NTT Mumbai 7 in Chandivali is 25.6 MW of critical IT load inside a 38 MW facility. That is a PUE of 1.48 — and it is the only power usage effectiveness figure in this entire registry that can be calculated rather than taken on trust. Every other PUE in the Indian market is an operator claim with no arithmetic behind it. This one is two numbers divided by each other. The building launched in 2026, is operational, and counts in full. Twelve point four megawatts of the 38 go to cooling, losses and plant.
1At a glance
2One number, calculated instead of claimed
The Indian data centre market quotes PUE constantly and publishes the arithmetic almost never. This record is the exception.
| Measure | Figure | Note |
|---|---|---|
| Total facility power | 38 MW | published |
| Critical IT load | 25.6 MW | published |
| PUE | 38 ÷ 25.6 = 1.48 | arithmetic, not a claim |
| Overhead | 12.4 MW | cooling, UPS losses, lighting, plant |
| Overhead as a share | 33% | a third of the power buys nothing computable |
A PUE of 1.48 is a perfectly respectable figure for an Indian facility. Chennai and Mumbai run hot and humid, which loads the cooling plant, and 1.4 to 1.55 is the realistic range for a well-designed air-cooled or hybrid building.
What makes it valuable is not the number. It is that a buyer can now check an operator’s efficiency claim against a building NTT actually runs, and see what a real published pair looks like.
3What PUE 1.48 costs
Efficiency is not an abstraction once it is priced. Here is what the overhead does at Indian industrial tariffs.
| Measure | Figure | Assumption |
|---|---|---|
| Overhead power | 12.4 MW | published and derived |
| Annual consumption at 85% load factor | about 92 GWh | an estimate |
| At ₹8 a kWh industrial tariff | about ₹74 crore a year | an estimate |
| At ₹7 a kWh | about ₹64 crore a year | an estimate |
| At ₹9 a kWh | about ₹83 crore a year | an estimate |
| Per megawatt of IT served | ₹2.9 crore a year | the efficiency dividend |
The reason this matters is that PUE is one of the few places where an Indian data centre can improve its economics after it is built. Cutting PUE from 1.48 to 1.35 on this building would save roughly ₹6 to 7 crore a year at current tariffs, without selling a single extra kilowatt.
Which is why the absence of PUE data across the rest of the file is not a documentation nicety. It is the single largest unmeasured operating cost in the Indian market.
CtrlS publishes 1.40 for its Kolkata campus. NTT’s Mumbai 7 calculates to 1.48. Those are the only two efficiency figures in this file, and only one of them can be checked.
For every other building in this registry, cooling efficiency is unknown.
4Why nobody else publishes it
Publishing facility power alongside IT load is free, and almost nobody does it. The reason is not secrecy.
| Reason | What it means |
|---|---|
| Most operators publish one number | they may not measure IT at the rack and facility at the meter separately |
| A design PUE is not an operating PUE | a design figure can be quoted forever without being true |
| Facility power is commercially sensitive | it reveals the utility contract and the tariff band |
| The two figures often come from different teams | one from engineering, one from marketing |
| Nobody asks | until a tenant with a large load starts negotiating on efficiency |
NTT publishes both for this building and the IT figure alone for the rest. That inconsistency is worth noting: three of the eight NTT Mumbai records carry a facility figure, three carry IT only, and two carry both.
For the registry the consequence is simple. Where both are published, the IT load is what gets counted and the facility figure is held as context. That is the conservative choice — the IT number is smaller and it is the one a tenant can actually use.
5Procurement
6Limits and open questions
- Design or operating PUE is not stated. 1.48 could be either, and the difference matters.
- No rack count or floor area. Density cannot be computed despite the power pair.
- No date precision. "Launched 2026" is the best available.
- Load factor unknown, so the annual energy figures are estimates.
- The three MVA-based records (Vikhroli and Vikhroli IV) are not directly comparable and are flagged as such.
- What would close it. A rack count and an operating PUE. This record is two fields from being complete.
7Sources
Sources: NTT Global Data Centers Mumbai 7 facility page, Data Center Dynamics reporting on the 38 MW Mumbai 7 launch, NTT India corporate site, and Indian market commentary on data centre power tariffs and efficiency.