Blog

Building the future of real estate development, in public

Notes on regulatory depth, product decisions, and what we learn running real sites through Keystone.

2026-10-21

Why do data-centre cost-per-MW numbers differ so much?

Published data-centre build costs run from under $9 million to over $23 million per MW. What sits inside that range, what is left out, and eight questions to ask before you trust a figure.

2026-10-21

Where does the floor area go between FAR and what you can sell?

A plot's FAR gives counted floor area. Buyers pay for carpet area. How the gap opens, what the RERA definition excludes, and a worked example showing why one point of efficiency matters.

2026-10-21

India's data-centre capacity by city: what the numbers say and what to check on a site

India's data-centre capacity grew from 375 MW to over 1,500 MW in five years. Where it sits, why the sources disagree, and the checks that matter before you pick land.

2026-10-21

How long does it take to get grid power for a data centre?

Why power, not land, decides most data-centre sites. What the research says about demand, queues and delays, and ten questions to put to the utility before you buy a plot.

2026-10-18

What should a development underwriting report disclose?

A practical review checklist for an early development underwriting report: sources, area definitions, cost basis, timing, assumptions and unresolved checks.

2026-10-09

A site can be real before it is ready to model

A clear project stage and a usable site boundary help an early feasibility model show what the available evidence can support.

2026-10-03

What a Feasibility Model Should Show When It Does Not Know

A useful early feasibility model separates verified inputs, working assumptions and unresolved questions instead of turning missing evidence into false certainty.

2026-10-03

Verify the Inputs Before You Trust the Masterplan

A masterplan is only as trustworthy as the site inputs behind it. Three checks we now make before treating an early development scenario as useful.

2026-09-08

How Developers Evaluate a New Land Opportunity

A land opportunity is more than a location and a price. Here's a practical India-and-USA framework for deciding which sites deserve a full feasibility pass.

2026-09-08

From a Land Boundary to a Development Scenario

A site boundary is not a feasibility answer. Here's the decision chain that turns a parcel into comparable development scenarios, with every assumption visible.

2026-09-08

Buildable Area Is Not the Same Question as FAR

FAR tells you how much floor area you're allowed. Ground coverage, setbacks and marginal open space tell you where on the plot you're allowed to put it — and they can bind before FAR ever does.

2026-09-08

Buildable Area Is Not the Same Question as Zoning (US Edition)

Zoning tells you what's allowed on a US parcel. Setbacks, parking minimums and FAR-exclusion rules tell you what actually fits — and each varies by city, not by a national code.

2026-08-11

US Zoning Due Diligence: A Source Hierarchy

A generic building code does not answer a parcel's permitted use, FAR or setbacks. Start with the local ordinance and official zoning map.

2026-08-11

How Early Site-Planning AI Should Be Evaluated

Four questions that separate a real early-stage planning tool from a demo: reproducibility, ground truth, disclosed limits and enforced human sign-off.

2026-08-11

FAR/FSI Is Not a Single Number

FSI/FAR is not one figure per zone. Base FSI, premium FSI, TDR and fungible area can change what is actually feasible; Mumbai's DCPR 2034 shows why.

2026-07-23

"Code-compliant" is doing a lot of work in that sentence — here's what it actually requires

Every feasibility tool claims its layouts are 'code-compliant.' Almost none of them define what that word is carrying. Here's the honest version, and why we don't use it.

2026-07-17

Why Land Selection to Concept Design Takes Months

A founder’s account of the handoffs between land selection, concept planning and feasibility, and how Keystone aims to connect the early decision.