A problem of when, not how much

India’s clean power just set a record. According to grid data tracked by CEEW, at 11.46 on the morning of 6 July 2026, more than half of the country’s electricity came from clean sources for the first time ever, even as system demand climbed to 221 gigawatts. It is a genuine milestone, and a real signal of where the grid is headed. But look again at the timestamp. The record fell at 11:46 in the morning, the one hour of the day when such a feat is easiest to achieve, with the sun directly overhead. By nightfall, that clean share would slide back toward a fifth.

A triumph at noon that quietly evaporates after dark: that single arc is the story of India’s power sector today. For most of its history, the central anxiety of that sector was whether the country could generate enough. India has answered that question emphatically. It added capacity at remarkable speed, built a solar fleet of over 157 GW, and clean energy now makes up more than half of the nation’s installed capacity. Yet on an average day those sources supply only about a quarter of the electricity actually consumed. That gap, between what India can build and what it can deliver at any given moment, is the new problem. The binding question is no longer how much electricity India can produce.

Advertisement

It is when it can produce it, and how fast it can change what it produces. Consider the other record India set this year. On 21 May 2026, peak demand touched 270.8 gigawatts, higher than any single moment in the country’s history. But the peak was not the real story. The real story was the speed. At 8 a.m., demand sat at 224 GW. In under eight hours, the grid absorbed an additional 47 GW, more than the entire British grid drew at its highest point in all of last year. India’s appetite for power is now not merely large; it is violently volatile within the span of a single day.

The price of electricity told the same story, only sharper. On the Indian Energy Exchange’s day ahead market, a unit of power scheduled for delivery at 1 p.m. cleared at Rs 1.56. The same unit at 6.30 p.m. hit Rs 10, the market’s ceiling. Six times the price for identical electricity, a few hours apart. And note the twist: the demand record at 3.45 p.m. was not the costliest hour, because the sun was still up. Prices spiked only once solar faded. Peak demand and peak price have decoupled, and the widening gap between them is now a defining feature of the Indian grid.

Grid operators have a name for the new shape of demand. In summer, solar carves a deep midday trough into the load that conventional plants must serve, followed by a steep evening climb as the sun withdraws: the “duck curve.” India now runs a pronounced duck. In winter, with a weaker sun and two separate demand peaks, the same grid traces a “double humped camel.” Across three summers, the belly of the duck has barely moved. Solar now supplies almost all of the extra midday demand. Yet the evening ramp that thermal and hydro plants must climb has doubled, from about 36 GW to 74 GW.

The grid is being asked to swing harder and faster every year. The clean energy peak of 6 July is simply the brightest point on that curve, the top of the duck’s back, and it says nothing about the hours when the duck’s neck rises after sunset. Three signals confirm that the stress is real, and that it lives after dark. Prices: the intraday spread in May has widened to a peak-to-trough ratio approaching nine. Curtailment: On an average May day, around 24 GWh of clean solar power was simply thrown away because the grid could not absorb it at noon, enough to power more than a quarter of Delhi for a full day.

Shortage: Across April and May, the grid fell short of demand on 36 of 61 days at the night time peak, but on only six days during solar hours. India is not running short of electricity in the middle of the day. It is running short when the sun is not shining. Here is where the conversation usually ends, in a technical debate about batteries and tariffs. It should not. Because this problem collides directly with India’s most ambitious economic bet: artificial intelligence. Every discussion of the AI race fixates on semiconductors, frontier models and engineering talent. Almost none of it reckons with the physical substrate beneath them: electricity, and specifically firm power around the clock. India’s clean energy record was set at 11.46 a.m.

An AI data centre draws just as hard at 11.46 p.m., when that clean share has collapsed and the grid leans back on coal. It needs uninterrupted, high-quality power at every hour of day and night, which is the capability India is shortest on. The numbers are converging ominously. India’s data centre capacity is projected to grow from roughly 1.4 GW today to around 9 GW by 2030, requiring an additional 40 to 45 terawatt hours of electricity. In the first quarter of 2026, India lost some 300 GWh of already generated renewable p ower b ecause the transmission network could not deliver it where it was needed.

The country is, in the same breath, wasting clean power at midday and straining to serve firm demand at night. AI sits squarely in the second category. This makes energy adequacy a question of sovereignty, not merely engineering. If India cannot supply firm, affordable power for AI at scale, that computing will migrate to data centres abroad, most of them owned and operated by entities outside India. Data governance, national security, and the very terms on which Indian institutions access AI would follow the electrons offshore. The uncomfortable truth is that India’s path to AI leadership runs not through its chip design labs but through its power grid, and more precisely through the hours after sunset.

The remedy is conceptually simple: store energy when it is abundant and release it when it is scarce, and use time-of-day tariffs and demand response to nudge load into the solar hours. California, where the term “duck curve” was coined, now runs a battery fleet large enough to convert an evening swing of nearly 28 GW into about 10 GW. India’s fleet is nowhere close. Halving the evening ramp on a single summer day would require roughly 130 GWh of storage discharge; the country’s entire pumped hydro and battery fleet discharges under 24 GWh a day. The shortfall is overwhelmingly a battery shortfall. Encouragingly, the regulatory foundation is being rebuilt.

The draft Electricity (Amendment) Bill, 2025, read with the draft Electricity (Rights of Consumers) Amendment Rules, 2026, would for the first time write energy storage into law, empower the regulator to deepen power markets through instruments such as contracts for difference, convert the obligation to buy renewable power into a broader clean energy mandate backed by penalties, introduce demand response, set firm deadlines for time of day tariffs, and even enlist rooftop “prosumers” to bank their midday surplus for the evening. The direction is right, and it deserves support.

But it carries a caution that cannot be repeated often enough: more solar without commensurate storage does not ease the grid’s stress. It deepens it. Which brings us to the real failure, one of coordination. The Draft National Electricity Plan, which will shape the sector to 2047, contains no dedicated assessment of AI-driven electricity demand. The IndiaAI Mission and the Ministry of Power are, in effect, planning the country’s compute future and its power future in separate rooms. Yet the reforms that would smooth the duck and the camel, namely storage, flexible markets, demand response and time-of-day pricing, are the very same reforms that would underwrite firm power for AI.

They are not two problems. They are one. India spent a decade proving it could generate enough electricity. The next decade will test something harder: whether it can deliver the right power at the right hour. The headlines will rightly celebrate the day clean power first crossed half. The more important question is what happens at 11.46 p.m. Get it wrong, and the country will keep spilling sunshine at noon while its data centres, and its AI ambitions, go looking for electrons elsewhere. Get it right, and the same grid that meets the evening peak will quietly power the country’s climb up the AI ladder. The question is no longer how much. It is when.

(The writer is a tech and social entrepreneur and Programme Director, Eastern India, at WHEELS Global Foundation. He writes on AI, economy and geopolitics, and tweets as i pravinkaushal.)