Tata’s AI Fibre Build Shows What Data Centres Still Need
Matthew Leo · Published October 31, 2026 · World News
Tata Communications plans to spend almost US$450 million on roughly 10,000 kilometres of new long-haul fibre in India. The company’s announcement is framed around artificial intelligence, but its most useful lesson is more basic: a data centre is only as useful as the networks connecting it to customers, other facilities and international routes.
Reuters reported the plan on October 26. Moneycontrol also reported that Tata is targeting a 30 to 40 per cent share of the relevant enterprise connectivity market. That target is a company ambition, not an observed market result.
The build connects the missing layer
AI infrastructure coverage often begins with accelerators and power. Those are constraints, but training and inference also move large datasets between users, cloud regions, storage systems and data centres. Long-haul fibre carries that traffic.
Tata says the project will be built in phases. The first is expected to link Mumbai, Pune, Hyderabad and Chennai. Later phases would add an east-coast route and extend western coverage to locations including Visakhapatnam, Bhubaneswar, Jamnagar and GIFT City.
The route is intended to connect cloud providers, global capability centres, banks and other enterprises to data centres. The announced length and spending describe planned construction, not completed capacity.
More fibre is not automatically resilience
A longer network can add capacity while still sharing physical corridors, landing points or power dependencies with existing routes. Resilience depends on diversity that survives a common failure.
Buyers and planners should ask:
- Route diversity: Are primary and backup paths physically separated?
- Interconnection: Which data centres, cloud regions and internet exchanges are reachable?
- Performance: What latency and service-level commitments apply between specific endpoints?
- Operations: Who repairs cuts, and what restoration times are contracted?
- International reach: How do terrestrial routes connect to subsea cables and cross-border networks?
- Customers: Is demand supported by signed contracts or only projected growth?
The Canadian relevance
Tata’s Indian network does not directly add Canadian capacity. It does give Canadian planners a useful checklist.
Canada has attracted proposals for large AI facilities, including an Anthropic-backed data-centre plan in Nova Scotia. Provinces are also deciding how projects should pay for electricity, as Mapletechie examined in its report on Alberta’s data-centre grid rules.
Those debates should include fibre. A remote site with inexpensive power may face higher costs for redundant routes and interconnection. A facility near a cable landing point may gain international reach but still need diverse inland paths. Capacity announcements should identify the network plan alongside megawatts and accelerators.
Equipment policy matters too. Proposed US restrictions on some Chinese optical transceivers used in AI data centres show that the fibre layer has its own supply-chain and security questions.
What the announcement does not yet prove
Tata has disclosed a budget, approximate route length and phased geography. Public reporting does not yet establish final completion dates for every segment, utilization, customer commitments or the percentage of spending already contracted.
Those details will decide whether the project becomes a durable competitive asset. For now, it is a clear signal that the AI infrastructure race is moving beyond the data-centre fence.
Sources: Reuters; Moneycontrol.
Tags: AI infrastructure, data centres, Optical networking, Tata Communications