India's AI traffic surge threatens to outrun telco revenue gains

Daily multimodal AI use in India is projected to nearly double from 22% today to 43% by 2030, well above the global forecast of 30%, according to Ericsson ConsumerLab research — and India's mobile networks, built for downloading rather than uploading, are not ready for the shift.

The figure matters because of what AI traffic does to a network. Indian consumers increasingly share images, video and audio with cloud-based AI models. That makes mobile traffic more uplink-intensive and less predictable. Traditional networks, in India and everywhere else, were dimensioned for downlink: social media, browsing and video streaming all pull far more data than they push.

"Multimodal AI will introduce new use cases that reshape mobile data consumption, driving not just higher traffic but also changing how it flows," said Siddhant Cally, senior analyst at Counterpoint Research.

Why is uplink the weak point?

The technical problem is structural. Arun Menon of MTN Consulting laid it out plainly: India's 5G networks were built primarily for high-volume downlink traffic, using mid-band spectrum, particularly 3.5 GHz. Uplink performs worse indoors and at cell edges because phones transmit at much lower power than base stations.

AI applications that send live video, images and audio to cloud-based models will press directly on that weak spot. On-device processing, by contrast, will generate less additional traffic.

Traffic patterns add a second strain. AI usage arrives in spikes — a user uploads an image or a voice clip, then waits for a response. "These short sessions can increase radio scheduling demands and latency. Unlike video streaming, where buffering can hide delays, AI interactions require a responsive exchange," Menon said.

Can today's networks cope?

For now, yes. AI traffic remains a small share of total network traffic on India's 4G and 5G networks, and existing infrastructure can absorb it. The challenge lies ahead.

"The networks are ready for AI but they have to be optimized as per the particular requirements of the specific AI application," Nitin Bansal, Managing Director for India and Head of Networks for Southeast Asia, Oceania and India at Ericsson, told Light Reading in a recent interview.

As volumes grow, operators will need upgrades across radio capacity, uplink performance, spectrum and backhaul. Not uniformly, though. "Operators should add sites where demand warrants it, rather than assume a uniform upgrade cycle across India," Menon said.

Cally went further, cautioning that incremental radio work will not be enough. "Delivering consistently responsive experience at scale will require more than incremental radio upgrades. Better uplink performance, increased site fiberization, smarter spectrum utilization will be equally critical," he said.

What role does 5G-Advanced play?

The transition to 5G-Advanced (5G-A) should help manage AI-driven traffic, but India's operators start from different positions. Reliance Jio, the country's largest service provider, has already built a 5G Standalone (SA) network; Bharti Airtel only recently began moving to SA.

Menon ranked the priorities: uplink capacity, fiber backhaul and completing the move to 5G-SA. 5G-A can improve uplink performance, support latency-sensitive traffic better, handle lightweight devices such as AI glasses and wearables, and use AI itself to cut radio network energy consumption.

He added a caveat that applies to any 5G-A marketing: the underlying architecture determines the payoff. A 5G-A network delivering limited benefits is the likely outcome if the operator still runs a 4G core under NSA or has inadequate backhaul. Microwave-linked towers can also become bottlenecks as radio capacity grows.

Will the traffic pay for itself?

Here the story turns uncomfortable for operators. TRAI data for FY26 shows wireless data usage grew 24.74%, while data revenue rose just 11.53%. Average industry revenue per gigabyte fell 10.59%, from INR8.97 ($0.09) to INR8.02 ($0.08) — roughly 11 gigabytes of traffic per dollar of revenue, and falling.

In that gap lies the monetization problem. Traffic is growing at more than twice the pace of revenue, and per-gigabyte realization is sliding.

"In India's price-sensitive market, operators will struggle to pass on network investments through higher data tariffs alone. Monetizing premium connectivity and enterprise services, improving networking efficiency, and partnering with AI and cloud players will be key to protecting margins," Cally argued.

Indian operators now face a narrowing window: upgrade uplink, backhaul and core architecture for traffic that may never pay proportionately — or risk an unresponsive network as multimodal AI use doubles by 2030.