Every commercial electricity connection carries a number that the building almost never reaches. It is called sanctioned load, or on larger connections contract demand, and it is the figure this company is built on. This note explains what the number means, how large we think the stock of it is, how we work one city from the bottom up, why we cut that estimate hard before we use it, and what we measure in a room before we believe any of it.
Sanctioned versus drawn
Sanctioned load is the maximum load, in kW or kVA, that the distribution licensee has agreed to supply to a connection. It is fixed when the connection is granted and it changes only by application. The licensee sizes its transformer, its feeder and its protection around it, and the consumer pays a fixed or demand charge against it every month whether or not the load is ever drawn.
Drawn load is what the meter records: the actual demand, in half-hourly intervals, rising and falling with occupancy, weather and the hour. In an office building it peaks on a humid afternoon with every chiller running and every floor occupied, and it collapses after seven in the evening and at weekends.
The difference between the two is provisioned capacity that earns nothing. It was applied for at design time against a peak the building may reach for a few hours a year. It survived tenancy churn, more efficient lighting, and a shift to hybrid work that emptied floors on Fridays. It sits in the building's fixed charges every month.
Two things are worth being precise about. First, none of this is free electricity. Every kWh a mill draws is billed at the tariff and, in our model, reimbursed to the host at actuals. What is idle is not energy; it is permission, and the copper that comes with it: a transformer, a feeder, a panel, all sized for a load that is not there. Second, the idle margin is a property of a particular building on a particular day, not of a city. We can estimate the city. We can only measure the building.
Why the connection is the scarce part
A new connection of any size is slow, and it is getting slower. In the largest interconnection queue anyone publishes, 233 GW has been asked for, 9 GW has been approved, and 3.9 GW is actually being drawn. The median new grid interconnection runs to 265 weeks. In primary global markets, grid interconnection for new data centre capacity waits four years or more, and up to a decade in the worst markets. Even where a regulator sets a standard of performance, as Maharashtra's does, a connection that needs a new sub-station is allowed a year before it must be made.
None of that queue is anyone's spare capacity. It is the reason the queue exists, and it is why the interesting question is not how to get to the front of it but how to avoid joining it.
An existing connection with an unused margin skips the queue entirely, because nothing new is being asked of the grid. The building was already allowed to draw the load. It simply never did.
How much of it there is
Our bottoms-up census counts roughly 2,000,000 rooms worldwide that already have power, fibre and spare capacity in them, holding on the order of 15 GW of stranded capacity that is energised and already paid for. Reaching it requires zero new grid connections, substations or transformers.
That figure is modelled, and we label it that way everywhere it appears. It is built from building stock by class and by market, load norms per square foot, and an eligibility cut applied at each stage; it is not a survey of two million rooms. What is not modelled is the grading: rooms are graded against the actual screen, we do not publish a count of them, and every one either passed or told us something about why rooms fail. The census is a screening prior. The grading is the evidence, and it is small on purpose because it is real.
One city, worked from the bottom up
To show how the census is built, it is easier to work one city than to describe a method. We use the Mumbai Metropolitan Region because its office stock is published and its load norms are conventional. This is a local derivation and nothing more: the arithmetic is the same in any city where buildings carry sanctioned load they do not draw.
The input is office stock. Knight Frank's H1 2025 review puts the region's office stock at about 169 million square feet. Distribution licensees sanction commercial connections against load norms expressed in watts per square foot, and for office space those norms run from roughly 10 W per square foot for a conventionally serviced building to something above 20 W per square foot for fully air-conditioned, equipment-dense floors. We applied a band of 10 to 24 W per square foot.
169,000,000 sq ft × 10 W per sq ft = 1.69 GW
169,000,000 sq ft × 24 W per sq ft = 4.06 GW
Band: 1.7 to 4 GWThat gives 1.7 to 4 GW of sanctioned commercial load across one region's office stock. The band is wide on purpose: it spans the norms rather than picking one. It is also a floor on the city-wide figure rather than the whole of it, because it counts only offices. Hospitals, diagnostic centres, cold stores, retail, the server rooms of banks and light industrial premises all carry sanctioned load too, and none of them are in the 169 million.
Why one percent
We do not know what fraction of that band is idle. Nobody does; it is not a published statistic, and the honest answer is that it varies from building to building by more than any average would suggest. So we chose a fraction we would be comfortable being wrong about in either direction, and we chose it small.
1% of 1.7 GW = 17 MW → 1,700 mills at 10 kW
1% of 4.0 GW = 40 MW → 4,000 mills at 10 kWEven at one percent, one region holds room for 1,700 to 4,000 ten-kilowatt mills. That figure is not a market size and we do not use it as one. It is a screening prior: a reason to believe the supply side exists at a scale worth building an engine for, before any particular room has been measured. The eligibility cut is real and it is severe. A building needs about ten kilowatts of idle margin after its own peak, a lockable room, a panel with a spare way, air handling that can reject the heat after hours, and an owner willing to sign. Each of those shortens the list.
The assumption is replaced with measurement room by room, as each is graded and scanned. A realised idle fraction by building class is only worth stating once there are enough rooms behind it.
What a scan actually measures
A scan does not take the sanctioned-load figure on trust, because that figure describes the licensee's promise, not the building's behaviour. It measures three things.
The meter. Larger commercial consumers are metered in half-hourly intervals, and the interval data is available from the licensee or from the building's own energy management system. A year is 17,520 half-hours. We take the 99th percentile of demand, not the mean, and we look separately at the after-hours profile, because the mill is most valuable at night and the building's own draw is lowest then. Headroom is sanctioned load minus that 99th-percentile demand, minus a safety allowance, and the mill's ten-kilowatt wall must fit inside it with room to spare.
The thermal margin. A ten-kilowatt mill puts ten kilowatts of heat into a room continuously. The scan measures whether the room can reject it: what air handling exists, what it is rated for, whether it runs after the building's central plant shuts down, and where the exhaust actually goes. We have written separately about why this, and not the electrical margin, turns out to be the binding constraint in most rooms.
The tie-in. Spare ways on the board that already feeds the floor for a 415 V three-phase circuit at twenty amps a phase, with no new connection; a cable route that does not cross a fire compartment; an earthing arrangement a licensed contractor will sign for; and a room with no water service above it. These are the questions a Chartered Electrical Safety Engineer asks, and the scan asks them first, so the answer is known before the design starts.
Only a room that passes all three is eligible, and only then does the one percent become a real number for that building.
Where this leaves the thesis
Mills went to the rivers because that is where the power was. The power now is in buildings that applied for more than they draw, in every commercial district, sitting under a fixed charge. The census says the supply side is large. The scan says whether any particular room is part of it. We hold the census loosely and the scan tightly, and we go by what the meters say.