A second case option, scoped but not built. Kept in the repo because choosing a case is itself part of the work, and a choice is only defensible if the alternatives are written down rather than remembered favourably.
India is building data centre capacity faster than it is building the power to run it. Which segments of the value chain capture the returns, and which absorb the risk?
Same shape as the aviation case: a market sizing, a competitive map, a profit pool across the value chain, and a recommendation about where to deploy capital.
It sits in the AMS real estate and construction sub-sector, which the posting names alongside commercial aviation. The timing is live rather than retrospective. And the value chain genuinely splits into segments with different economics: land and shell, power and cooling, colocation operations, and hyperscale tenancy.
Data availability, which decided it. The best India data centre numbers sit in broker research from JLL, Colliers and Cushman & Wakefield, published as PDFs, behind gates, and not machine readable. The aviation case runs on DGCA, Eurostat, World Bank and OurAirports, all free, all machine readable, all reproducible by anyone who clones the repo. A portfolio project whose figures cannot be recomputed by the reader is a slideshow.
Forecast quality. The headline capacity projections diverge wildly by source, from around 4.7 GW to 8 GW by 2030 depending on who is publishing and who commissioned it. Those are sponsor-adjacent forecasts, and building a case on them means inheriting someone else's assumptions without being able to inspect them.
Framing risk. The most interesting questions in the data centre story are about power procurement, land acquisition and cooling. That is infrastructure operations, and it drifts away from the market strategy and profit pool work the role is actually about.
No Middle East angle. The aviation case has the India-Gulf corridor built into its spine, which matters for a role that names interfacing with Bain case teams from ME offices. The data centre case would have to reach for that connection.
- A machine-readable capacity dataset, most plausibly assembled by hand from operator
disclosures with the same provenance discipline used in
data/data_dictionary.md - State-level power tariff and availability data, which does exist through CEA and state regulators but needs assembly
- An honest position on the forecast spread: report the range across sources and refuse to pick one, in the same way the aviation sizing reports a band and refuses to average
Aviation stays the flagship. This stays a one-page scope note, revisited only if a reproducible capacity dataset appears. Building both would produce two half-cases, and one finished case beats two started ones.