EDGX launches AI computing platform in orbit on SpaceX Transporter-16 using NVIDIA Accelerated Computing
Market
Commercial onboard compute and compute-as-a-service / satellite operators, payload developers and real-time Earth-observation users.
Trend
EDGX reports that two Sterna payloads have been in orbit since March, with one returning operational data within two weeks while the other still awaits host commissioning; two follow-on flights are booked for 2027 and initial compute-as-a-service customers are signed. This is flight evidence with an important mixed result—not proof of a scaled service—and it exposes host-satellite commissioning as a transfer bottleneck.
Tech Highlight
Sterna wraps a commercial NVIDIA Jetson Orin NX with power, thermal, mechanical, radiation and fault-recovery engineering, dynamically scaling between roughly 10 W and 45 W while preserving the CUDA/TensorRT toolchain. Transfer requires measured radiation upset and recovery rates, host interfaces, secure software upload, thermal headroom, workload isolation, downlink economics and an operational boundary between the payload computer, satellite bus and ground controller.
6-Month Outlook
Watch the second payload finish commissioning and EDGX publish real in-orbit workload, power, thermal and fault data before the February 2027 Super Mode mission. Repeatable customer-software execution with bounded recovery would validate an orbital cloud primitive; another delayed host activation would show integration, not TOPS, is the limiting resource.