This Advisor series explores the emergence of on-orbit data centers — space-based platforms that enable real-time, AI-powered data processing and analysis directly in orbit. In Part I, we examined the enabling technologies, functionality, benefits, and transformative potential of these systems for supporting autonomous space operations, Earth observation, defense, manufacturing, and beyond. Here in Part II, we focus on the current state of on-orbit data centers, including:

• Countries and organizations leading their development and deployment

• Commercial offerings

• Projects planned and underway

Countries, Companies & Organizations Leading the Way

Different nations and their respective space agencies and commercial enterprises are taking different approaches to implementing orbital data processing and AI infrastructure. These range from complex, on-orbit data centers to real-time, satellite-based Earth observation analytics and autonomous control systems. This Advisor provides a country-by-country overview of the major on-orbit data center and space AI computing initiatives, including major players, capabilities, and goals. This is not an exhaustive study; however, we do feel it provides a good overview of the types of projects and commercial developments in this exciting area.

US

The US is seeing considerable activity around on-orbit data centers and space-based AI computing initiatives.

Axiom Space

Axiom Space is building a network of orbital data center (ODC) nodes in low-Earth orbit (LEO) to deliver secure, scalable cloud computing and AI/ML (machine learning) capabilities directly in space for defense and commercial use. ODCs are designed to operate independently of terrestrial infrastructure, enabling real-time data processing for satellites, spacecraft, and defense systems. Axiom plans to launch two ODCs this year.

Key features include:

• On-orbit data processing for real-time exploitation and dissemination of satellite data

• AI/ML and large language models (LLMs) for autonomous decision-making and threat detection

• Cybersecurity with Earth-independent endpoint detection and response

• Optical intersatellite links for high-speed data transfer

• Interoperability with Space Development Agency (SDA) standards for defense mesh networks

Use cases include national security operations involving multi-sensor fusion, space threat tracking, autonomous satellite control, and commercial applications like Earth observation, disaster response, and industrial analytics.

ODC nodes will integrate with the Kepler Communications optical relay constellation, a next-generation space infrastructure designed to create a real-time, high-throughput, secure communications backbone for satellites and other space assets.

Axiom & Red Hat

Axiom has partnered with open source software provider Red Hat to develop a prototype data processing unit — AxDCU-1, which was delivered in August to the International Space Station (ISS) aboard SpaceX’s 33rd commercial resupply services mission for NASA. This effort, sponsored by the ISS National Lab, will test the feasibility of providing increased data storage and real-time data processing in space — capabilities crucial for commercial space stations.

Key AxDCU-1 components include Red Hat Device Edge, a lightweight Kubernetes platform for handling hybrid cloud workloads in resource-constrained environments; automated rollback and self-healing capabilities for detecting and recovering from system failures; and AI/ML workloads for supervised autonomy, cyber-intrusion detection, and space weather analytics.

In addition to enabling autonomous spacecraft operations, advanced computing could allow astronauts to troubleshoot spacecraft operations and equipment, perform predictive maintenance, and monitor astronauts’ health (via sensor-equipped spacesuits that send data such as heart and breathing rate to a central location). If an anomaly is detected, edge computing would initiate predictive AI models to determine an appropriate response. This effort marks a significant advancement in orbital edge computing, allowing real-time decision-making without relying on Earth-based data centers. It’s also a testbed for future deployments on Axiom’s commercial space station project, Axiom Station.

Starcloud

Starcloud aims to train large generative AI (GenAI) models in orbit using Nvidia graphical processing units (GPUs) and eventually scale to gigawatt-level compute capacity. Its first commercial satellite, Starcloud-2 (with a launch target of 2026), implements a GPU cluster, persistent storage, and proprietary thermal/power systems in a smallsat form factor. Supporting infrastructure includes micro data centers designed for real-time, high-volume data analysis directly in orbit.

Starcloud is also developing a hypercluster architecture, a future iteration planned to scale massively when SpaceX Starship-class and other super-heavy launch vehicles become commercially available.

Starcloud aims to become the first company to train an LLM entirely in space and hopes to offer secure, Earth-independent data storage and compute for governments and enterprises.

Orbital AI clusters represent a new paradigm in compute infrastructure. If Starcloud (and other companies) can pull this off, benefits could prove significant:

• Resilience — immune to terrestrial outages, cyberattacks, and geopolitical disruptions

• Scalability — unconstrained by land, water, or permitting constraints

• Sustainability — powered by the sun, cooled by space, and potentially carbon-neutral

Satlyt: Decentralized Space Computing

Satlyt is one of the more intriguing players in the emerging field of decentralized space computing. It aims to democratize access to space computing and reduce the environmental and logistical burdens of terrestrial data centers, including reducing emissions by shifting compute workloads to solar-powered satellites and minimizing latency for time-sensitive applications like disaster response and autonomous systems. It also seeks to create a sovereign, resilient infrastructure for governments and enterprises wanting Earth-independent data processing.

Satlyt is developing a space computing platform to transform satellites into virtual cloud nodes — essentially orbital data centers. By interconnecting satellites from different operators into a federated satellite system, it seeks to create a distributed computing mesh across LEO.

Key features include:

• Inter-satellite network (ISN). A mesh of satellites that share compute tasks and data, reducing latency and bandwidth strain.

• AI edge processing. Satellites run AI/ML models directly on board, enabling real-time analytics for Earth observation, defense, and industrial applications.

• Tools. These tools can be used to build custom space applications for the ISN.

• Compute monetization. Satellite operators can lease excess compute capacity to third parties, creating additional revenue streams.

NASA-HPE Spaceborne Computer Project

Through their Spaceborne Computer program, NASA and Hewlett Packard Enterprise (HPE) have collaborated on radiation-hardened computing systems since 2017. These efforts have evolved into the Spaceborne Computer-2 and -3 (introduced in 2021 and 2024, respectively), which allow astronauts to run sophisticated AI and ML models on the ISS. This architecture is designed to simulate terrestrial high-performance computing environments but adapted for the harsh conditions of space.

In April, Meta and Booz Allen Hamilton deployed Meta’s Llama 3.2 LLM aboard the ISS as part of the “Space Llama” initiative. Running on HPE’s Spaceborne Computer-2 equipped with Nvidia GPUs, the project aims to allow astronauts to run GenAI workloads in a space environment, thereby reducing reliance on Earth-based computing. Astronauts can access real-time insights, troubleshoot onboard anomalies, and replace paper-based documentation with AI-expert assistance.

Beyond LLMs, Spaceborne Computer-2 has supported over 24 experiments, ranging from natural disaster modeling and 3D printing to healthcare analytics and image-based anomaly detection. These experiments demonstrate how edge computing in orbit can accelerate time-to-insight from months to minutes — a critical advantage for missions where bandwidth is limited and latency matters. They are helping lay the groundwork for autonomous, AI-driven spacecraft and habitats — a critical capability for lunar bases, Mars missions, and orbital stations.

EU

Rather than directing its efforts at massive ODCs, the European Space Agency (ESA) places a strong emphasis on AI and edge computing, focusing on developing modular, intelligent satellite platforms that can autonomously process and analyze Earth observation data in space.

Φsat-2: ESA’s Flagship AI-Enabled Satellite

Φsat-2 (“PhiSat-2”), which launched in August 2024 aboard SpaceX’s Transporter-11, is ESA’s most advanced project involving onboard AI and edge computing for Earth observation. It employs a multispectral camera, AI accelerator, and payload pre-processor to support various onboard AI applications, including deep image compression, cloud detection, street and vessel detection, and fire and marine anomaly detection. Φsat-2 has a full data-processing pipeline on board, enabling data to be first transformed into usable insights before transmission to Earth.

The Φsat-2 mission showcases how satellites can autonomously interpret complex scenes, prioritize data transmission, and reduce latency for various time-sensitive applications, including:

• Climate monitoring — faster insights from multispectral and thermal imagery

• Disaster detection — onboard wildfire and flood identification for rapid response

• Maritime surveillance — vessel tracking and anomaly detection without the latency associated with ground transmission

• Autonomous operations — enabling spacecraft to make decisions without Earth intervention

ESA is partnering with European start-ups and firms to accelerate on-orbit data processing and analysis. ReOrbit is working with ESA’s InCubed (Investing in Industrial Innovation) program to launch a satellite (in 2025) to demonstrate secure space-to-space and space-to-ground data transfer. It will employ a software-first architecture to enable in-orbit autonomy, interconnectivity, and reconfigurability. It will also host a commercial Earth observation payload and test optical terminals for high-speed inter-satellite communication — capabilities critical for deep space missions, lunar infrastructure, and European digital sovereignty in space.

China

China is undertaking one of the most ambitious and technically fascinating space computing initiatives. On 14 May 2025, it launched the first cluster of 12 satellites — the initial deployment for a massive orbital computing network called the “Three-Body Computing Constellation,” which will eventually form a constellation of 2,800 satellites, each functioning as a node in a distributed AI supercomputer. When fully operational, this orbital system will deliver a combined computing power of 1,000 peta operations per second — that’s 1 quintillion operations per second, according to the Chinese government.

Key capabilities include:

• Real-time in-orbit data processing

• Climate modeling and global weather prediction

• 3D urban modeling and satellite imagery analysis

• Cosmic event detection (some satellites will carry X-ray telescopes for gamma-ray bursts and solar flares)

ADA Space is the lead commercial developer of satellite hardware for this multi-institutional project, and Zhejiang Lab is the project coordinator and AI systems architect. China National Space Administration (CNSA) provides strategic oversight and launch logistics, with Tsinghua University and the Beijing Institute of Technology responsible for research and design of AI frameworks and satellite systems. Baidu Cloud AI Lab is developing AI algorithms.

China’s Three-Body Computing Constellation represents a strategic undertaking intended to position the country as a leader in global Earth observation, autonomous space-based AI, next-gen climate and disaster forecasting, and space-based surveillance for government, military, and national security.

Japan

Japan’s key orbital data infrastructure initiative is Space Compass, a joint venture (founded in 2022) between NTT and satellite operator SKY Perfect JSAT. It seeks to build a space-integrated computing network that includes on-orbit data centers and laser-based optical relays to support Earth observation, climate research, and defense.

 Key features include:

• Satellite constellation with onboard compute and storage — forming a virtual ODC

• Space radio access network — stratospheric and orbital communications layer designed to support beyond-5G/6G connectivity

• Laser-based inter-satellite links and downlinks to Earth — for high-speed, low-latency data transfer

A key component is the GEO optical relay satellite. Part of a Japanese Ministry of Defense project, this satellite will use laser-based optical communication to relay data between satellites and ground stations. It will also support space-domain awareness by transmitting large volumes of surveillance and environmental data and facilitate real-time analytics for defense, disaster response, and climate monitoring.

Space Compass aims to build a sovereign, sustainable infrastructure for Japan and its partners by offloading compute from terrestrial data centers to solar-powered satellites — cutting emissions, strengthening national security through independent space-based data processing, and enabling future space missions. This initiative aligns with Japan’s broader space ambitions, including its role in NASA’s Artemis program and the development of lunar communications and navigation systems.

Conclusion

ODCs are starting to evolve from concept to deployment. We are now witnessing the emergence of a new computational frontier — one where satellites no longer just observe but also interpret, decide, and act in real time. From Axiom’s AxDCU-1 and ESA’s Φsat-2 to Japan’s Space Compass and China’s planetary-scale AI constellations, the race to build sovereign, space-based data infrastructure promises to reshape how we understand and respond to Earth’s dynamic systems. These platforms represent a shift in governance, autonomy, and resilience. By processing data in orbit, they reduce latency, enhance security, and unlock new capabilities — from disaster detection and climate modeling to autonomous spacecraft operations and deep space mission support. In short, the convergence of AI, edge computing, and satellite architecture not only accelerates Earth observation but also redefines it. As these orbital systems become more modular, intelligent, and interconnected, they will form the backbone of a planetary information ecosystem.