Three reasons a charging site is dark, and why nobody publishes which

If you drive an EV in India you have had this happen. The app says the charger is live, you make the detour, and nothing works. If you operate chargers in India you have had the other version. The site sat energised all day and nobody came.

Both of those are usually described with the same word. Broken. It is the least useful word in this industry, because a charger being off is not one story. It is three, and almost nobody publishes which one it was.

One. The supply

The grid went. Or the transformer was sanctioned and never delivered.

This is the largest single cause in everything we have collected so far, and it is nobody's hardware failing. On 2 April 2026 a driver reported that every charger past Satara was down: "all chargers after Satara were not working... due to a power cut." Nothing was faulty. The supply was not there.

The permanent version is worse. Tribune India reported nine charging stations in Dharamsala, installed under a Ministry of Heavy Industries scheme, lying defunct for about three years with no power supply and missing transformers. Twenty-three stations installed in Chandigarh in November 2022 were still non-operational a year later.

Notice that both of those are state-owned. That distinction matters and almost nobody makes it. State-owned chargers in our collection fail permanently. Commercial ones fail on a given day and work the next. Averaging the two into one national percentage is the central analytical error in the existing literature, and it quietly defames commercial operators by association.

Two. The maths

Average utilisation at an Indian public charging site runs at around five percent. For most operators it is under ten. The utilisation needed to cover power, rent and maintenance is thirty to thirty-five percent. Those figures come from ORF Occasional Paper 497 and from NITI Aayog.

Sit with the gap for a second, because it explains almost everything else.

Chargers in India are not failing from overuse. They are rotting from disuse. At five percent utilisation there is no margin to pay a technician with, no margin to keep a spare gun in a van, and a real question every morning about whether today's demand charge is worth paying. A site that is switched off is very often a site that cannot afford to be on.

This is the half of the story that gets left out of every reliability headline, including some we have written ourselves. It is also the most defensible version of the whole argument, and it is on the operator's side.

Three. The choice

And then there is the third kind, which is a decision.

In July 2026 the Free Press Journal reported that a bus depot in Thane had its power cut off over unpaid dues, and that 22 of 63 electric buses could not operate. The operator and the transport body spent the coverage trading blame over who owed what, so we will not pin it on either of them. What we will say is that nobody there was short of a transformer.

A Jio-bp station at Ambur was reported by a driver as "inoperable for the last 2-3 months" in September 2025. The duration is the argument. The grid does not stay off for three months.

In June 2022 a driver arrived at a Tata dealership in Nerul at 8 percent state of charge and was turned away, because the chargers were "reserved for test-drive and delivery cars". The charger worked. That is the point.

There are softer versions of the same category. Paying for fast and getting slow: a Shell Recharge unit at Varve delivered 82 kW against a 120 kW rating, on a car capable of taking 150. A Reliance charger at RR Mall in Kolhapur delivered about 14 kW and, in the driver's words, "charging time doubled." Nothing was broken and nothing was disclosed.

Why the three have to be separated

Not out of politeness. For a much harder reason.

While every dead charger is equally "broken", nothing is anybody's fault. The grid cut and the unpaid bill get the same word, which means the unpaid bill never has to answer for itself. Being straight about the supply is the only thing that makes "this one was a decision" land.

A reliability score with no cause attached is not a measurement. It is an accusation. And it cuts the other way just as hard: the grid does not stay off for three months.

What none of the three change

Here is the part that is true in all three cases, and it is the thing we are actually building.

You arrive. The app said live. The detour is already made. The trip breaks.

And nobody records that you came.

A charging session record only exists if somebody charged. Every driver who arrived, found the site dark and drove on is missing from every report the operator has. Not because anyone is hiding it. Because nothing in the system was ever built to count an arrival that did not become a session.

So the cost of being switched off is invisible on the operator's side of the glass. The power bill is real and visible. The customers are real and invisible. Which means switching off reads as free when it is not, and the decision about when to energise gets made on instinct, against data that structurally excludes the people who could not charge.

That is also why low utilisation and poor reliability are not two separate problems. A failed arrival suppresses demand. Suppressed demand makes the site look not worth energising. The site stays dark, and produces more failed arrivals. Nobody in that loop can see the step where the customer was lost.

The two numbers

There is one missing measurement here, and it reads differently from each end.

For the driver: of the people who arrived intending to charge, how many left with energy.

For the operator: how many arrived while you were dark.

We are building both, and when we publish we will say which of the three reasons it was. Grid took you down, that goes on the record. Nobody paid the bill, that goes on the record too.

Operators will be able to buy their own missed-demand data, because that number does not exist in any report they have. No operator can buy a number in the index, and nobody sees one before it is public, including us moving one because somebody asked. The first Mumbai index publishes on 5 November 2026, with its method.

Sources

• Team-BHP, 2 April 2026, Satara corridor power cut
• Tribune India, 12 December 2024, Dharamsala stations defunct about three years
• Tribune India, 18 November 2023, 23 Chandigarh stations non-operational after a year
• ORF Occasional Paper 497, 8 October 2025, and NITI Aayog, on utilisation and breakeven
• Free Press Journal, 31 July 2026, Thane bus depot power disconnection over dues
• Team-BHP, 30 September 2025, Ambur station reported inoperable two to three months
• Team-BHP, 14 June 2022, charging access refused at a Nerul dealership
• Team-BHP, 13 June 2024, Varve delivered output against rating
• Team-BHP, 2 April 2026, Kolhapur delivered output

Each incident above is one report from one visit on one date. It is not a rating, and we will not present it as one.