Aitech Integrates NVIDIA IGX Thor Into S-A2300, Lifting Satellite AI to 5,581 FP4 TFLOPS
Updated
Updated · Interesting Engineering · May 28
Aitech Integrates NVIDIA IGX Thor Into S-A2300, Lifting Satellite AI to 5,581 FP4 TFLOPS
5 articles · Updated · Interesting Engineering · May 28
Aitech said its rugged S-A2300 space computer is being upgraded with NVIDIA’s IGX Thor to sharply expand onboard AI processing for satellites and other edge missions.
The new setup is aimed at Low Earth Orbit tasks such as autonomous navigation, image analysis, climate monitoring and debris tracking, reducing reliance on Earth-based processing and cutting latency.
S-A2300 currently delivers up to 248 TOPS on Jetson AGX Orin Industrial modules; with IGX Thor, Aitech expects up to 5,581 FP4 TFLOPS and 400 GbE connectivity via ConnectX-7 SmartNICs.
Aitech already has IGX Thor evaluation hardware, which it said could shorten design and validation cycles before launch, while Thor’s Functional Safety Island supports mission-critical operations.
The move fits a broader shift toward autonomous satellite networks that process more data in orbit under tight power, radiation and communications constraints.
As data centers expand into orbit, can AI solve the crippling data bottleneck back to Earth?
How will orbital AI supercomputers reshape global strategic competition miles above our heads?
How does a 20-fold power jump in orbital AI change the risk of catastrophic satellite failure?
5,581 TFLOPS in Orbit: Aitech’s S-A2300 AI Supercomputer Redefines Space Autonomy with NVIDIA IGX Thor
Overview
Aitech has launched the S-A2300 AI COTS Supercomputer, marking a major leap in space-based artificial intelligence. By integrating the NVIDIA IGX Thor platform, Aitech’s upgrade delivers up to an 8x boost in performance, bringing unprecedented AI computing power directly to orbit. This breakthrough enables satellites to process complex data and make real-time decisions in space, reducing reliance on Earth-based processing. The S-A2300’s robust, radiation-tolerant design supports advanced applications like autonomous navigation and real-time Earth observation, setting a new standard for intelligent, self-sufficient space missions and opening new possibilities for the future of orbital computing.