Google Research Expands Earth AI Access, Pledges $10 Million as 100 Million Datasets Join Platform
Updated
Updated · Google Research · Oct 14
Google Research Expands Earth AI Access, Pledges $10 Million as 100 Million Datasets Join Platform
3 articles · Updated · Google Research · Oct 14
Summary
Google Research used its Research@ 2026 event to widen early access to Earth AI, adding three new access routes plus Google Dataset Search across more than 100 million public datasets.
Over 70 geospatial tools now power the platform, including RSFM, PDFM and TimesFM, which Google said can automate complex spatial workflows and support forecasting for public health, clean energy and business planning.
$10 million in Google.org funding will back crisis resilience and public health work, including $2 million for the Milken Institute to address infrastructure bottlenecks in housing, broadband, water and energy.
Google also broadened its AI push beyond geospatial tools: Deep Review is opening to the public after helping review 16,000 research papers, while TimesFM and TabFM are set to be integrated into Gemini for business.
Health and disaster-response updates framed the wider strategy, with MedGemma topping 10 million downloads, AMIE tested with 98 patients, and a new flash-flood model trained on 2.6 million flood events.
How does Google's new AI agent map thousands of at-risk people in minutes without relying on traditional language models?
Could an AI predicting disaster zones using virtual satellites accidentally introduce dangerous biases when analyzing data-poor regions?
0.82 Accuracy in Minutes, Not Weeks: Inside Google’s 2026 Earth AI and Geospatial Reasoning Agent Revolution
Overview
In 2026, Google expanded global access to its Earth AI platform, introducing tools like Ask Google Earth and the Geospatial Reasoning Agent API. These innovations, powered by Gemini, allow organizations to quickly analyze complex environmental, demographic, and infrastructure data, shifting crisis management from slow, manual processes to fast, proactive prevention. During the Ebola outbreak in the DRC, these tools enabled health teams to identify at-risk communities and deploy aid in minutes instead of weeks. The platform also supports low-bandwidth regions with compressed data models, making advanced analysis possible even where internet is limited. However, the rollout of AI-generated satellite imagery raised concerns about data authenticity, leading Google to pause the feature after reports of easily created fake maps.