Google.org announced the selection of 17 organizations to receive total funding of $36 million through the Impact Challenge: AI for Science, with the goal of using artificial intelligence to accelerate research related to climate, health, and basic science. The beneficiaries will be able to join a six-month accelerator program, whose participants will receive technical guidance, engineering support, and access to Google infrastructure.
The selection followed an open call that received applications from 30 countries. The projects cover diverse fields, from storm forecasting and fusion energy development to treatment design and understanding genetic changes.
Examples of Innovation in Climate and Energy
AfriClimate AI Limited is analyzing historical weather indicators for dangerous storm systems in Africa, with the goal of increasing the warning time for flash floods from less than three hours to more than six hours. Inria Chile Research Center is building a multimodal foundation model for the ocean by creating a digital twin of Earth’s marine ecosystems.
In the energy field, Max Planck Institute for Plasma Physics, in collaboration with TU Eindhoven, is using physics-informed artificial intelligence to simulate turbulent conditions inside fusion reactors and narrow the gap between simulation and real-world application. University of Texas at Austin is also developing neural-network models to accelerate the design of compact fusion reactors.
Other projects include automated materials discovery at Massachusetts Institute of Technology, the development of interpretable climate models at UC San Diego, improvements to sodium-ion batteries at University of Cambridge, and protein design using natural language at University of Washington for applications including plastic degradation.
Artificial Intelligence in Health Research
AITHYRA Research Institute, in collaboration with EPFL, aims to understand how viruses manipulate cells and to create an AI-designed protein library for discovering treatments for proteins that are difficult to target with drugs. Allen Institute is working to design DNA sequences that target types of brain cells most vulnerable to neurodegenerative diseases, while openly sharing the models, sequences, and results.
Other health initiatives include developing precision chemotherapy at Australian National University, studying the effect of genetic variants on RNA splicing regulation at National University of Singapore, and discovering unknown causes of mutations at New York University Grossman School of Medicine.
University of California, Berkeley is also developing methods for extracting interpretable clinical biomarkers from black-box medical models, while University of Cape Town is creating a platform to map genetic variation in African genes that metabolize drugs. The list also includes engineering models for comparing heart regeneration at University of Pittsburgh and a foundation model for peptide drug design at University of Toronto.
What Is Changing in Practice?
The program does not announce a single artificial intelligence model or a ready-made product. Instead, it directs funding and engineering support toward specialized research projects that may benefit from expanded computing and improved models. The grant’s direct value lies in combining scientific expertise with technical infrastructure, while the final outcomes—such as improved prediction accuracy or access to usable treatments—remain tied to each project’s progress and its ability to move from research to a verifiable application.
The nature of the projects points to growing reliance on multimodal models, interpretable artificial intelligence, and foundation models in fields that require specialized scientific data and experiments. However, the announcement does not specify the funding amount for each organization, nor does it provide timelines for project results or standardized measurement criteria across them.