STT GDC states 23 MW of IT load for Bengaluru across three facilities. The cards read 0.6, 4 and 18 MW — a sum of 22.6, which is what a rounded headline should look like. After Noida’s 4 MW gap and Pune’s 41 MW gap, this is the portfolio that behaves: the basis is IT load, nothing is hidden under construction language, and the arithmetic closes to the nearest megawatt. It is also 25% of counted Karnataka capacity and contributes nothing to it, because every STT portfolio in this file is flagged non-additive.
1At a glance
2The portfolio that reconciles
Across eight STT city portfolios, this is one of two where the cards and the headline tell the same story.
| City | Headline | Cards | Difference | Reading |
|---|---|---|---|---|
| Mumbai | 133 | 133 | 0 | exact |
| Bengaluru | 23 | 22.6 | −0.4 | rounding, as expected |
| Kolkata | 9 | 9 | 0 | exact |
| Delhi | 12 | 11.9 | −0.1 | rounding |
| Noida | 63 | 67 | +4 | perimeter undefined |
| Hyderabad | 6 | 4 | −2 | short |
| Chennai | 68 | 61 | −7 | only closes if a building from another campus is counted — see PRJ-138 |
| Pune | 40 | 81 | +41 | scope conflict |
A headline that rounds up its cards is not a reconciliation failure. It is what a well-maintained page looks like: the items are published, the total is published, and the two agree to the level of precision anyone would quote.
Two details make this portfolio better than most in the file. The basis is stated as IT load rather than an unlabelled power capacity, and no card is labelled under construction — so unlike Noida, the headline is not quietly including buildings that do not exist.
Publish the items, publish the total, state the basis, do not hide construction inside a current figure. That is four things, and only two operators in this file do all four.
It is also the reason Bengaluru is the easiest STT city to model, and the one where the registry’s counting defect is most obviously wrong.
3Three buildings, a thirty-fold range
A 0.6 MW building and an 18 MW building in the same city portfolio is the widest internal range in the STT set.
| Building | MW | Share of 23 | What it is |
|---|---|---|---|
| Bengaluru 1 | 0.6 | 2.6% | a suite — see PRJ-130 |
| Bengaluru 2 | 4 | 17.4% | the standard small module — see PRJ-131 |
| Bengaluru 3 | 18 | 78.3% | the workhorse — see PRJ-132 |
| Total | 22.6 | 98.3% | against a 23 MW headline |
The Mumbai portfolio ranges 28-fold and this one 30-fold, both driven by the same structure: a legacy city suite at the bottom and a modern hall at the top. Bengaluru compresses the whole history of Indian data centres into three cards.
The 0.6 MW building is the interesting one. It is too small to be a data centre in any commercial sense, and it is still published, still inside the headline, and still counted in the 23 MW that gets quoted. Whether that is honesty or inflation depends on whether the headline claims to be capacity or to be a list of sites.
4Where 23 MW sits in Karnataka
Karnataka counts 90.7 MW across 19 records. STT’s 23 MW would be the second largest share in the state, and it is counted at zero.
| Operator | Counted | Published |
|---|---|---|
| Nxtra | 45.4 MW | 45.4 MW — four cards |
| STT GDC | 0 MW | 23 MW — three cards |
| NTT, Bengaluru 2 and 3 | 16.5 MW | 16.5 MW plus a 67.2 MW campus outside the count |
| CtrlS | 13 MW | 13 MW |
| Sify | 11.8 MW | 11.8 MW |
| Iron Mountain | 4 MW | 4 MW |
| State total | 90.7 MW | at least 113.7 MW once STT is counted |
That last row is the size of the defect in one state: counted capacity is 90.7 MW and published capacity is at least 113.7 MW. And that number is itself a floor, because NTT’s 67.2 MW Bengaluru campus and its 22.4 MW buildings are also flagged non-additive.
Karnataka is also the clearest case for fixing the flag rather than re-researching the market: every figure needed to correct the state total is already published by the operators. The research is done. The rule is what is wrong.
5Procurement
6Limits and open questions
- No stage on any card. None of the three says whether it is operating.
- No racks, floor area or PUE. Density cannot be computed for any Bengaluru building.
- The 0.4 MW difference is unexplained but consistent with rounding, which is the most benign explanation available.
- Counted at zero despite being one of the best-documented portfolios in the file.
- Single source for all three capacities.
- What would close it. A per-building stage and rack count. Two lines per building.
7Sources
Sources: STT GDC India Bengaluru city page, STT GDC India facility pages for Bengaluru, STT GDC India corporate site, Nxtra and NTT Bengaluru material for the market comparison, and Indian market commentary on the Bengaluru data centre market.