ECMWF took part in Adopt AI, the major Paris event dedicated to artificial intelligence (AI), on 25 and 26 November at the Grand Palais. Through a hands-on demonstration of new AI capabilities at a dedicated booth and two speaking sessions, ECMWF, together with representatives of European partner institutions and Member States, showcased the growing role of AI in weather and climate. The activities highlighted how these methods work in practice, introduced the latest developments at ECMWF, including in EU’s Destination Earth (DestinE) initiative, and demonstrated how European collaborations and supercomputing investments make such advancements possible while supporting climate resilience.

Bringing together 25,000 attendees and 500 speakers from industry, public administration, research, and policy, Adopt AI followed on from the earlier “AI Action Summit”, led by President Macron and European leaders, to spark further engagement and discussions about AI across a wide range of sectors.

AI for the planet – showcasing capabilities for weather and climate

It was under the umbrella of the “AI for the Planet” ecosystem that ECMWF contributed to this event, helping audiences understand the role of AI and machine learning (ML) in weather and climate. In recent years, data-driven forecasts for weather predictions have advanced rapidly, and ECMWF has been an active player in this revolution. Operational systems such as the ECMWF’s Artificial Intelligence integrated Forecasting System (AIFS) and the Anemoi software framework,, developed in close partnership with national meteorological services, are already up and running, illustrating Europe’s role in advancing AI methods for weather and climate applications. These advances rely on high-quality datasets, decades of scientific expertise, and the power of EuroHPC supercomputers involved in DestinE.

A masterclass on data-driven weather forecasts

Uncovering the details of this revolution, Matthew Chantry, Strategic Lead for Machine Learning at ECMWF, led a lively masterclass exploring the evolution of data-driven weather forecasts. The session highlighted key milestones in AI-based forecasting and the importance of high-quality datasets, such as the Copernicus Climate Change ERA5 reanalysis, for building these AI-based weather models.

Data-driven forecasts have been shown to be more skilful in certain metrics than physics-based forecasts, for lead times of up to 10–15 days; supporting applications such as weather prediction – including of extreme events, and renewable-energy planning. In addition, data-driven forecasts can run in minutes on a single graphics processing unit (a processor built for highly parallel computations) rather than an hour, using hundreds of central processing units (standard computer processors).

In Destination Earth, ECMWF is extending development of AI-based models, beyond the atmosphere, to other parts of the Earth system, including land, hydrology, ocean and sea ice, and to longer, climate-relevant timescales. The new ML based prototype earth system models are already showing promising results, simulating events with high accuracy. These developments leverage the EuroHPC supercomputers, as well as the Anemoi software infrastructure developed by ECMWF and its Member States.

Discover more about AI-based model developments in DestinE with our recent blog posts on hydrology, wave and land models.

AI for climate resilience: how European and national efforts operationalise AI applications

The capabilities offered by artificial intelligence are especially critical in the current context of increasing climate change impacts and extreme weather events. This is where Europe’s investments in data, computing, and science translate into scalable solutions that strengthen prediction capabilities and enhance society’s ability to anticipate, adapt, and respond effectively to environmental challenges.

Gathered for a round table to discuss this topic, Virginie Schwarz, President and CEO at METEO FRANCE, and Christian Canton, Associate Director of the Barcelona Supercomputing Center (BSC), engaged in a lively discussion during the session “Harnessing AI to Confront Climate Change: Europe’s Next Layer of Weather and Climate Intelligence,” moderated by ECMWF Director General-elect Florian Pappenberger.

The discussion explored how European and national initiatives, including Destination Earth, EuroHPC, the European network of AI Factories, and national efforts, are driving the development and operationalisation of AI-based models for weather and climate resilience.

Reminding the importance of physics-based models and data to build quality AI models for weather and climate, Virginie Schwartz stated: “Weather and climate are great sectors to exploit artificial Intelligence methods because we have all this amazing data we have collected for years.”

Bringing voices from European Commission, national meteorological services, international organisations, and leading supercomputing centers, the panel highlighted the importance of collaborations across Europe to pool expertise, data, and computing resources, demonstrating how shared efforts can accelerate innovation in AI for weather and climate.

Live demonstration: “Forecast-in-a-box”

At the booth, scientists and Machine Learning experts from ECMWF engaged with visitors and explained the Centre’s AI activities, including those carried out under the Destination Earth initiative. The booth featured a hands-on demonstration of the “Forecast-in-a-Box” concept, showcasing how a complete AI-based forecasting system can be run on different computing infrastructures.

The Forecast-in-a-Box is a prototype solution that combines open-source software and AI models from ECMWF and Member States, including Anemoi, EarthKit, AIFS and Met Norway’s Bris. It packages these elements into ready-to-use pipelines, from data input and simulation through post-processing to visualisation, enabling users to run and adapt AI simulations on their own systems or on cloud infrastructure, such as the DestinE platform. This approach shows how AI-based modular forecasting workflows can be shared, adapted and reproduced across a range of computing environments – from central supercomputers to local systems, lowering the technical barriers for using such systems.

Showcased by Harrison Cook, scientist at ECMWF playing a key role in piloting the Forecast-in-a-Box concept, this interactive display gave participants a tangible sense of how AI methods work in practice and sparked lively discussions about AI applications for weather and climate.