Google DeepMind’s Farm AI Reaches Six African Countries as FAO Plans Data Integration

Google DeepMind’s agricultural models have started providing insights for six African countries: Kenya, Uganda, Ghana, Rwanda, Nigeria and Zambia. The new footprint moves the technology beyond its India-first origin, while a planned Food and Agriculture Organization integration aims to bring its datasets into country-facing agricultural intelligence tools.

Google DeepMind’s AnthroKrishi team developed Agricultural Landscape Understanding, or ALU, and Agricultural Monitoring & Event Detection, or AMED, for agricultural insights. Their outputs are available through application programming interfaces, or APIs, which allow other software to request results. Google has also integrated the outputs into Google Earth.

A path from India to six countries

The rollout followed an earlier sequence: the models initially focused on India, and their outcomes were shared with trusted testers in Asia-Pacific last year. Google says the ALU layer is among Google Earth’s most popular layers globally; that is a company-supplied usage characterization.

The FAO project is a planned infrastructure test

FAO’s geoAI4stats initiative plans to integrate ALU and AMED datasets into CROPGRIDS. The intended functions are automated crop detection, agricultural maps and faster updates to agricultural statistics for member countries. Google.org is providing $2.5 million through the AI Collaborative: Food Security, but the integration remains a plan rather than an operating result.

India’s deployments show the operating model

In Telangana, the state government has integrated both datasets into its Agriculture Data Exchange, which supports collaborations and innovations intended to benefit more than 5 million farmers. ALU helps reconcile past field-survey findings, while NaPanta uses AMED in applications for early-warning indicators and crop-stress diagnosis.

Terrastack combines the APIs with land records, satellite data, crop activity, climate signals and market data across more than 140 million hectares of Indian farmland. Google lists land-record reconciliation, biofuel procurement and farm-revenue estimates as applications. Separately, Karnataka’s water-information system uses ALU and AMED with local weather and remote-sensing data across 2.6 million hectares of irrigated land.

CarbonFarm uses ALU and Gemini to provide field-level insights for rice farming and support access to carbon credits and climate finance. Its platform operates in 12 countries. Expansion to 20 countries and support for 2 million hectares of low-carbon rice by 2030 remain expectations and an ambition, not completed results.