Tata Communications said on Monday, October 26, that it will build what it calls India's "AI Super Highway": about 10,000 km of new long-haul fibre meant to carry traffic between the country's fast-growing data centre clusters. The company gave a cost range of $400 million to $450 million. Mint, which converted the upper figure to roughly ₹4,282 crore, reported a build period of three to five years, and Reuters carried the $450 million plan the same morning.
The network will arrive in stages. The first leg covers the Mumbai-Pune-Hyderabad-Chennai corridor, which the company describes as a core route for AI and hyperscale growth, and Mint reports that this phase is expected to go live within about three years. Later phases push out to newer hubs on both coasts, such as Vizag, Bhubaneswar, Jamnagar and GIFT City. The operator also plans to connect these domestic corridors to its international cable landing stations and global backbone, so that traffic could travel from an Indian data centre to AI hubs abroad on a single operator's infrastructure.
According to CNBC-TV18, the intended buyers are hyperscalers and neocloud providers, AI and digital-native businesses, global capability centres, enterprises and emerging cloud platforms. Chief executive Ganesh Lakshminarayanan told the channel he expects the Indian market for data centre-to-data centre connectivity to exceed $1 billion over the medium term, and said a growing portion of the expansion rests on commitments from anchor customers. The company's Monday release, as quoted by Mint, made the same point about contracted demand.
Rupesh Chokshi, the chief product and technology officer who also runs network services, laid out the financial case to Mint. The build will be paid for from operating cash flows, he said, and he pointed to an industry-benchmark return on investment of 15% to 20%, with revenue starting as each stretch of capacity comes online. His technical argument is that AI applications must shift large volumes of data quickly between computing sites, while companies keep backups in several places, so slow links between facilities drag on how efficiently AI systems run. Asked about long equipment lead times, he said fibre and electronics makers are expanding output and he did not foresee major delays.
The move comes in a crowded field. Cushman & Wakefield counts nearly 1.8 GW of operational data centre capacity in India over the first six months of 2026, plus a further 3.86 GW being built or planned through 2030, Mint reported, and CBRE puts cumulative investment commitments in the sector since 2021 at about $173 billion. Singapore-based Lightstorm, which says it has already spent $500 million on an AI corridor in southern India and a subsea cable to Singapore and Malaysia, competes directly for this connectivity business, while Airtel's Nxtra, Google, Adani and Reliance are investing heavily in the facilities themselves.
Tata Communications starts from a sizeable base: more than 80,000 km of terrestrial fibre and 30,000 km of metro fibre in India, with links to 118 data centres, according to Mint. Its revenue for the September quarter rose 9.6% to ₹6,688 crore, and its shares traded more than 1% higher on Monday afternoon.
Several of the figures behind the plan come from the company alone, including the market-size estimate, the expected return and the role of anchor contracts, and none has been independently tested. At the time of writing the full release was not on the company's own newsroom, no anchor customer had been named, and no capacity figure or completion date had been given for the later phases. Whether the fibre arrives in time to matter will depend on how quickly AI compute actually fills India's newer data centre hubs.