A University of Chicago team has won a $10 million Laude Institute grant to build an AI weather Almanac for agencies serving places where useful local forecasts are difficult to produce. The university’s 5 November announcement describes a free, open-source platform that will help meteorological agencies, governments and NGOs generate, evaluate and deploy forecasts tailored to local decisions.
The practical question is not just tomorrow’s temperature. A smallholder farmer deciding when to plant needs useful information about monsoon timing, while a health official may need warning of dangerous heat. The university says the proposed UChicago–Laude AI Almanac will combine forecasting methods with tools that agencies can use without building a supercomputing operation or assembling a dense observation network.
AI weather Almanac targets locally useful forecasts
The team brings together Rebecca Willett, Pedram Hassanzadeh, Ian Foster and Michael Kremer, spanning AI, atmospheric science, computing and development economics. Its approach blends existing AI and physics-based forecasts, then adapts them to geography, available observations and the decisions faced by farmers, health officials and emergency managers.
The university argues that global temperature and precipitation scores may not reveal whether a forecast will identify monsoon onset at a particular location. Its platform is intended to make that local evaluation and adaptation more accessible, including where observational data and computing resources are limited.
That is a different contribution from simply releasing another global forecast model. TSN’s coverage of MIT and Microsoft’s subseasonal bias-correction research describes another way to improve existing forecasts rather than replace them. The Almanac announcement, however, is about funding and a planned platform, not a newly published peer-reviewed forecast-skill result.
The university reports that over the preceding six months, the team and collaborators reduced the work needed to benchmark and adapt forecasting models from four to five months of expert labour to a single day of expert review. That is a claim about an adaptation workflow. It is not a statement that all weather forecasts previously took months to generate or that their prediction horizon has changed to one day.
Earlier farmer outreach is separate from the new grant
According to UChicago, governments in South Asia and Africa used AI forecasts developed with the team to give more than 50 million smallholder farmers rainy-season timing information over two seasons. The release describes delivery through government agricultural ministries and text messages. That is historical reach attributed to the university, not a promise that this award has already brought in 50 million new recipients.
Laude’s grant announcement places UChicago in the accelerating-science category of four new multi-year labs receiving $10 million each, or $40 million overall. The other winners are Stanford for workforce reskilling, Cornell for civic discourse and a joint UCSF–UC Berkeley team for healthcare. The full $40 million is not a weather-only award.
The weather lab is intended to strengthen the AI and software behind forecast delivery and extend work to more regions and hazards. The free platform is being built; the grant announcement does not establish that every intended agency can already deploy it or that its local predictions have identical accuracy across climates.
Reach, adaptation speed and forecast quality measure different things
The strongest practical case for this project is the gap between having a forecasting method and making it useful to a local institution. Shortening adaptation work could lower one barrier, while an open platform could reduce the need for each agency to build its own tools. Those are plausible routes to wider access, but they are not interchangeable with evidence that a particular rainfall prediction is correct. Delivery, local evaluation and the decision someone makes from a forecast remain separate parts of the chain.
The historical farmer figure and workflow comparison illustrate that distinction. Reach describes information being distributed; an adaptation-time claim describes work done by experts. Neither supplies a general accuracy score, an economic return for recipients or a guarantee that the next region will obtain the same result. Funding gives the team resources to develop the platform, not automatic validation of every future use. The useful test is whether local agencies can assess forecasts against the events and decisions that matter to their communities, rather than treating a large audience or a faster setup process as a substitute for that assessment.
Other forecasting projects occupy different parts of that chain. ECMWF’s AI-DOP prototype explores generating forecasts directly from observations, while China’s Fengqing ensemble system concerns multiple global forecast runs. Neither is a published head-to-head validation of the planned Almanac.
Laude says it expects the labs to ship open software and evidence of impact over the coming years. For the weather team, delivery of locally evaluated tools is the next milestone beyond the grant award.
Sources
- Laude Institute: Moonshots // ONE Lab Winners, 5 November 2026
- University of Chicago: $10-million Laude grant for AI weather forecasting, 5 November 2026
Related stories
MIT and Microsoft Researchers Use Machine Learning to Improve Weather Forecasts Two to Six Weeks Out moves from this platform announcement to a specific published forecasting method, with its authors’ performance claims kept separate.
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
