Organizations have access to a massive amount of satellite data today. However, connecting these different datasets and making use of them effectively has historically been challenging. Last year, Google DeepMind introduced the AlphaEarth Foundations model to address these challenges by combining multiple sources of satellite imagery and data into analysis-ready geospatial embeddings, and we launched the annual Satellite Embedding dataset in Earth Engine and Google Cloud Storage available for each calendar-year.
Over the last year, our partners have used the Satellite Embeddings dataset to transform their workflows. By translating disparate, complex datasets into easy-to-use embeddings, AlphaEarth Foundations has allowed organizations to easily map forest carbon, detect landscape change, and find agricultural facilities, bypassing the need for massive compute or specialized ML expertise.
Building on that foundation, today we are announcing the Private Preview of Custom Satellite Embeddings, powered by AlphaEarth Foundations. Custom Satellite Embeddings are flexible, on-demand embeddings that allow organizations to transition from annual mapping to monitoring and change detection for custom time periods and regions of interest. This new data product from Google Maps Platform complements our global annual Satellite Embedding dataset in Earth Engine and Google Cloud Storage, by providing higher-frequency, on-demand data. Sign-up to express interest in Custom Satellite Embeddings.
The power of AlphaEarth Foundations
AlphaEarth Foundations and our annual Satellite Embeddings dataset simplifies complex remote sensing workflows. By translating multiple sources of satellite data into highly efficient geospatial embeddings, AlphaEarth Foundations reduces manual preprocessing bottlenecks, data overload, and fragmentation.
AlphaEarth Foundations acts as a “virtual satellite” that synthesizes different data types—including optical imagery, elevation, radar signals, and 3D laser mapping (LiDAR)—into a single, unified 64-band image representation. Leveraging these observations allows the model to produce an unobstructed, consistent view of the planet—even in persistently cloudy regions.
Agricultural monitoring using Custom Satellite Embeddings. A 6-step monthly sequence from mid-November 2025 to mid-May 2026 in Meloland, California, USA, capturing crop growth and harvest in the Imperial Valley.
From annual mapping to active monitoring
With Custom Satellite Embeddings, organizations can request custom data summaries tailored specifically to their areas of interest and timeframe, such as quarterly, monthly, weekly, or non-calendar-annual intervals down to a 5-day frequency (based on AlphaEarth Foundations input data sources). This enables businesses and public sector teams to shift from annual mapping to active, rolling monitoring and change detection, making it possible to monitor crop growth cycles, track climate-driven supply chain risks, and evaluate event-based damage after disasters occur.
Components of Satellite Embeddings.
How organizations can apply Custom Satellite Embeddings
Organizations across industries can leverage Custom Satellite Embeddings to solve complex monitoring challenges.
-
Agriculture: Improve crop classification, track growing seasons in near real-time, and assess frost or storm damage. Analyze during the growing season to better inform agricultural finance and market analysis.
-
Public sector: Monitor forest health, track urban expansion, and assess damage and responses following natural disasters.
-
CPG & Retail supply chains: Track climate risks, inform EU Deforestation Regulation compliance, and identify agricultural disruption risks.
-
Utilities & Telecom: Monitor and detect changes in vegetation and water that may affect communication lines infrastructure and large construction areas.
Custom Satellite Embeddings for the Palisades Fire region of Los Angeles, around the time of the fire. Before (left) and after (center) visualized embedding states over the Pacific Palisades, California area (January 2025 fire). The change map (right) uses dot product similarity to isolate the burn scar in bright white.
What customers are saying about Custom Satellite Embeddings
At NGIS, we see Google's Custom Satellite Embeddings as a highly valuable product for fast-moving industries like the forestry sector. The biweekly or monthly cadence has the potential to power near-real-time forestry supply chain maps supporting our clients' sustainable sourcing initiatives. The analysis-ready data saves significant compute and development time, allowing our team to focus on delivering impactful insights faster.Senior Earth Observation Scientist, NGIS
AlphaEarth Foundations is genuinely impressive work and we believe it's the right architectural bet for the next decade of remote sensing.CTO at TerraFort
Woolpert has found Google’s sub-annual custom satellite embeddings to be a game-changer for agricultural and other land-use analytics. Integrating Google's temporal, multi-sensor representations at a 10-meter pixel resolution with freshly tasked 50cm resolution satellite imagery allows us to determine crop types and inform the annual assessment process for state and local government clients. These embeddings provide the essential environmental context needed to augment our human analysts and automate the workflow.Director of Scientific Outreach, Woolpert
Join the Enterprise Private Preview
We are providing early access to customers interested in joining the Preview for Custom Satellite Embeddings. Sign-up to express interest in Custom Satellite Embeddings and visit our website to learn more.
For academic researchers
We are launching an application program offering selected academic researchers a free sample dataset via Google Earth Engine to support their studies. Prior experience with Earth Engine, affiliation with a university or research institution, and an intent to share findings are strongly encouraged. Apply here by September 1, 2026.
Explore our other geospatial AI offerings
AlphaEarth Foundations is part of Google Earth AI, Google’s collection of geospatial models and datasets. Learn more today.
