Tata Communications Bets $450 Million on AI-Ready Fiber Build

Tata Communications will spend between $400 million and $450 million of its own cash on a 10,000km terrestrial fiber network across India, built explicitly for AI data center traffic rather than the enterprise workloads most of the country's long-haul routes were designed to carry.

The company is solely bankrolling the multi-phase build through its operating cash flows — roughly the cost of a mid-sized subsea cable system, spread instead across land routes connecting India's fastest-growing AI corridors to global cable landing stations.

"Most terrestrial networks in the market were built for general enterprise and consumer traffic. They were not planned around AI data center clusters," Rupesh Chokshi, Tata Communications' chief product and technology officer, told Light Reading. "AI is changing what a network has to carry. Training and inference workloads move far more data between data centers than traditional enterprise or cloud traffic did, and they are far less tolerant of latency."

What does the build cover?

The first phase targets the Mumbai-Pune-Hyderabad-Chennai corridor, which Tata describes as one of India's most important AI and hyperscale growth routes. Later phases will extend east coast and western India connectivity to emerging data center ecosystems in Vizag, Bhubaneswar, Jamnagar, GIFT City and other strategic locations.

On the technical side, the system is designed for total capacity of more than 7,200 Tbit/s, using dense wavelength division multiplexing (DWDM) paired with coherent optics. Tata says the network is being built from the ground up for the scale, speed and latency requirements of AI workloads.

That capacity figure is a design ceiling, not a service launch number, and Tata has not disclosed an official project timeline. According to Submarine Cable Networks, the build will take three to five years, with commercial entry and initial revenue expected in FY2028-2029 and scaling through FY2030-2036. In plain terms: no meaningful revenue from this asset for at least three fiscal years.

Why own the fiber rather than lease it?

The new network will plug into Tata Communications' existing global backbone — more than 500,000km of subsea optical fiber and over 200,000km of terrestrial fiber. Chokshi framed ownership of the full fabric as the strategic point.

"This buildout provides control across subsea, backbone, metro, and data center connectivity as an alternative to relying on leased capacity," he said. "Because we own and operate the fabric end to end, we can design for performance, resilience and data residency together. As US and other global companies expand AI and cloud operations in India, they need reliable and high-capacity connectivity."

The data residency argument matters here: India's data localization expectations make owned, domestic infrastructure a selling point for hyperscalers that leased international capacity cannot match as directly.

Can India become an AI connectivity hub?

Tata positions the investment as infrastructure for India's emergence as a global AI and cloud connectivity gateway. "It will enable global hyperscalers, digital platforms and multinational enterprises to scale their operations in India with confidence as the country enters its next phase of digital transformation," Chokshi said. "The goal is to build the country's infrastructure and readiness for the global AI economy."

The commercial backdrop gives that claim some grounding. India's data center connectivity sector is predicted to be worth more than $1 billion within the next few years, and once operational the new system is expected to generate business there.

The fiber build extends an 16-month run of infrastructure investment by Tata Communications, focused on its submarine assets. In June 2026 it planned a new subsea cable between Mumbai and Singapore and joined as a consortium member in a Chennai-to-Singapore system expected ready for service in Q4 2029. In 2025 it integrated the TGN-Intra-Asia 2 cable into the original TGN-IA system, adding capacity to its intra-Asia network.

The bet, then, is a coordinated one: subsea routes carrying AI traffic to and from India, and a new terrestrial spine moving it between domestic data center clusters — with returns concentrated in the back half of the decade.