Power: The old is undermining the new amid rising electricity demand

India's power sector faces a growing gap between rising electricity demand and investment in conventional generation and grid infrastructure, risking supply shortages during peak hours

In fifteen years, India has doubled power supply, virtually eliminating official shortages, which were once a tenth of peak demand. September 2026, with evening shortages and surging market prices, was an augury of the next challenge. It is tempting to treat this year as an outlier, the product of war, a weak monsoon and heat. That would be a mistake because the shocks exposed a structural problem: Under-investment in the old power sector.   

The old sector has two parts: Non-intermittent generating capacity (NIC) — comprising coal, gas, hydro and nuclear — and distribution capacity, namely the local network that carries power to homes, shops and factories. The new sector consists of dynamic but intermittent power sources: Solar and wind. 

Capacity in the new sector has grown from 13 Gw in 2011 to more than 220 Gw and helped meet rising demand in this difficult year, especially during the daytime. The success of the new sector, and official shortage figures that hide a more sober reality, have obscured the serious under-investment in the old — a gap that will only widen as demand soars from an AI-addled economy as and energy shocks become more frequent. 

Consider generation. Figure 1 plots annual growth in demand against growth in capacity, split into NIC and intermittent sources. Growth in overall capacity consistently exceeded demand growth between 2010-11 and 2016-17 but fell short in most years thereafter. Since 2016-17, demand grew by about half while capacity grew by about a third. The key finding is that NIC capacity grew even more slowly, by less than a fifth. That matters most in the evenings and at night, when solar is of no help: the NIC installed for every 100 MW of evening peak demand has fallen by about a quarter since 2016-17, back to where it stood in 2010-11. 

For some years the system coped well, because plants from the investment boom of the 2000s could be run harder. Coal plants ran at under 60 per cent of capacity in 2016-17 and at about 70 per cent by 2024-25. That was a good use of what had been built. This September there was little slack. In the evenings almost no coal plant that could run was standing idle and, after sunset, supply could not keep up with demand. 

Take distribution next. While the data is less precise, we can estimate investment in the distribution network — what discoms spend each year on lines, substations, transformers and other equipment. Figure 2 plots such investment against underlying sales. From 2020-21 to 2022-23, discoms invested about as much each year, after inflation, as they had in the first half of the 2010s, while selling about 50 per cent more power. Even allowing for the uncertainty, we can say that in the last several years, the level of real investment has fallen behind the growth in power that the network has to carry, and even further behind what a modern system geared to incorporating increasing amounts of renewables will require. Investment has picked up since 2023-24, but has not yet caught up.     

This under-investment is both a symptom of deeper problems and a potential aggravating factor. Discoms invest too little largely because many are loss-making and cash-strapped — where tariffs do not cover costs and subsidies arrive late, little is left for the network. Some discoms show that it can be done. 

Gujarat’s four state-owned discoms rank among the ten best in the power ministry’s latest ratings. In Odisha, a private operator took over the discoms in 2020 and 2021. Within three years the share of power lost or not paid for fell from 31 per cent to 18 per cent, losses were turned into profits, and investment had risen. What these discoms share is not ownership but managers who are left to manage and are held to account. 

 

 

There is a subtle but no less important link from discoms’ weak finances to investment in generation, not in renewables but in NIC. A discom short of money buys as little power as it can in advance and relies on the exchange. In a normal year that is cheap. This September, with little power on offer and prices at their ceiling, discoms that had counted on the exchange had to cut supply. Nor will such a discom sign the long-term power purchase agreements (PPAs) that would insure it against a bad year. And without such contracts, generators, wary of being paid late, are reluctant to build new plants, NIC above all. As it is, investment in NIC is hobbled by the sword of Damocles: The risk that better batteries will leave new plants stranded.   

Moreover, as the new sector expands, investment in the network will have to rise sharply. In an OECD study, the grid cost of each unit of solar and wind power roughly triples as their share of electricity goes from a tenth to three-quarters. 

The network also has to help manage and reduce demand. As demand grows, time-of-day pricing (much higher prices during evening peaks) will become essential, and that requires smart meters. Despite the Centre sanctioning the funds, only about a quarter of the planned smart meters have been installed. 

What next? Part of the framework exists. Guidelines issued in 2023 already require every discom to line up enough capacity to meet its demand. States now need to act on them: Contract for evening power that depends on neither rain nor imported gas, whether from new plants, pumped storage or batteries, and invest in the local network. Coal supply to plants also needs to become more reliable. By mid-September, plants held about a week’s coal, half as much as a year earlier. The Centre can tie its support to reform, starting with hour-by-hour data on power cuts, and back the states that reform first. The binding constraint is political economy, not technology. 

China is becoming one of those rare countries with surplus green power, geared to exploiting the AI revolution, because it has been building both sectors at once: Clean power, and the firm capacity and grids behind it. India has shown with solar that it can build fast. The next step is to bring the old sector up to the capabilities and standards of the new.

Abhishek Anand is visiting fellow, Madras Institute of Development Studies; Praveen Ravi is principal consultant, CEGIS;  

Navneeraj Sharma is an independent consultant; Arvind Subramanian is senior fellow, Peterson Institute for International Economics 

Note: The analysis, not the writing, in this piece is based on extensive use of AI