By A.S. Prasad, Vice President of Product Management and Engineering, Vertiv India
Following the revolution of the telecom sector in India, network operators are now navigating a scale that few planners could have imagined a decade ago. Nokia’s latest Mobile Broadband Index puts average monthly mobile data use at over 31 GB per user, with total traffic above 27 exabytes a month and 5G carrying nearly half of it. The trajectory is steep; Ericsson’s June 2026 Mobility Report expects per-user consumption to nearly double within five years. And this is merely the baseline, before intelligence fully migrates into the network itself through AI-assisted radio optimisation, inference at the edge, and AI services delivered directly from network sites.
In my role as a product engineering leader at Vertiv, I view this evolution through a distinct lens. The defining challenge of this era is not predicting what the network will do but proactively enabling the physical layer beneath it to be robust and agile enough to absorb those demands.
Designing a Pathway, Not Guessing a Peak
AI workloads do not adhere to a predictable rollout calendar. Vertiv’s Frontiers 2026 report outlines the rise of distributed AI, where inference is deployed across numerous edge environments rather than concentrated in a few large facilities. For the telecom sector, this necessitates dense, compute-capable capacity in regional hubs, local exchanges, and edge locations. Because many of these sites were designed for lighter, traditional loads, introducing hardware that runs inference which can draw several times the power of a conventional rack presents a critical bottleneck.
Attempting to forecast an exact peak capacity is a losing exercise. At Vertiv, we advocate for a more disciplined approach: designing a resilient path to that peak. This enables the site to grow in a known, validated pattern without the cost and disruption of being ripped out and rebuilt.
Modularity: The True Unit of Scalability
While modular design is frequently praised for its speed advantage, its ultimate value lies in predictability. Standardised, factory-tested infrastructure blocks eliminate site-to-site variation, which is a critical advantage in a country where one operator’s footprint spans coastal humidity, desert heat, dust-heavy corridors, and fluctuating grid quality.
The industry is also evolving to validate designs before committing capital. Vertiv’s Frontiers 2026 report identifies digital twin-driven design as a defining trend, enabling infrastructure to be simulated, stress-tested, and managed in real time before and after construction. For a network planner, this shifts the paradigm from asking “will it work?” to stating “we have already seen how it behaves under load.”
Incremental Capacity: Deploying Power and Cooling with Precision
The greatest commercial risk in AI-era infrastructure is stranded capital. Build too much too early and the assets sit idle; build too little and the network chokes. Modern power and cooling architectures must empower operators to add capacity in precise increments that track real demand.
This evolution is most visible in power distribution. Vertiv’s Frontiers 2026 report points to higher-voltage DC distribution as a strategic response to rising densities, as fewer conversion steps improve efficiency, and lower current helps support denser loads. This approach builds on the telecom sector’s native expertise, where DC power has long been the norm, and aligns perfectly with renewable sources like solar and fuel cells that generate DC natively.
While the report is candid about implementation hurdles such as stricter safety requirements, limited skilled labour, and higher upfront costs, the recommended strategy is to evaluate and pilot these solutions in high-density areas rather than committing wholesale prematurely. Thermal management demands the same agile logic. Sites will not flip from traditional air cooling to liquid cooling overnight.
Serviceability: The Underrated Imperative
If modularity dictates how a site grows, serviceability dictates whether it stays online. Yet, this remains the requirement that product roadmaps most frequently underestimate.
The design brief must address a blunt reality: who fixes this at two in the morning, at a site hundreds of kilometres from the nearest engineer, and with what tools? The solution lies in providing components that generalist technicians can seamlessly swap, intelligent remote monitoring that flags degradation before failure occurs, and advanced analytics that transform maintenance from reactive to planned.
Serviceability engineered at inception is less expensive and disruptive than serviceability retrofitted after an outage.
Three Mandates Before the Network Turns AI-Native
The immense scale of India’s networks leaves little room for infrastructure that requires constant retrofitting. Before AI-native architectures arrive in earnest, operators and their technology partners must be able to definitively answer three questions:
Can capacity be added in modular steps, without demanding a site rebuild?
Can every site be intelligently monitored and serviced without a specialist visit?
Can the design be validated digitally before capital is committed?
If the answer to all three is yes, “scale without boundaries” ceases to be an aspiration and becomes a well-planned engineering outcome. The networks that lead the AI era will be those that partner with experts to treat the physical layer as a strategic design priority today, rather than a procurement concern tomorrow.