India knows where extreme weather is coming. It doesn’t know what it will cost

On August 26, a glacier broke away above the Nepal-Tibet border, leaving Nepal with a $4-5 bn reconstruction bill, close to a 10th of its economy. India doesn't need a glacier to face the same problem.

Until recently, a deficient monsoon thinned the harvest and lifted food prices for a season. Today, a week of unseasonal rain takes out infrastructure. A week of extreme heat removes working hours from every factory floor. What was once a sectoral shock has become a general one, transmitted through prices, through power system planning and, in time, through public finances.

Weather has become a macroeconomic input that India has yet to learn how to price. In early June, RBI lowered its growth forecast to 6.6% and raised its inflation forecast to 5.1%, citing energy prices alongside a weak monsoon. Peak demand touched 270.8 GW on May 21, against roughly 180 GW in 2019, even as 8,133 GW hours of solar were curtailed at midday. But a shock that cannot be located cannot be priced.


Monsoon rainfall in 2025 stood at 108% of the long-period average (LPA), while east and the northeast received 80%, the region's second-lowest since 1901, an average that was accurate and described no actual place. This monsoon has made the point more sharply: districts of Upper Assam with little recent experience of inundation are going under as cloudbursts upstream sent Dikhow into the plains ahead of Brahmaputra's peak.

Bharat Forecast System runs on a 6   km grid, finer than any other national service, while Mission Mausam promises 5 km forecasting and household-level warning by 2030. Researchers later examined Sentinel-1 radar data from the Nepal slope and found it had been creeping at 10 mm a month beforehand - a signal that wouldn't have alarmed anybody. Nothing scans thousands of slopes for such slight motion. SERVIR-Hindu Kush Himalaya, run by ICIMOD with Nasa and USAID, was building towards such monitoring when its US funding stopped in 2025.

Remarkable as India's systems are, they are half of what climate intelligence means, which covers not only seeing a hazard approach but also saying what it will cost, to whom it will fall, and who pays. A forecast that cannot be turned into a number has no standing in a budget.

Turning one into a number begins with knowing what stands in its path. Assam can report how many kms of embankment need repair. But no asset-level register shows what sits inside a floodplain or what it is worth. IMD's annual report counts more than 2,690 deaths from extreme weather in 2025, based on media reports and state agencies. A model can infer a building from a photograph, but not a loss that was never written down.

Reinsurance company Swiss Re finds 93% of India's catastrophe exposure uninsured. Pilots are slight, with one heat product examined by Prayas paying ₹1,100, too little to change what a household does. Instruments exist, a UNFCCC paper lists early warning linked to automatic payouts across developing countries. What is missing is the series of defensible trigger must rest on.

The obstacle is not money but architecture. Assam's disaster allocation has risen across 3 finance commissions, from ₹2,541 cr to ₹4,742 cr, and a recommended ₹5,825 cr, weighted 70% on past disaster expenditure and 30% on a risk index. Past expenditure is a defensible measure of recurring need. Weighting is what fails, since a formula anchored so heavily in spending recognises a new geography of risk only after damage accumulates. Nor can Assam's allocation be spent upstream in Nagaland, where much of Dikhow's catchment lies, since disaster finance is organised by state while a river basin is not.

The remedies are institutional, and won't arrive unbidden. Markets left alone will not point AI at public goods. If budgets are to apply forward-looking criteria, those criteria need something to be forward-looking about: an exposure register published at panchayat resolution; a disaster information system turned into an audited loss series recording economic damage alongside lives; money moving on measured indices through a standing budget line along basins rather than boundaries.

Automatic payment is not an argument for relief over prevention but its precondition. A state able to price a flood in advance can justify spending that averts one, and should be judged by the capacity that spending creates, rather than the sums allocated.

India has made its weather legible. The next decade of climate intelligence will be judged by whether it can make the consequences legible too, in registers of what stands exposed and of what was lost. An economy that can price its weather can insure it, budget for it and build against it. One that cannot, will keep discovering the cost afterwards.