STT GDC states 9 MW of IT capacity for Kolkata across two facilities of 3 and 6 MW. The cards reconcile to the headline exactly — no gap, no scope conflict. The number matters less than the setting. West Bengal publishes roughly 94 MW across four operators — CtrlS 60, NTT 25, STT 9 — and this registry counts 1.08 MW for the entire state. That is 87 to one, the widest gap in this file, and it is a counting rule failure rather than a research gap: every megawatt is published.
1At a glance
2Nine megawatts, reconciled exactly
Two cards, one headline, no gap. After eight STT cities this is the second of only two exact reconciliations.
| City | Headline | Cards | Gap |
|---|---|---|---|
| Mumbai | 133 | 133 | 0 |
| Kolkata | 9 | 9 | 0 |
| Bengaluru | 23 | 22.6 | −0.4 |
| Delhi | 12 | 11.9 | −0.1 |
| Noida | 63 | 67 | +4 |
| Hyderabad | 6 | 4 | −2 |
| Chennai | 68 | 61 | −7 |
| Pune | 40 | 81 | +41 |
Kolkata is also the smallest portfolio in the estate, which makes it the easiest to keep honest — two buildings and one number. There is nothing here for an operator to get wrong, and the operator did not.
The basis is stated as IT capacity, which puts Kolkata alongside Chennai, Noida, Bengaluru and Pune in the group of five cities where STT says what the megawatts measure. Mumbai and Hyderabad say "power capacity" without qualification.
3A state that counts one megawatt
This is the most consequential page in the STT set, and it has almost nothing to do with STT.
| Operator | Published | Counted | Why not counted |
|---|---|---|---|
| CtrlS Kolkata | 60 MW campus, 16 MW DC 1 | 0 | campus aggregate flagged non-additive; the 16 MW child inside it too |
| NTT Kolkata 1 | 25 MW campus | 0 | campus aggregate flagged non-additive |
| STT GDC Kolkata | 9 MW across two cards | 0 | portfolio and both children flagged non-additive |
| Sify Kolkata 01 | 1.08 MW | 1.08 MW | a standalone facility with no parent aggregate |
| State total | ~94 MW | 1.08 MW | 1.3% of published capacity |
One rule produces this: never add a child to its parent. Every metro operator in Kolkata publishes a campus or city aggregate and then the buildings inside it. Both are flagged non-additive, so neither reaches the state total, and only the one operator that runs a standalone 1.08 MW site is counted.
The consequence is that this registry, which exists to measure Indian data centre capacity, reports West Bengal as a 1.08 MW market. Any buyer using that number would conclude the state has no meaningful capacity, when it has a 60 MW campus, a 25 MW campus and a 9 MW portfolio.
West Bengal is not under-researched. Every figure needed to correct the state total is published by the operators and already in this file.
The registry is wrong because of an additivity rule, not because of missing data. Fixing the flag for five records moves the state from 1.08 MW to roughly 95 MW — and the same fix corrects Maharashtra, Karnataka, Tamil Nadu and Telangana at the same time.
4Why the gap exists
Double counting is a real risk, so the rule was right in principle. Here is where it fails.
| Rule | Protects against | Produces |
|---|---|---|
| Do not add a child to its parent | counting a building and its campus twice | a campus and all its buildings both excluded |
| Flag aggregates non-additive | inflated totals from design ceilings | operator portfolios vanishing from state totals |
| Keep unresolved records at zero | counting unverified claims | verified capacity disappearing alongside the unverified |
| Result in West Bengal | — | 1.08 MW counted against 94 published |
The fix is not to relax verification. It is to separate two things the current model conflates: a facility, which is countable, and a ceiling, which is not. A campus claim of 60 MW is a ceiling. DC 1 at 16 MW is a facility. The facility should count; the ceiling should sit beside it as context.
That is a data-model change of perhaps two hours, and it is worth more to the product than any new research in this file. Until it is made, every state total on the public registry is understated, and the four most industrialised data centre states are understated the most.
5Procurement
6Limits and open questions
- No stage on either card. The portfolio is described as operational and expanding, but neither building is dated.
- No racks, floor area or PUE.
- Counted at zero, which is the defect this page documents.
- Single source for both capacities.
- The state total is not corrected here — this record reports the defect rather than silently patching it, so the published number stays traceable to the source file.
- What would close it. Two things: a rack count from the operator, and an additivity fix in the registry.
7Sources
Sources: STT GDC India Kolkata city page, STT GDC India facility pages for Kolkata, STT GDC India corporate site, CtrlS and NTT Kolkata material for the state comparison, and Indian market commentary on the Kolkata data centre market.