Tata Communications to Lay 10,000km of AI Fibre

Tata Communications plans to build about 10,000km of new long-haul fibre across India to link the country’s main data centre and AI hubs, at a cost of USD 400m-450m. The first phase will run along the Mumbai-Pune-Hyderabad-Chennai corridor.

The company set out the plan, which it calls India’s “AI super highway”, in a release from New Delhi and in posts on X. It describes the investment elsewhere in the release as “close to USD 450 million”. It gave no completion date.

Mumbai to Chennai first

The first phase covers the corridor that Tata calls one of India’s most important routes for AI and hyperscale growth. Mumbai is one end of it, and new capacity is still being added there, such as Colt DCS and RMZ’s 64MW Mumbai 3, which is under construction.

Later phases will reach data centre clusters in the east and west of the country, including Vizag, Bhubaneswar, Jamnagar and GIFT City, the financial centre in Gujarat. The network will also connect these clusters to international cable landing stations, where subsea cables come ashore, and to Tata Communications’ global backbone. Tata says this will give seamless links between India and major AI hubs abroad, making the country a “strategic gateway” in global digital infrastructure.

The company says the new routes will be “one of the country’s largest AI-ready connectivity network[s]”. In parallel, it will keep expanding its existing national long-distance and metro networks for current customers.

Built for hyperscalers and neoclouds

Tata says the network is aimed at hyperscale cloud companies, neocloud providers that rent out GPU capacity, digital businesses, Global Capability Centres (the offshore operations multinationals run in India) and enterprises. A later post on X lists six customer segments, including data centres, security platforms and content delivery networks. The pitch to all of them is the same: lower latency, more capacity, greater resilience and faster data movement between sites.

“The AI economy needs more than compute. It needs a network built for speed, scale and resilience,” said Rupesh Chokshi, Tata Communications’ chief product and technology officer and global head of network services. He said the company calls the new network its AI Network Fabric, and that it would give global customers “the ultra-low-latency, high-capacity and resilient connectivity they need to scale AI with confidence”. “By connecting India’s AI and data center hubs to the world, we are enabling our customers to grow.”

The release links the plan to growth in India’s data centre industry, driven by cloud adoption, AI, rules requiring some data to be stored in the country, and business demand for resilient capacity. It says connectivity is becoming “as critical as compute power and energy availability”.

Fibre is the next bottleneck

Most attention in the AI build-out goes to chips and power. Fibre matters because training and inference spread across several sites need fast, reliable links between them, and new data centres are going up away from the traditional hubs. India’s sites cluster around a few cities, and the push towards Jamnagar and Vizag shows capacity spreading out.

The timing also matters for supply. Optical fibre prices have jumped as AI demand outruns production, which could raise the cost of a 10,000km build or slow it down. Tata has not said how long each phase will take or how the budget splits between them.

There is a competitive angle too. Network operators are racing to offer AI customers links between their compute sites, as CoreWeave did by extending its own backbone into Asia. Tata’s advantage is that it already runs a large global network and its own landing stations. Long-haul fibre and cable landing points are also attracting more attention as critical infrastructure, as shown by UK proposals on subsea cable sabotage.

The first sign of progress will be a completion date for the Mumbai-Chennai route.