business · 2026-09-06
TCS's ₹70K Cr AI Bet Has One Tenant
Photo: Metropolitan Transportation Authority from United States of America / Wikimedia (CC BY 2.0)
TCS is building 1 GW of AI data centre capacity in Hyderabad for ₹70,000 crore, but only 100 MW is actually booked by OpenAI, the rest is just an option, not a signed commitment.
What exactly does 100 MW booked by OpenAI mean?
OpenAI and TCS's HyperVault signed a multi-year deal for an initial 100MW of AI data centre capacity in India, a firm first phase, not a projection. Alongside it, HyperVault got an option to scale that same partnership up to 1GW, ten times larger, but that expansion depends on future demand and is not a locked-in order.
Why do developers pre-book anchor tenants before building at all?
Hyperscale leasing de-risks new developments by locking in an anchor tenant before construction, which makes a project financeable to project-finance lenders, sovereign funds and REITs. A single large tenant can anchor an entire campus and attract co-location tenants seeking proximity, so the 100MW deal matters more for financing than for revenue today.
How crowded is the race for India's AI data centre capacity?
TCS competes for the same hyperscaler demand as Reliance-Meta's 168MW Jamnagar site (part of a planned 3GW campus), Google's $15Bn Visakhapatnam hub, Microsoft's $17.5Bn commitment, Amazon's $35Bn plan, and L&T's 250MW Dholera project. India's data centre capacity is projected to grow from 1,500MW in 2025 to 7-8GW by 2030 — whether TCS fills its remaining 900MW depends on winning share of that fivefold expansion.
Who will fill the other 900 MW if not OpenAI?
No one is named yet. TCS says the campus will be built in phases based on market demand, and is pitched at frontier AI firms and hyperscalers generally, not a specific second buyer. The wider market has appetite: Microsoft, Amazon, Google and Meta are each pouring billions into Indian AI infrastructure, and Nomura estimates $400 billion has entered India's AI ecosystem in the past year, so demand exists, just not yet contracted to TCS.
Why do developers pre-book anchor tenants before building at all?
Hyperscale leasing de-risks new developments by locking in an anchor tenant before construction, which makes a project financeable to project-finance lenders, sovereign funds and REITs. A single large tenant can anchor an entire campus and attract co-location tenants seeking proximity, so the 100MW deal matters more for financing than for revenue today.
How crowded is the race for India's AI data centre capacity?
TCS competes for the same hyperscaler demand as Reliance-Meta's 168MW Jamnagar site (part of a planned 3GW campus), Google's $15Bn Visakhapatnam hub, Microsoft's $17.5Bn commitment, Amazon's $35Bn plan, and L&T's 250MW Dholera project. India's data centre capacity is projected to grow from 1,500MW in 2025 to 7-8GW by 2030 — whether TCS fills its remaining 900MW depends on winning share of that fivefold expansion.
Source: inc42.com