Tata Communications plans to invest between US$400 million and US$450 million in a new, approximately 10,000-kilometre fibre network designed to connect India’s major artificial-intelligence and data-centre hubs. Announced on October 26, the project—branded the “AI Super Highway”—is intended to provide high-capacity, low-latency links between computing clusters, cloud platforms and enterprise customers as AI infrastructure expands across the country.
The plan targets a less visible constraint in the AI boom: data centres need more than chips and electricity. Training and serving large AI models moves enormous volumes of data between accelerators, storage and other facilities. If networks cannot move that data quickly and reliably, adding more computing capacity alone will not deliver the full benefit. Tata Communications says the new terrestrial network will connect domestic AI corridors with its global backbone and international cable landing stations. Its official announcement describes a three-phase expansion.
Which cities will the AI Super Highway connect?
The first phase will focus on the Mumbai–Pune–Hyderabad–Chennai corridor, which the company identifies as one of India’s key AI and hyperscale growth routes. Later phases are planned to extend connectivity into eastern and western India, with the company naming Vizag, Bhubaneswar, Jamnagar and GIFT City among the intended locations.
The company says the network will serve hyperscalers, newer “neocloud” providers, AI platforms, Global Capability Centres (GCCs), software and digital businesses, banks and other enterprises. These customers use data centres for different reasons: cloud firms host customer workloads; AI labs train or run models; and large enterprises often need resilient connections between their own systems and external cloud environments.
The investment is planned for new long-haul fibre, with links integrated into Tata Communications’ existing national, metro and international infrastructure. The company already operates a substantial network, including more than 500,000 kilometres of subsea optical fibre and over 200,000 kilometres of terrestrial fibre, according to its press release. The announced project is an expansion of that backbone rather than an entirely separate network.
Why AI needs a different kind of network
Traditional enterprise connectivity has long carried email, business applications, video calls and routine cloud traffic. AI data-centre traffic can be more demanding. Large model training involves frequent exchanges between accelerators and storage, while AI inference—the process of producing answers from a trained model—requires data to travel reliably between users, applications and compute systems. When workloads are spread across facilities, poor connectivity can become a bottleneck even if each facility has enough processing power.
High-capacity fibre and carefully designed routes can help reduce congestion and improve resilience, but kilometres of cable alone do not guarantee low latency. Actual performance depends on the route taken, network equipment, congestion, routing design and how traffic is managed. Tata Communications is presenting the investment as a way to give customers higher capacity and more reliable connectivity; measurable gains will depend on the final routes and services delivered.
The company is entering a market growing alongside India’s data-centre build-out. Fragna Tech has previously covered new AI capacity plans such as the 240-megawatt CoreWeave–AdaniConneX deployment in Navi Mumbai and Microsoft’s India South Central cloud region in Hyderabad. Those projects illustrate why network routes between data-centre clusters and business centres are becoming a bigger part of the infrastructure equation.
How much will it cost, and when will it be built?
Tata Communications has set out a planned investment range of US$400 million to US$450 million and a total network length of around 10,000 kilometres. In an interview with Moneycontrol, an executive said the capital would be deployed over roughly three to five years, with rollout priorities shaped by expected traffic, capacity requirements and the pace of data-centre development.
The company has not publicly named all anchor customers or disclosed a complete corridor-by-corridor construction timetable. Its announcement says a growing share of the expansion is expected to be supported by anchor customer commitments and contracted demand, but the public release does not list customer names or commercial terms. That leaves the final pace of rollout dependent on detailed project execution, demand and the development of the connected data centres.
Tata Communications estimates that the data-centre-to-data-centre connectivity market in India could exceed US$1 billion over the medium term. That is the company’s market outlook, not an independently verified forecast or a guarantee of revenue. In its interview, the company also said it was targeting a 30–40% share of the market it can address. The scale of that opportunity will depend on how quickly customers move from discussions to live workloads and how rivals expand their own networks.
What it could mean for India’s AI ambitions
India’s AI infrastructure debate often focuses on GPU supply, power availability, land and cooling. The connectivity layer is less visible, but it can determine how efficiently those assets are used. Fibre connecting Mumbai, Pune, Hyderabad and Chennai could help cloud operators, AI service providers and large companies move data between locations without relying exclusively on longer or more indirect routes. Links from domestic data-centre clusters to international cable landing stations could also support workloads that span India and overseas markets.
For Indian businesses, the most direct benefits—if the network meets its design goals—would be for data-intensive services and organisations that need predictable, resilient links between multiple sites. It does not mean consumer internet speeds will automatically increase, nor does it guarantee cheaper cloud or AI services. The project is aimed primarily at enterprise and data-centre connectivity, and any downstream price or performance impact will depend on customer adoption and competition.
The initiative also reflects a wider shift: AI capacity is a system, not just a collection of chips. A data centre needs compute, memory, power, cooling, storage and strong network links to function effectively at scale. Investment in fibre does not solve every infrastructure constraint, but without enough network capacity, extra compute can be harder to use efficiently.
The milestones to watch
The first tangible test will be the rollout of the Mumbai–Pune–Hyderabad–Chennai corridor: which segments are built first, when customers can order services, and what capacity and resilience the live network delivers. Further indicators will include the pace of expansion toward the east coast and western India, disclosed anchor customers and whether data-centre operators sign up for long-term connectivity agreements.
For now, the US$400–450 million figure is a planned investment rather than proof that all 10,000 kilometres are already funded, constructed or operational. Tata Communications is betting that India’s AI build-out will require a stronger fibre backbone connecting its compute hubs. If data-centre growth continues, that network layer may become as strategically important as the facilities and accelerators it is designed to connect.
