Google.org awards $36 million to 17 teams pursuing AI-powered science

Longer flood warnings, better disease treatments and advances in fusion energy are the goals—not results delivered by this award. On October 14, 2026, Google.org announced $36 million for 17 organizations through its Impact Challenge: AI for Science, backing research across climate, health and foundational science with funding and an optional technical accelerator.

The selections follow an open call that drew applications from organizations in 30 countries. Recipients can choose to join a six-month Google.org Accelerator offering engineering support, technical mentorship and Google infrastructure. Participation is optional: the offer pairs the grants with access to people and computing resources intended to help teams turn their research ideas into solutions that can operate at greater scale.

Flood forecasts and the physics of energy

Google’s descriptions of the funded projects show several different roles for AI, from spotting atmospheric signals to interpreting laboratory images. AfriClimate AI Limited is studying the atmospheric conditions that precede deadly storms across Africa. Its stated aim is to extend flash-flood warning times, giving the project a concrete measure of success beyond building a model: how much earlier a warning can arrive.

Two other recipients are approaching fusion energy from different directions. The Max Planck Institute for Plasma Physics, working with TU Eindhoven, will use AI informed by physics to model chaotic reactor conditions and narrow the gap between simulations and reality. The University of Texas at Austin is developing neural-network models to speed the design of compact fusion reactors. Both projects target research and design, rather than announcing an operating energy system.

Ocean research also spans distinct approaches. Inria Chile Research Center is building a digital representation of Earth’s marine ecosystems. UC San Diego’s planned AI pipeline would discover understandable climate physics from Earth observations, build a climate model and decide where to collect new ocean observations. That project links analysis with the choice of what scientists should measure next.

Other funded projects focus on materials and sustainability:

• MIT is training AI agents to interpret experimental microscope images on their own, with the aim of accelerating automated materials discovery.
• The University of Cambridge is combining physics-based AI models with microscopy to forecast degradation in battery cathodes, aiming to develop more durable sodium-ion batteries.
• The University of Washington is using natural-language instructions to design proteins for sustainability applications, including proteins intended to help break down plastic.

Making disease research more specific—and understandable

In health research, several recipients are working on the biological differences that can shape treatment. The University of Cape Town is developing an AI platform to map genetic variation in African genes involved in processing drugs, aiming to guide clinical choices and improve safety. Australian National University is leading an international collaboration intended to move chemotherapy away from trial and error toward more precise, equitable treatment.

UC Berkeley’s project tackles a different obstacle: understanding what a medical AI prediction means. It plans to extract interpretable hypotheses from otherwise opaque models and turn those predictions into clear clinical biomarkers—measurable indicators relevant to health. The Allen Institute, meanwhile, is designing DNA sequences to reach brain cell types more vulnerable to neurodegenerative diseases. It plans to openly share every model, sequence and result from that work.