traffic is relatively straightforward as it grows organically. But B2B is a different challenge, with sudden location-specific spikes, often in terabits and not gigabits. Our first step is demand forecasting. We work closely with partners and customers to anticipate future requirements. Second, from a market perspective, we have identified high-consumption zones. In these zones, we are upgrading our fibre expressways, strengthening infrastructure, upgrading equipment, and improving latency and resilience.
One major issue across the ecosystem is frequent fibre cuts, a growing concern for operators, enterprises and data centre players. To address this, optical ground wire (OPGW) is increasingly being adopted as part of backbone infrastructure, providing both protection and capacity.
The second focus area is metro networks, where demand is surging, especially for FTTH. An FTTH user typically consumes 50-100x more data than a mobile subscriber. Yet, India lags behind global FTTH penetration benchmarks, even compared to other developing economies, leaving substantial room for growth.
From both mobile and B2B perspectives, metro cities face heavy traffic and diverse demands. That is why metros are a priority and not just the top seven. A tiered approach should be followed, starting with the top 100 cities, followed by the next 1,000, then 2,000 and so on, with rural regions coming next. We prioritise fibre deployment based on a combination of factors including population, user profiles, base station density and demographic trends, ensuring that investments are data-driven and future-ready.
What are the key trends in AI and Gen AI with respect to infrastructure and software-defined architecture? Is AI already being used and what is Nokia’s road map for AI?
Chandan Kumar
We have developed several AI-driven use cases, and AI is often the first topic customers raise during discussions. One key area is fibre condition monitoring. At Nokia, we have built a programme called “The Network That Senses, Thinks and Acts”. The idea is straightforward, the system must sense its environment, process the data and then act on it. In our fibre sensing use case, the digital signal processor in our chipset detects changes in fibre polarisation.
Our second AI use case focuses on power efficiency. We developed an AI-based network management system that identifies idle modules and places them into deep sleep, significantly cutting power usage. We have extended this functionality to the port level so that only active ports stay powered while unused ones sleep. This use case has progressed from proof of concept to a general availability product that is now in the deployment stage.
Another promising application is AI-driven system interaction, such as ChatGPT. For example, an operator can ask, “Why am I getting this alarm?” and the system will search relevant documentation, suggest possible causes and recommend corrective actions, all through natural language interaction. We have launched this capability under the name Tara, developed at our Bengaluru lab. It is already live, and we are continuing to explore more AI-driven innovations.
How is the growing use of AI impacting Airtel’s network? What steps is the telco taking to manage this rising demand?
Deepak Sanghi
I believe we are only seeing the tip of the iceberg. Internally, there is significant discussion across teams, IT units and within Airtel Clouds, which powers many AI-driven applications. But these are still internal use cases.
Across the country, data centres are beginning to evolve towards AI-centric services, bringing two key requirements. First, compute demand is set to grow exponentially. This raises challenges like GPU-to-GPU connectivity, which is becoming a major focus for data centre architecture. Traditional top-of-rack set-ups will not meet the needs of modern GPU clusters. As data centres scale, we are already observing multi terabit interconnect requirements between racks. While early, these could soon evolve into B2B offerings. Even server interfaces and hardware design are being reimagined.
Second, there will be immense pressure on latency and bandwidth. AI workloads are often bursty, with short-duration spikes that strain networks despite low average utilisation. We need tailored strategies for this.
Further, while forecasting future latency and capacity needs is manageable, the real complexity lies in the dynamism of AI workloads. Packet loss and network interruptions, especially in remote GPU clusters, pose a serious risk. We have debated this internally, and I will be candid, we benchmarked fibre cut incidents in India versus Europe and the US. India sees up to 300 times more fibre cuts. Some over-the-top customers abroad have never even experienced one and are shocked that this is a common occurrence here. As AI-centric data centres scale in the future, customers will demand absolute reliability. Meeting that expectation, especially in the face of infrastructure vulnerabilities, will be our greatest challenge.
