By Thomas Gorbold
For most people, a forecast being a degree off rarely changes much. If it is 26°C rather than 27°C, you are still wearing a T-shirt. For a power generator deciding how much electricity to produce, however, that degree can be worth millions.
In the United States, weather forecasting is a public good: it is non-excludable and non-rivalrous. Anyone can use a storm warning, and one person’s use does not prevent anyone else from using it. Public goods fill gaps where private businesses cannot profitably provide services that benefit society as a whole, so the government pays for the infrastructure through taxes.
The National Oceanic and Atmospheric Administration (NOAA) operates the satellites, radar, ocean buoys and weather balloons behind American forecasting. It says a five-day forecast is accurate about 90 per cent of the time and a seven-day forecast about 80 per cent. That is remarkably good for something most people receive for free, and expensive to produce. The public system provides the data; private firms sell more specialised forecasts to businesses that need them. Today, a private weather industry worth around $7bn has grown on top of that public data.
NOAA calculates that improving its wind forecasts alone saves the energy industry around $150m a year, because utilities can schedule generation more precisely and hold less spare capacity in reserve. Work with Colorado State University put the cost of under-forecast rainfall at $117m in lost working time. These are just two examples of the savings that better forecasts create across the economy.
The private layer
By ten days, NOAA says forecast accuracy falls to around half. For most people, that makes little difference. For energy companies, weather can directly affect prices. A colder winter increases demand for natural gas and electricity, while a mild one does the opposite. A utility that can anticipate that change can buy cheaper fuel or avoid producing expensive power, while traders can position for the resulting moves in energy prices and other financial markets.
The Chicago Mercantile Exchange has traded weather risk for more than 25 years, allowing utilities and other companies to hedge this exposure using contracts based on heating and cooling degree days. But the value of weather information extends beyond hedging. If weather moves the price of energy, commodities and securities, having a better forecast than the market can create a significant trading edge.
Hedge funds are taking the next step: trying to find an edge in the forecast itself. Bloomberg reported they hired 23 per cent more weather specialists in 2024 than the year before, with the best paid between $750,000 and $1m, against a median of about $93,000 for atmospheric scientists. Citadel started early, hiring the head of a weather-focused trading firm in 2018 along with roughly 20 traders and analysts. It now runs about two dozen weather specialists.
Forecasting is getting cheaper too. Google DeepMind is another sign that weather forecasting is becoming commercially interesting. Its GenCast model can produce a 15-day forecast in minutes, and has outperformed the leading European weather model.
Where should the boundary sit?
NOAA has been buying from the private sector for years. In 2017 Congress instructed it to assess and, where appropriate, purchase commercial weather data. Its first purchase came in 2020, worth $23m, from Spire Global and GeoOptics, and in June 2026 it awarded $7.3m to Tomorrow.io of Boston and $2.7m to Weather Stream of Boulder for satellite measurements of temperature and moisture. NASA, meanwhile, paid SpaceX about $152.5m to launch NOAA’s latest geostationary satellite.
The test is whether the value can be captured. A hedge fund will pay for a forecast it can trade on, and an insurer for a model that prices policies more accurately than a competitor’s. Neither has reason to fund a national radar network at the coverage society has currently, because the benefit is spread across everyone who ever receives a warning. A study commissioned by the National Weather Service put the annual value of America’s radar system at $8.9bn, almost all of it in injuries that never happened and flights that were not delayed.
Weather modelling is now a business in its own right. Fermat Capital Management, a Connecticut fund and the world’s largest catastrophe-bond investor, employs meteorologists to build its own risk models. Bloomberg found one case where Fermat put the probability of a named storm hitting Florida and Louisiana at 2.16 per cent, against 1.6 per cent in the bond’s offering document. It bought only a fifth of the deal, demanding a higher return for the risk. The fund does not need a better forecast than the public one. It needs a better forecast than the market’s.
Hurricane Melissa showed what that machinery can do for a country. In July 2025 the National Hurricane Center signed a research agreement with Google DeepMind. That October, DeepMind said its WeatherNext model helped the centre anticipate Melissa’s rapid intensification and Jamaican landfall five days out. Melissa became the strongest hurricane on record to strike Jamaica, and the country’s $150m World Bank catastrophe bond, which pays out on the storm’s measured central pressure and track rather than on assessed damage, was fully triggered. Private research improved a public forecast, while private capital had already agreed to absorb part of the loss if the storm arrived.
However, a year earlier Hurricane Beryl hit Jamaica without triggering the bond, because the pressure thresholds were not breached even as the government declared a disaster. And Melissa caused $12.23bn in damage and losses, around 57 per cent of Jamaica’s annual output, against a $150m payout.
Private forecasting may well improve the public system. Markets are good at funding information someone can profit from, unlike the current state of forecasting. The observation network also has to operate at national scale. If several private firms owned it, they would have an incentive to focus on the most profitable areas. If one firm owned it, the result would be a private monopoly over infrastructure that everyone depends on.
That is why the boundary matters. Private firms can make parts of forecasting cheaper and create incentives to improve the models built on public data, but the underlying network is still difficult to replace with a private market. The Trump administration has proposed steep cuts at NOAA, though Congress digressed and funded it above the request, and even AccuWeather, a commercial forecaster with everything to gain, opposes full privatisation of the National Weather Service.
The views expressed in this article are the author’s own and may not reflect the opinions of The St Andrews Economist.
Image Credit: Wikimedia Commons