Tata Communications Bets ₹4,350 Crore on Data Centre Connectivity for India’s AI
The release gives an investment range and a distance pertaining to data centre connectivity. The company’s own executives gave the timeline, the capacity, the funding source and the utilisation figure that explains the decision.
Tata Communications will spend up to 4,350 crore rupees, roughly 450 million dollars, on about 10,000 kilometres of new long-haul fibre connecting India’s AI and data centre corridors. The company calls it an AI super highway. The number that explains the decision appears nowhere in the announcement: its existing network runs at 1.25 petabits per second and sits at 70 to 80 per cent utilisation. An operator at that level has exhausted the room on the network it owns, and the next increment is a build rather than an upgrade. The announcement also says the company will keep scaling its existing National Long Distance and metro footprint in parallel, which is the sentence that distinguishes this from a replacement programme. The highway is incremental capacity for a new class of customer, not a refresh of the network serving existing ones.
The investment was announced on 26 October. Coverage the same day added what the release omits: a three-to-five-year deployment, three phases running to FY36, around 6 Pbps of new capacity, funding from operating cash flows, and a stated ambition to take 30 to 40 per cent of India’s data-centre-to-data-centre connectivity market.
What is being built
The plan is a dedicated long-haul network linking data centre clusters directly rather than routing their traffic through general telecom infrastructure. Phase one covers the Mumbai-Pune-Hyderabad-Chennai corridor and should become operational over about three years. Later phases extend to Vizag, Bhubaneswar, Jamnagar and GIFT City, and the network connects those clusters to international cable landing stations and to the company’s global backbone.
AI-ready is doing a lot of work in this announcement, and it is worth being precise about what it can mean for a fibre network. A cable does not know what traffic it carries. What changes for AI workloads is the design above the glass: dedicated rather than shared paths, direct data-centre-to-data-centre routing that avoids metropolitan handoff points, diverse routes so one cut does not take out both directions, service-level agreements written against latency rather than availability alone, and optical electronics sized for 400 and 800 gigabit wavelengths.
Those are engineering and commercial choices, and they are the substance behind the label. The existing backbone is what makes the plan credible. Tata Communications operates more than 500,000 kilometres of subsea optical fibre and more than 200,000 kilometres of terrestrial fibre. The new build integrates with that estate rather than replacing it, so the company adds capacity on a route network it already owns, maintains and knows how to permit.
Why now: the utilisation problem
Fibre economics turn on utilisation, and the company disclosed the figure that makes its case. Capacity of 1.25 Pbps at 70 to 80 per cent utilisation leaves roughly a quarter of the network free. The build adds about 6 Pbps between FY30 and FY36, close to a fivefold increase on today’s capacity.
That is a large multiple for a business whose core is carrying other people’s traffic. It is also why the funding structure matters. The capital expenditure comes from operating cash flows rather than debt, which means the build proceeds at the pace the business can pay for rather than the pace a lender demands. Chokshi has also cited 25-year asset longevity and owner economics as reasons to own the fibre rather than lease it.
The market and the share target
Tata Communications puts India’s data-centre-to-data-centre connectivity market above 1 billion dollars over the medium term. Chief executive Ganesh Lakshminarayanan told CNBC-TV18 the company targets 30 to 40 per cent of it. That is an ambitious share of a market that does not yet exist at that size, and it explains why the company builds ahead of contracted demand rather than behind it.
The demand context is real and it cuts both ways. India’s operational data centre capacity reached nearly 1.8 gigawatts in the first half of 2026, with a further 3.86 gigawatts under construction or planned by 2030, according to Cushman and Wakefield. If that pipeline arrives, the data centre connectivity market the company targets will exist. If a meaningful share slips, 6 Pbps of new fibre sits idle against a 25-year asset life.
That is the risk in any build-ahead strategy, and it is why the utilisation figure matters so much: the company is not guessing at demand, it is extrapolating from a network it has already filled. Capacity of that scale has to connect to something, and the connections carry latency requirements as demanding as the compute they serve. Chokshi has noted that customers now ask for 400 gigabit, 800 gigabit and even terabit-scale links rather than the 10 and 100 gigabit connections that were standard.
Geography is Informative
The phase geography is informative. Phase one links the four cities where India’s hyperscale capacity already concentrates. Later phases reach Vizag and Bhubaneswar on the east coast, and Jamnagar and GIFT City in the west. Jamnagar is a refining and industrial complex that has attracted large-scale data centre interest; GIFT City is a financial special economic zone with its own regulatory regime. Both sit where new capacity is planned rather than where it currently runs, which tells you the company expects the next wave of Indian data centres to be built away from the existing three clusters.
The alternative to building
The alternative to building is to buy. The data centre connectivity market has long-distance capacity from several operators, and a hyperscaler can lease wavelengths from more than one. Tata Communications is arguing that leased capacity at the required latency, with the required diversity and the required service levels, is not available at the scale AI workloads will need. That argument is testable, and the test is whether the first contracts come in at the prices the company needs rather than at commodity rates.
What the release left out
The gap between the announcement and the company’s own public statements is unusually wide. The release gives an investment range and a distance. It gives no timeline, no capacity figure, the rupee amount, the market size, the share target, the utilisation rate of the existing network or the funding source.
All of those appeared in interviews the executives gave the same day to BusinessLine, CNBC-TV18, Moneycontrol and Mint. A reader of the release alone would not know whether the network takes two years or ten, or what it will carry when finished. That is not a criticism of the journalists, who asked the right questions and received detailed answers. It is an observation about how the company chose to disclose a 4,350 crore commitment: fully to the business press, thinly in the announcement.
The contracts question
One detail deserves care. Rupesh Chokshi told Moneycontrol the company is in advanced discussions with customers for the highway and has not yet announced contracts for it. Separately, Lakshminarayanan said a growing proportion of the company’s expansion is driven by anchor customer commitments and contracted demand. Those statements are consistent and not identical: the second describes the expansion programme generally, the first describes this project specifically.
For a build of this size the difference matters. A network with anchor tenants on long-term leases is an infrastructure asset. A network built in anticipation of tenants is a bet that the market arrives on schedule.
The subsea half
The highway is one part of a two-part story. Tata Communications plans a new subsea cable between Mumbai and Singapore and holds a consortium position in a Chennai-Singapore system with ready-for-service expected in the fourth quarter of 2029. It integrated the TGN IA2 intra-Asia cable in 2025.
Domestic fibre without international capacity moves data between Indian cities. International capacity without domestic fibre cannot distribute what it lands. The company is building both, and the question is whether the subsea systems arrive alongside the terrestrial network or three years behind it.
What to watch
Three markers. The first is the first contract for the highway, because that turns a commitment into a lease. The second is whether phase one becomes operational on the stated three-year horizon, since a fibre build crossing four states carries permitting and right-of-way risk the release does not mention. The third is utilisation: a network built to 6 Pbps against 1.25 Pbps of current capacity needs a great deal of AI workload to arrive, and the company’s own share target assumes it does. A fourth is whether Reliance, Bharti Airtel and the other holders of Indian long-haul fibre respond with comparable dedicated builds. If they do, the data centre connectivity constraint is being priced correctly. If they do not, and Tata Communications takes the share it targets, it will have bought a position its competitors have to route around.
What it signals
The announcement marks a stage where India’s AI conversation turns from chips to pipes. Compute attracts the attention, but a data centre full of accelerators with an inadequate link to its customers is an expensive building. Tata Communications is betting that connectivity becomes the binding constraint before compute does, and the utilisation figure on its existing network suggests it already has. The bet is large, funded from its own cash, and spread over a decade, which is what an infrastructure conviction looks like rather than a marketing one, and it is the kind of commitment a balance sheet can carry only when the operating cash flow is already there.