Tata’s 10,000-km AI Super Highway: What the Rollout Actually Says

What Tata has announced

On 26 October 2026, Tata Communications announced an investment of US$400–450 million to build around 10,000 km of new long-haul fibre for its India “AI Super Highway”. The company describes the planned network as high-speed, low-latency connectivity for data-centre and AI corridors, designed for hyperscalers, AI platforms, neo-cloud providers and enterprises. The announcement says the new network will be integrated with Tata Communications’ global network backbone.[1]

The exchange filing records the capital expenditure as ₹4,350 crore and says it will be financed from operating cash flows. It also lists a proposed addition of approximately 6 Pbps of network capacity, to be added during FY2030–FY2036.[2]

The rollout: one first corridor, then east and west

Tata’s press release confirms a three-phase programme and names the first phase as the Mumbai–Pune–Hyderabad–Chennai corridor. It says subsequent phases will extend into eastern and western India data-centre ecosystems, naming Visakhapatnam (Vizag), Bhubaneswar, Jamnagar, GIFT City and other strategic locations. The release does not assign each later city to a numbered phase.[1]

A more granular phase description was given by Tata Communications’ network-services executive Rupesh Chokshi in a Moneycontrol interview published the same day: phase two would extend connectivity between Chennai, Hyderabad and Visakhapatnam; phase three would reach towards Bhubaneswar and Gujarat, including GIFT City and Jamnagar. The interview also described capital deployment over three to five years, prioritised according to traffic projections, capacity needs and data-centre development.[3]

Three numbers, three different things

Some early coverage compresses the project into a single “completion” window. The primary documents and the executive interview describe different planning markers; they should not be treated as interchangeable dates.

The public materials reviewed do not explain how the three-to-five-year capital-deployment horizon relates to the FY2030–FY2036 period for adding capacity. The latter is a capacity window—not a published construction or commercial-service schedule for Mumbai, Pune, Hyderabad, Chennai or the later locations.[2][3]

How the global backbone fits in

Tata says the new Indian fibre fabric will integrate with its global network backbone. The company also says it intends to link key Indian AI corridors and data-centre clusters with international cable landing stations, so domestic sites can connect to global AI hubs. The announcement does not name the landing stations or publish an end-to-end international route map.[1]

For context, the same company release describes Tata Communications’ existing network as spanning more than 500,000 km of subsea optical fibre and over 200,000 km of terrestrial fibre. Those are company-reported existing-network figures—not the size of this new project. Tata also says it will continue to scale its existing National Long Distance and metro network in parallel.[1]

The practical architecture is therefore a planned domestic long-haul layer connected into a much larger existing international backbone, with metro and national-long-distance expansion alongside it. But “integrated with the global backbone” does not, by itself, reveal which facility connects to which landing station, which paths are physically diverse, or what service a customer can order.

What “AI-ready fibre” does—and does not—tell us

Fibre is the transport layer: it moves data between computing sites, storage locations, cloud networks and international gateways. More route capacity and path options can matter when AI workloads move large datasets, checkpoints or inference traffic between sites. Tata describes the proposed fabric as built for speed, scale, resilience and low latency; these are the company’s design aims, not published measurements for a particular route.[1]

The label does not tell an operator or data-centre customer whether a specific campus will be on-net, whether a diverse second path is available, what latency or restoration commitment applies, or when a circuit can be turned up. Those details depend on the route, network equipment, interconnection points and commercial service schedule—and are not set out in the announcement or its filing.

What to watch in the next company update

• Route-level alignment: published paths, phase boundaries and kilometres assigned to each corridor.
• Build and turn-up milestones: a dated construction sequence and the first locations open for service—not just a broad capex or capacity window.
• Data-centre endpoints: named interconnection points or campuses, including how metro tails and cross-connects are handled.
• Resilience and service specification: physically diverse paths, available bandwidth, latency/availability commitments and restoration terms.
• International path detail: which cable landing stations and backbone routes will connect to the new domestic fabric.

Until those details appear, the most defensible status is simple: a substantial network plan has been announced; the first corridor and later geographic priorities are known; the operational map and city-level timetable are not.

Quick answers