NTT's BSE data centre in Fort carries 0.3 MW of critical IT load inside a 1 MW facility. Both numbers are published by the operator, and divided they give a facility-to-IT ratio of 3.33. That is the highest in this registry and 2.25 times the overhead of NTT's own Chandivali hyperscale halls. It is also the smallest record in all 242 — three hundred kilowatts, roughly sixty cabinets at five kilowatts each, about 0.2% of what NTT counts in Mumbai. This is what a legacy exchange-district site looks like when the numbers are finally put side by side: you are not buying efficiency at Fort, you are buying microseconds, and the electricity bill is the price of the location.
1At a glance
2Three hundred kilowatts is a category, not a rounding error
In a registry where one building is 62 MW, a 0.3 MW record looks like a data entry. It is not. It is a different kind of asset, and the difference is visible the moment you divide the two numbers the operator publishes.
NTT does something almost no Indian operator does: it publishes critical IT load and total facility load on the same page. For BSE that is 0.3 and 1.0. For Mumbai 7 it is 25.6 and 38. For Vikhroli it is 2.4 and 6 MVA. For Vikhroli IV it is 3 and 4 MVA.
Put the four pairs on one axis and the pattern is not about engineering quality. It is about scale. A megawatt of chillers, pumps, switchgear, UPS losses and battery rooms is a fixed cost. Spread it across 25.6 MW of IT and it is 1.48. Spread the same plant across 0.3 MW of IT and it is 3.33.
A small facility is not a small version of a big one. Below about 2 MW the fixed overhead stops scaling, so every rentable kilowatt carries a disproportionate share of the plant.
That is why edge and exchange-district sites are sold on latency, compliance and proximity — never on price per kilowatt.
3What 3.33 costs per megawatt
The ratio is not an abstraction. It converts into rupees per megawatt per year, at the registry’s standard assumption of Rs 8 per kWh and 80% average loading.
| Ratio | Facility load per MW IT | Annual energy | Annual cost | Premium vs Mumbai 7 |
|---|---|---|---|---|
| 1.48 — Mumbai 7 | 1.48 MW | 10,372 MWh | Rs 8.3 crore | — |
| 2.25 — Vikhroli, at 0.9 pf | 2.25 MW | 15,768 MWh | Rs 12.6 crore | +Rs 4.3 crore |
| 3.33 — BSE, Fort | 3.33 MW | 23,337 MWh | Rs 18.7 crore | +Rs 10.4 crore |
Read the last column slowly. For every megawatt of rentable IT it operates, a facility at 3.33 burns about Rs 10.4 crore a year more in electricity than a facility at 1.48. At BSE itself the absolute number is smaller — 0.3 MW does not move a national bill — but the unit economics are the worst in the file.
That is the arithmetic behind a fact the Indian market states without ever showing: small data centres cannot be sold on price. They are sold on where they are.
4Fort: what you are actually buying
Fort is the old financial district of Mumbai — the Bombay Stock Exchange, the Reserve Bank, the brokers, and the fibre routes that connect them. A data centre here is not a warehouse for servers. It is a latency instrument.
| Attribute | BSE, Fort | Chandivali, Mumbai 9 |
|---|---|---|
| Critical IT load | 0.3 MW | 62 MW facility, about 42 MW IT |
| Roughly | 60 cabinets | 6,000 cabinets |
| Distance to the exchange | minutes on foot | about 20 km through Andheri |
| Facility-to-IT ratio | 3.33 | about 1.48 |
| Cost per rentable kW | 2.25x | baseline |
| Bought for | latency and proximity | capacity and density |
Both of these are NTT buildings. Both are in Mumbai. Both are counted at full value in this registry, and the registry total of 161 MW is 0.2% BSE and 38.5% Mumbai 9.
The honest way to hold the two together: the 62 MW building serves the workloads that scale, and the 0.3 MW building serves the ones that cannot move. If a tenant needs both, the small site will always be the expensive one, and it is the one that will never be expanded.
5Procurement
6Limits and open questions
- 3.33 is arithmetic, not a published PUE. It assumes the 1 MW facility figure is total load including losses, and that both figures come from the same operating state.
- Rack count is unpublished. The 60-cabinet figure on this page is inferred from a 5 kW assumption and is labelled as such.
- Small-site ratios are not an efficiency verdict. Fixed plant does not scale below roughly 2 MW, so a high ratio at 0.3 MW is expected rather than diagnostic.
- The Rs 8 per kWh tariff is the registry’s standard assumption, not a BSE contract rate. Mumbai HT commercial tariffs have run higher.
- No PUE, no floor area, no cooling specification, no tenant list.
- Relationship to BSE itself is not stated. The facility carries the exchange’s name; whether the exchange is a tenant, the anchor, or a naming convention is not established by public record.
- What would close it. A rack count and a per-cabinet power figure. Two numbers would turn this from a curiosity into a comparable asset.
7Sources
Sources: NTT Global Data Centers Mumbai BSE facility page, NTT DATA global data center locations index, and this registry’s cross-record comparison of NTT Mumbai facility-to-IT pairs.