The power plant
nobody built.

Millions of batteries, rooftop solar systems and EVs already sit on India's grid. The next power plant probably won't get built. It will get coordinated.

Grid & ESS13 Aug 20267 min read

The constraint that quietly disappeared.

Every power grid in the world was built around one constraint: control used to be expensive, so it stayed in as few hands as possible. A large power plant could be metered and dispatched by a small control room because there were only a handful of decisions to make in a day. Everyone downstream was treated as a fixed quantity of demand you forecast and lived with, not something you talked to.

That constraint is mostly gone. A sensor costs a few dollars. A cellular modem costs less than lunch. Computing that used to sit in a utility's control room now runs in the cloud for a fraction of the cost. It used to be cheap to control a few big things and expensive to control many small things. That maths has flipped. Nobody rebuilt the grid around it; it just filled up, on its own, with things the old design never planned for: rooftop solar, batteries behind the meter, EVs and smart inverters.

What's already sitting on the grid.

India's rooftop solar capacity grew by 7.1 GW in calendar year 2025, a 123% jump over the 3.2 GW added in 2024, taking cumulative installed capacity to roughly 20.8 GW, according to Mercom India Research's Q4 2025 rooftop market report.

A large part of that jump has a name: PM Surya Ghar: Muft Bijli Yojana, the residential rooftop scheme launched in February 2024, with subsidies of up to 40% and a target of one crore households by March 2027. By Mercom's count, residential installations made up 76% of everything added in 2025, a reversal from earlier years when commercial and industrial buyers led.

7.1 GWrooftop solar added in India in 2025
~20.8 GWcumulative rooftop capacity, end 2025
76%of 2025 additions were residential

None of this capacity is centrally scheduled. It comes online and goes offline on the household's or the business's own terms.

Connected is not the same as intelligent.

Most companies in distributed energy stop at "connected": put a modem on the inverter, get telemetry into a dashboard, know the numbers. That is necessary, and it isn't close to enough.

Every battery has two numbers that matter. State of charge is how full it is right now. State of health is how much life it has left, because a battery wears a little every time it is charged and discharged. A connected battery can report both. That is a fact about the present and the past.

The harder question gets asked continuously: given its health, what its customer needs right now, what the grid or market is signalling, and what a few thousand other units in the same fleet are about to do, what should this battery do in the next five minutes? That is not a dashboard feature. That is a different kind of product.

Why dispatch is harder than buy low, sell high.

A single battery is usually juggling jobs that pull in different directions: hold enough charge for backup, because that is often why the customer bought it; avoid unnecessary cycles, because every extra cycle is wear the owner pays for; catch a good price if the market allows; help the grid if asked; feed the customer's own solar first. None of these wins permanently. The right call shifts hour to hour, so the software has to keep re-checking the plan as new information arrives. A fixed rule cannot do that.

India's power market makes this concrete. Supply and demand have to match almost exactly, every second, or frequency drifts off 50 Hz and equipment trips to protect itself. On India's real-time power exchange (IEX), prices in August 2025 averaged around ₹3.38 per unit. On one rain-heavy day that month, a single 15-minute block cleared at close to ₹0.01 per unit. A battery running one fixed rule either misses a swing like that or ends up cycling on noise instead of a real signal.

What India has opened up.

CERC opened ancillary services to storage-based resources in January 2022. A battery providing Secondary Reserve has to respond to a grid signal within 30 seconds and deliver its full committed capacity within 15 minutes: a real control-systems requirement, not a trading strategy.

The Central Electricity Authority's National Electricity Plan (2023) projects India will need 208 GWh of battery storage by 2030 to keep absorbing renewable generation. As of March 2026, 35.8 GWh was under construction, according to a written reply from the Ministry of Power in the Rajya Sabha.

State regulators have started treating flexibility as something a DISCOM has to procure. Maharashtra set a target of sourcing 1.5% of peak demand from flexible loads in 2024, and Karnataka followed with a similar framework in 2025. Tata Power's Mumbai DISCOM ran a pilot at the Bhandup water pumping station, shifting 345 kW of pumping load for three hours a day. Small, but a real result on a real site.

The diesel-shaped opportunity.

Underneath all of it sits an opportunity that doesn't appear in VPP playbooks written for California or the UK. India's diesel genset market was worth USD 995.95 million in 2025 and is projected to reach USD 1,974.42 million by 2034, a 7.03% CAGR, according to IMARC Group. Standby and backup power is the largest use case at 44.2% of the market, and industrial users the largest segment at 48.6%. More than 800,000 telecom towers still run on diesel.

A diesel genset just sits there until it's needed. A networked battery can work while it waits.

Why none of this happens on its own.

Hardware from different vendors often doesn't speak the same protocol without custom integration. Telemetry that is reliable in a metro can fall over in a tier-2 industrial estate with patchy connectivity, and a decision made on stale data is worse than no decision. Handing a remote system authority over a customer's backup power is a trust problem before it is a technical one. And the stakes are different from finance: a mistake isn't a bad trade. It's a battery, under load, doing the wrong thing.

Once the cell and the inverter turn into commodities, and they are heading there fast, the part that's hard to copy stops being the box. It becomes knowing what every box in a fleet should do next, and having earned enough trust from the owners to be allowed to make that call.

That is the order we build in. A FlexiPod has to earn its place first: beat the genset on economics and survive a real Indian site, with the heat, dust and load swings that come with it, before it does anything clever. FlexiTwin comes after that: the layer that knows what a Pod can do, predicts what it will do, and eventually helps decide what it should do. First for one customer's site. Then, if the fleet is large enough, for a grid that never has to build a new power plant to cover an evening peak, because a few thousand Pods were already sitting there, quietly capable of it.

The next power plant in India probably isn't getting built. It is already sitting on rooftops and in the back of buildings across the country, waiting for something to tell it what to do next.

See how the network works.

FlexiTwin and the VPP layer turn sites you already run into one coordinated resource.