EnduroSat Makes Orbital AI Compute a Standard Satellite Capability with NVIDIA

The collaboration will pre-integrate NVIDIA AI infrastructure into EnduroSat’s serial-production satellites and accelerate flight heritage for the latest generation of NVIDIA architecture

Satellite operators face orbital bottlenecks due to sheer volume of data generated, downlinking and ground processing capability. Processing data where it is collected on orbit closes that gap. EnduroSat, a space infrastructure builder, today announced a first-of-its-kind initiative to make data-center-class AI compute a standard, pre-integrated capability on application-agnostic serial-production satellites with NVIDIA.

The collaboration will progressively pre-integrate NVIDIA’s compute portfolio – NVIDIA Jetson Orin, Jetson AGX Thor, and NVIDIA IGX Thor today, and the Space-1 Vera Rubin Module as it becomes available – directly into EnduroSat’s FRAME satellite buses, which scale from the 150kg-class FRAME S to the 1+ tonnes FRAME Max. This will allow operators to receive a validated orbital AI platform off the shelf, giving them fast, scalable access to advanced AI computing capabilities in orbit. EnduroSat’s patented cableless satellite bus design lets a single technician assemble, integrate, and test a satellite in under eight hours, and its Space Center can build up to two ESPA-class satellites per day.

NVIDIA’s full AI software stack – including NVIDIA CUDA, NVIDIA® TensorRT™, and cuPHY –  can run natively on EnduroSat’s FRAME satellites. Native CUDA support means developers can write and train models on Earth and deploy them to orbit with no code changes, turning EnduroSat’s space service into a ground-to-orbit AI development pipeline, with the bus serving as the secure management layer for secure space-to-space and space-to-ground communications and power.

The partnership creates a rapid path to flight-prove successive generations of compute hardware – from today’s Jetson edge AI platform to the orbital data-center-class NVIDIA Space-1 Vera Rubin Module as it comes online. The first deployment of NVIDIA Blackwell-based Jetson Thor in space is already manifested on an EnduroSat customer mission, Aethero’s Titan, scheduled to launch in 2026.

The intended destination is a standing model in which EnduroSat serves as the space infrastructure provider for NVIDIA-based space computing: NVIDIA supplies the AI compute and software, EnduroSat supplies the satellites and operations, and customers buy orbital AI capacity as a product.