AI forecasting for smallholder farmers

For millions of smallholder farmers, a growing season can determine if a family has enough to eat, can afford to keep children in school, or falls into debt. Worldwide, nearly 500 million small family farms⁠ support the livelihoods of more than two billion people, many with little or nothing to fall back on if a harvest fails.

As climate change makes weather more volatile and extreme, smallholder farmers need timely, reliable weather forecasts to make decisions that will protect their incomes and food security. Yet for many farmers in low- and middle-income countries, high-quality forecasts have long been out of reach, in part because accurate weather forecasts have been prohibitively expensive to create and logistically complicated to deliver.

Recent advances in AI are changing that. Weather prediction systems that include frontier AI can now match or exceed traditional models at a fraction of the compute cost, making more localized forecasts available at much greater scale and giving farmers earlier information to protect their crops, incomes, and families. And AI-enabled farmer chatbots and related tools are making farmer engagement more efficient and effective.

That is why the OpenAI Foundation’s Civil Society & Philanthropy team is investing $60 million to significantly expand access to AI-powered weather and crop disease forecasts—bringing AI-powered forecasts to more countries and more farmers, and covering a wider range of weather and crop-disease threats. Delivered through mobile phones and other channels, this information can help millions more farmers decide when to sow, harvest, fertilize, and take action ahead of weather risks.

This $60 million portfolio brings together six extraordinary organizations—the University of Chicago⁠, UC Berkeley⁠, Precision Development⁠, AIM for Scale⁠ (receiving funding through the University of Notre Dame⁠), Digital Green⁠, and The International Maize and Wheat Improvement Center⁠ (CIMMYT). Working in deep partnership with national and local governments and institutions, these organizations aim to support government-led efforts to deliver weather and crop disease forecasts to 100 million smallholder farmers across South and Southeast Asia and East Africa over the next three years.

The cost of weather uncertainty

Smallholder farms produce roughly a third of the world’s food, and major crop losses can mean food security risk for entire countries. Weather is a major driver of income volatility for farmers: a false monsoon onset can lead farmers to plant too early and lose the crop to a following dry spell, while unexpected rainfall at harvest can spoil a crop before it’s picked. A study across six African countries found⁠ that adverse climate events hit about a third of sampled plots in a typical year, cutting national crop production by 29% on average. And NASA satellite data shows⁠ that such climate events are intensifying as the planet warms.

Before AI, accurate weather forecasting relied on numerical models that executed on expensive supercomputers, which means they largely cannot be run by meteorological offices in many low- and middle-income countries. Even where those forecasts existed, they often were not translated into useful information that a farmer can act on.

Now, frontier AI has powered a second revolution in weather forecasting. These forecasts are produced by AI models trained on huge amounts of atmospheric data, and they’re now often as accurate as (or even more accurate than) traditional forecasting models, at a fraction of the computing power. This opens the door to the mass production of forecasts tailored for smallholder farmers across low- and middle-income countries who are trying to decide whether and when to plant, harvest, and fertilize crops.

These advancements in AI unlock multiple approaches that together provide weather prediction for smallholder farmers at an affordable cost:

1. AI-driven forecasting systems help predict weather at a fraction of the cost, time, and compute of systems historically relied upon.

2. The speed and efficiency of AI weather models makes it possible to combine several of the best models to produce more accurate weather forecasts than any individual model alone, with accuracy improving even more as the models ingest local environmental data.

3. AI models can be fine-tuned by meteorological researchers on targets tailored for farmers (like the date that rains begin at a level to enable planting), enabling the sort of decision-relevant forecasting that was previously out of reach.

4. AI-powered voice, SMS, and chatbots deliver forecasts directly to farmers in their own language, and can also be used by the farmers to share direct feedback on what actually happened on their land, if the forecast was correct or not, and if the advice was helpful to address crop threats.

Multiple randomized trials provide promising evidence that when farmers get timely forecasts, they make different, better decisions—shifting what and when they plant, spend, and invest—and those decisions can raise their incomes and cut crop losses (Burlig et al., 2025⁠; Cole, Goldberg, Harigaya & Zhu, 2025⁠; Rudder & Viviano, working paper⁠).

How this work comes together

Our $60 million investment brings together organizations working across the path from forecast to farmer:

• Generating weather forecasts: The University of Chicago and UC Berkeley will support national meteorological agencies to tailor and operationalize AI weather models to create forecasts that are relevant to farmers. Initial forecasts will include monsoon onset and cessation, harvest-related rainfall, medium-range precipitation, and extreme heat. Alongside weather, CIMMYT will invest in its wheat pathogen forecasting capabilities, using AI and NASA’s Earth-observation data to surface outbreaks earlier.

• Delivering forecasts and related useful information to farmers: Precision Development will work with governments and civil society organizations to turn those forecasts into practical advice and deliver it to farmers through trusted channels. Digital Green will integrate improved forecasts into FarmerChat, its AI agriculture advisory app used by over 2.1 million farmers, and feed farmer response back to the forecast teams. Delivery teams will test message design, comprehension, trust, and action through focus groups, A/B tests, and impact evaluations.

• Sustaining this work: AIM for Scale will work with governments and development banks to build the partnerships and financing needed to sustain these services and expand them to more countries. The ultimate goal is durable public infrastructure, including national meteorological and agricultural institutions operating the services, with governments and development banks financing continued delivery.

The Foundation selected these organizations for their leadership in their respective areas of work, track record of effective collaboration with local government agencies and communities where they work, and demonstrated impact and ability to work together on this ambitious initiative. Our $60 million investment builds on and expands this work—bringing in new capabilities such as crop-disease forecasting, connecting forecasting to new trusted channels, and supporting extension to additional countries. The goal of this work is to give 100 million farmers better information when they need it, helping them avoid losses and make investments that can lead to higher, more stable incomes.

Moving forward, the opportunity is to build on this progress with advanced AI models, making forecasts more accurate and localized at scale, and delivering them to farmers through evidence-based methods. There is still much to learn about what works—and weather systems will need independent benchmarking, feedback, local evaluation, and adjustment across countries and crops.

Part of the OpenAI Foundation’s broader work

This investment is the OpenAI Foundation’s second initiative through our Civil Society & Philanthropy work. Our goal is to help ensure that AI’s benefits reach people and communities around the world—especially where better tools, information, and public-interest infrastructure can improve essential services and economic opportunity.

Across this work, we will focus on:

• Putting AI to work in essential services where it can meaningfully improve access and outcomes.

• Giving civil society organizations the tools, expertise, and support to adopt AI responsibly.

• Building shared infrastructure so that successful tools and lessons can benefit the entire social sector.

We will continue to engage with communities, leaders and builders globally as we develop this work. To reach our team contact [email protected]⁠.