General, Aerospace, Defense
Space-certified AI hardware expands options for LEO and deep space missions
As artificial intelligence (AI) and machine learning (ML) move from ground-based data centers to space-qualified electronic warfare (EW) systems, Mercury Systems is expanding its portfolio of radiation-hardened processing platforms for missions across low Earth orbit (LEO), geostationary orbit (GEO), and deep space. According to Mercury Systems, the move to onboard AI processing is reducing reliance on ground stations while enhancing autonomous decision-making for defense and commercial spacecraft.
For manufacturers and program developers, this move reflects continued demand for modular, space-certified computing hardware that combines AI acceleration with radiation hardening and open architecture standards, reducing redesign requirements as mission capabilities evolve.
AI processing allows electronic warfare to reach the ends of the universe
According to the Mercury AI/ML blog, modern spacecraft increasingly require onboard processing to analyze sensor data, detect anomalies, and respond without waiting for commands from Earth. This approach addresses the growing volume of Earth observation, intelligence, surveillance and reconnaissance (ISR), navigation, and electronic warfare data generated in orbit.
As Mercury notes, “AI-enabled electronic warfare systems are becoming an essential component of modern space-based defense operations.” The company says onboard processing reduces latency and limits the need to transmit raw data, allowing spacecraft to continue operating in competitive or communication-denied environments.
The same source identifies several operational applications, including jamming detection, spectrum awareness, autonomous threat response, sensor fusion, and signals intelligence processing.
Modular architecture supports future hardware upgrades
Mercury attributes much of this development to the Modular Open Systems Architecture (MOSA) standard. In its blog, the company highlights its adoption of SpaceVPX, SOSA™, and OpenVPX™ standards to simplify the integration of new processing technologies while avoiding a complete hardware redesign.
According to Mercury, these standards allow customers to “rapidly integrate new AI/ML capabilities” and replace computing elements as mission requirements change.
From a manufacturing perspective, open architecture can simplify long-term component selection by supporting hardware interoperability across multiple suppliers and program lifecycles, especially if missions remain in operation for many years.
Radiation-hardened AI hardware supports autonomous missions
Mercury said modern AI workloads such as convolutional neural networks, transformer models, and real-time inference require significantly more onboard processing power than traditional spacecraft computers.
To address these requirements, the company has identified several technology components.
- Radiation hardened GPU for AI and signal analysis.
- Low-power neural processing units (NPUs) for autonomous inference.
- Heterogeneous architecture that combines CPU, GPU, and FPGA.
According to the company’s analysis, these processing platforms support image classification, signal analysis, autonomous navigation, electronic threat localization, and multi-sensor data fusion while operating under radiation exposure, extreme temperatures, and fault-tolerant requirements.
Mercury writes that these systems “need to be designed with radiation tolerance, thermal management, and fault tolerance in mind.”
Stay ahead of electronics news
Join the electronics industry’s Astute Engineer newsletter
Space certified platform extends from LEO to deep space
The company’s extensive space technology portfolio positions processing, storage, RF electronics and secure computing as core components of space missions from LEO constellations to deep space exploration.
In its AI blog, Mercury highlights the SCFE6933 SpaceVPX board, based on the AMD Versal™ AI Core Adaptive SoC, as an example of onboard computing designed for machine learning inference, beamforming, and software-defined radio applications. The company says the platform is compliant with MOSA principles and supports sensor fusion and mission processing across multiple orbital environments.
Mercury also describes real-world applications across a variety of mission profiles. LEO ISR satellites can perform object recognition onboard, and GEO weather satellites can process large sensor datasets in orbit. Deep space probes, on the other hand, use machine learning for autonomous failure detection and recovery, and communication delays prevent immediate human intervention.
The company’s strategy reflects a broader move toward higher-performance onboard computing that can support increasingly autonomous spacecraft. For organizations planning future satellite programs, the combination of open architecture, radiation hardening, and AI acceleration is becoming a critical consideration in hardware selection, subsystem integration, and long-term platform support.
Franchise Marketing Manager Damien Semple commented, “As AI workloads move onboard the spacecraft, the focus expands beyond processing performance to ensuring long-term access to certified components throughout the program lifecycle. Early engagement with the supply chain and careful component selection will help reduce redesign risk and support production schedules that continue to increase demand for space-grade computing hardware.”
View Mercury Systems’ product portfolio
