Google DeepMind is applying its artificial intelligence models across agriculture and food research, from predicting protein shapes to help breed hardier crops to giving smallholder farmers better guidance on what to plant, according to FoodNavigator.com. Ramine Tinati, a director at the company, described the work at the Asia-Pacific Agri-Food Innovation Summit 2026 in Singapore, the outlet reported.
Tinati framed the effort around solving core scientific problems first, arguing that progress there unlocks many practical applications further down the line. FoodNavigator.com reports that the company's research spans three areas: protein structure prediction through AlphaFold, understanding how DNA changes affect biological function through AlphaGenome, and a longer-term goal of building virtual cells that would allow more biological work to be done by computation.
Why it matters
Agriculture is under pressure from less predictable weather, and the crops relied on today may not hold their yields as conditions shift. Tinati told the summit that changing climate is producing wider swings in weather, which means existing crops need to become more resilient to keep producing at current levels or better, according to FoodNavigator.com.
The company's bet is that faster science shortens the path from an idea to a result. Tinati said much of the time in research goes into processing data, running computations and analysing outcomes, and that AI can compress those steps. The stated aim is not to replace scientists but to let them test far more possibilities before spending money on physical experiments.
Protein prediction put to work in plant breeding
AlphaFold addresses a basic biological problem. Proteins are chains of amino acids that fold into three-dimensional shapes, and that shape determines what a protein does. Working out those structures in the lab has traditionally taken months or years per protein; DeepMind's system predicts them computationally instead, and its database now holds predicted structures for hundreds of millions of proteins, FoodNavigator.com reports.
That capability is being turned toward crops. Tinati said the technology is being used to develop plants with stronger resistance to disease, pointing to research at the Sainsbury Laboratory in the UK that could help build plant immunity, according to the outlet.
He also cited work on sorghum, a major food and feed crop, where AlphaGenome tools are being used to map varieties that may cope better with external conditions. Tinati said the results are being released as an open atlas that any researcher can use, with the hope that they will improve the plants or seeds they work with, FoodNavigator.com reported.
Farm-level data and a widening rollout
Separately, DeepMind has been working with smallholder farmers and soil scientists in India, combining limited data from individual farms with broader environmental information to sharpen decisions about what to grow and when. Tinati said the approach links details about individual land parcels with long-term data and factors such as soil condition, weather and a shifting climate, according to the outlet.
The promised benefit is practical: a smallholder would know which seeds to sow, the best time to harvest and when to let a plot rest so that output is maximised. The programme started in India and is now being extended to 11 countries, Tinati said.
Speeding up the scientific process itself
Beyond individual models, DeepMind is examining whether AI can accelerate research as a whole. Tinati described researchers starting with an idea and using the system to design experimental approaches, working through thousands of possible combinations and drawing on scientific literature, including material from other disciplines, FoodNavigator.com reports.
The longer-term vision is for more biology experiments to run in simulation, so that fewer need to go straight into a physical lab, or so that the handover from computer to lab is quicker. Tinati also said such tools can counter human habits in experimentation, allowing scientists to rank and prioritise options systematically, including ones they might otherwise have overlooked.
What to watch: whether the open sorghum atlas and the smallholder programme's expansion to 11 countries produce results that other breeders and farmers can adopt.
