Xoople and L3Harris team up to build satellites for ‘Earth AI’ 

COLORADO SPRINGS — L3Harris Technologies has struck an agreement with Madrid-based startup Xoople to develop a satellite constellation designed for a still-emerging concept in the space industry: Earth AI.

The companies say the system will go beyond traditional Earth observation, feeding artificial intelligence models with a continuous stream of data about activity on the planet rather than delivering images for human analysis.

Xoople and L3Harris describe this as “a first of its kind” constellation optimized for the AI era where satellite data is treated less as imagery and more as input for machine learning systems.

Earth AI refers to systems that ingest large volumes of Earth observation data — optical imagery, synthetic aperture radar and radio frequency signals — and apply machine learning to extract structured information about what is happening on the ground and that data is continuously updated. Algorithms could query a system for changes over time, patterns of activity or anomalies, with the output delivered as data.

Designing satellites for AI

Xoople operated in stealth for seven years before beginning to outline its plans in 2025. The deal with L3Harris, announced earlier this month, is to build a satellite constellation explicitly designed for AI workflows, meaning they generate continuous Earth data streams optimized for machine learning.

The company raised $130 million in a Series B round, bringing its total funding to $225 million from investors including Nazca Capital, MCH, CDTI, Buenavista Equity Partners and Endeavor Catalyst.

The funding is to support development of the constellation and the data platform.

Fabrizio Pirondini, Xoople’s chief executive, said the AI era requires satellites “capable of monitoring the whole world in real time and ensuring the exquisite measurements that AI requires.”

He said Xoople and L3Harris have spent years developing the technical requirements for  the constellation.

Xoople’s system would combine data from its own satellites with cloud-based infrastructure, including Microsoft’s Planetary Computer Pro, a platform designed to bring geospatial insights into analytics workflows.

Pirondini said the goal is to combine Earth AI data with other data sources to gain better understanding about changes on the Earth’s surface using natural language queries.

The concept is gaining traction across the technology industry.

Google launched an Earth AI initiative, which applies large-scale artificial intelligence to global geospatial data, making physical, environmental and infrastructure information queryable through advanced models.

“We are developing what we call the ‘Earth system of record,’ because it’s basically a data infrastructure that is capable of measuring with precision what’s going on everywhere on the Earth,” Pirondini said in an interview.

L3Harris moves up the stack

L3Harris says the project marks a shift in its role in the space industry.

The company builds  a wide range of remote sensing instruments across multiple sensing modalities, including optical, infrared, radar and signals intelligence payloads.

Carolyn Cossavella, general manager of exquisite Imaging at L3Harris Technologies, said the Xoople constellation is unlike anything the company has done before.

“It’s more than Earth observation. It’s providing information and real time data that you can make decisions on, that AI can build and learn on,” she told SpaceNews. “And in order to do that, Xoople and L3Harris have defined the requirements of what the system must do to provide that level of accuracy and that level of detail that AI can then take and ingest and build on.”

Cossavella said the contract with Xoople “is a very big thing for us.” The company has been a payload provider to government and commercial constellations but this is a step up the value chain, helping to design the satellites and the broader system architecture.