Tata Communications said on Monday, 26 October, that it will invest $400-450 million over the next three to five years to build a 10,000 km fibre network linking data centre and artificial intelligence clusters across India, a bet on the connectivity layer that sits between the country’s computing buildout and the customers who will rent it.
A 10,000 Km Grid Built For AI Workloads
The network, which the company calls an AI superhighway, will be built in three phases and will connect hyperscale campuses to cable landing stations before plugging into the company’s global backbone. Ganesh Lakshminarayanan, managing director and chief executive, said the data centre to data centre connectivity market in India could exceed $1 billion over the medium term.
Read more on newzshort:
L&T Tech Wins Five-Year Nokia R&D Deal In India
Capacity is the constraint the company is trying to break. Its present network carries 1.25 petabits per second and runs at 70 to 80 per cent utilisation, according to the regulatory filing, and the new build adds about 6 Pbps. That is close to five times the headroom it has today.
The money comes from operations rather than debt. Rupesh Chokshi, chief product and technology officer, said the capex will be funded through operating cash flows, which stood at Rs 1,562 crore in the first half of FY27 against Rs 4,479 crore for the whole of FY26.
Read more on newzshort:
Jio IPO Subscribed 22.19 Times Ahead Of Listing
Why The Middle Mile Matters Now
India’s data centre count is projected to rise from 118 to 180, carrying 19 gigawatts of IT workload, as generative and agentic AI move from pilots into production. Those workloads need low latency between campuses, and the long haul fibre between cities has been the thinnest part of the stack.
Tata Communications already runs more than 500,000 km of subsea cable and 200,000 km of terrestrial fibre. A Mumbai to Singapore cable announced in June 2026 and a Chennai to Singapore consortium system due in the fourth quarter of 2029 extend the same map offshore, which matters because a large share of India’s AI traffic still terminates abroad.
Owning the glass rather than leasing it changes the arithmetic. Chokshi said the asset has a 25 year life and that capacity will be leased on long term contracts instead of being kept captive, the usual route to owner economics on a multi tenant network.
The Numbers Behind The Bet
At $450 million for 10,000 km, the outlay works out to roughly $45,000 a km, or about Rs 43 lakh. Spread across three to five years it is a manageable slice of cash flow, but it lands before the revenue does. Chokshi expects meaningful financial contribution only around FY28 and FY29.
The rupee figure has been reported at Rs 4,350 crore by the Economic Times, citing the filing, while the company’s own statement puts the spend at $400-450 million. At current rates the two are close, so the gap reflects the exchange rate used rather than a different project.
The base business is steady enough to carry it. For the quarter ended September 2026, Tata Communications reported a profit of Rs 335.6 crore on revenue of Rs 6,704 crore.
Three Phases, And Two Open Questions
Phase one covers the Mumbai, Pune, Hyderabad and Chennai corridor, the route with the densest hyperscale and gateway concentration. Phase two links Hyderabad to Vizag and Chennai. Phase three reaches Bhubaneswar, Bengaluru, Ahmedabad and Jamnagar, and runs into GIFT City.
The timeline is less clear than the route. The company describes a three to five year build, while the Economic Times reported the three phases as running between FY30 and FY36, a window that sits oddly beside the FY28 and FY29 revenue guidance. The two accounts have not been reconciled.
Risks And What To Watch
The clearest risk is demand concentration. A network built for hyperscalers, neo cloud providers and frontier model labs depends on a handful of tenants signing long term leases, and anchor customer commitments are exactly what the company leans on to justify the capital.
Competition is the second. Reliance Jio now has a listed currency for infrastructure spending, and Airtel and the large data centre operators are building the same links. Fibre is not scarce the way spectrum is, only slow and expensive to lay.
The editorial view: this is a sensible, unfashionable investment. It will not produce a headline quarter, and the returns arrive after the depreciation does, but a 25 year asset funded from operations, in a market the company thinks can pass $1 billion, is a better use of cash than another acquisition.
