From Satellite Data to Planetary AI: How TESSERA Is Transforming Earth Observation

Large language models transformed how we work with text. Foundation models are now transforming how we understand our planet.

Every day, Earth observation satellites capture petabytes of data, revealing how crops grow, forests change, cities expand, and renewable energy infrastructure evolves. Turning that data into meaningful insights, however, has traditionally required enormous compute resources and task-specific AI models.

That's beginning to change.

TESSERA demonstrates what's possible when world-class research is paired with production AI infrastructure. Our latest press release explores how Vultr and AMD helped support University of Cambridge researchers in bringing planetary-scale Earth observation AI to life.

“Our goal is to democratize access to planetary-scale environmental monitoring. By making TESSERA's embeddings freely available under a CC-BY license and publishing the complete training pipeline, we're ensuring that any researcher, government, or organization worldwide can deploy this technology for their specific conservation needs." — Professor Anil Madhavapeddy, Professor of Planetary Computing, University of Cambridge

A foundation model for the planet

Traditional Earth observation AI often requires building a new model for every application, whether monitoring crops, mapping biodiversity, or detecting land-use change.

TESSERA changes that approach.

Instead of repeatedly processing raw satellite imagery, it transforms years of Sentinel-1 and Sentinel-2 observations into reusable AI embeddings that represent every 10-meter square of Earth's land surface. Researchers can then build new applications using significantly less data, labeling, and compute.

This opens the door to applications such as:

Precision agriculture
Biodiversity and conservation
Land-use change detection
Renewable energy infrastructure monitoring

Scaling from research to planetary AI

Building a foundation model is only the first step. Running it across years of satellite observations covering the entire planet requires AI infrastructure capable of efficiently moving, processing, and analyzing massive volumes of data.

Using Vultr Cloud GPU and Bare Metal infrastructure powered by AMD Instinct™ MI325X GPUs and AMD EPYC™ processors, the University of Cambridge (Department of Computer Science and Technology) generated planetary-scale AI embeddings that can be reused across a wide range of Earth observation applications.

Projects like TESSERA demonstrate that production AI depends on more than just accelerator performance. It requires a complete AI infrastructure stack that brings together CPUs, GPUs, storage, networking, and software to support the entire AI lifecycle.

Example applications include:

Agriculture: Improve crop monitoring, yield forecasting, and precision farming using high-resolution satellite intelligence.

Insurance and risk management: Support environmental risk assessment, underwriting, and claims analysis with up-to-date Earth observation data.

From RAISE to Advancing AI

At the recent RAISE Summit in Paris, Sadiq Jaffer, Assistant Research Professor at the University of Cambridge, showcased how TESSERA transforms satellite imagery into reusable AI embeddings, making Earth observation AI faster, more accessible, and easier to deploy.

As we head to AMD Advancing AI, TESSERA highlights what's possible when open research meets open AI infrastructure, accelerating scientific discovery while enabling practical applications that help us better understand and protect our planet.