Welcome to Issue 2 - Autonomy in Orbit, Ahead of the Data Centers
The orbital data center thesis is being stress-tested in public while the AI workloads it would eventually host are already moving into orbit on their own. Last week’s activities spanned various aspects of defence and security elements.
On the Space4AI side, a Breaking Defence analysis declaring the sector at "peak hype" but not yet ready for national-security workloads sets the frame, while a panel at SmallSat Europe names the specific obstacles — chip lifespan, launch, and cybersecurity. Against that backdrop, the incremental progress shows up as a hardware refresh on an existing in-orbit AI supercomputer as well as a UK–US partnership to host sovereign storage on orbital power platforms.
On the AI4Space side, the arc runs from research access to operations to automation: ESA's Φ-lab opens commercial thermal data to researchers, an Israeli satellite is already tracking moving targets day and night with onboard AI, and Chinese researchers unveil an LLM-agent system to automate satellite targeting and surveillance workflows end-to-end.
The compute is not yet in orbit at scale, but the autonomy that will run on it is shipping ahead of the racks.
Specifics below.
Space4AI
Despite 'peak hype,' orbital data centers for AI not yet ready for National Security
On May 26, Breaking Defense published an analysis concluding that orbital data centers for AI have hit "peak hype" but remain unready for national security workloads. The piece draws on Quilty Analytics and roughly a dozen officials, and notes that the sector pulled in $541M in venture funding in Q1 2026 alone. Quilty adds that the current excitement around the topic will further increase investment in ODCs.
However, the gating constraints to deploy ODCs are physical rather than algorithmic: launch cost per kilogram, the scale of solar arrays required to power dense compute, and the performance ceiling of radiation-hardened chips. Those (current) limits keep on-orbit compute below the thresholds that national security customers would need to treat it as a primary tier. In practical terms, current ODC designs, which are planned to serve commercial markets, go by orders of magnitude beyond the scope and needs of national security users.
SmallSat Europe: orbital data centers face chip lifespan, launch, and cybersecurity hurdles
The scepticism about orbital data centers continues at SmallSat Europe. A Via Satellite dispatch from the event in Amsterdam reported on an industry panel that framed the near-term obstacles to orbital data centres, with chip lifespan, launch availability, and cybersecurity singled out as the core challenges. The discussion was situated against the backdrop of SpaceX's initial IPO documents, which have pushed the question of moving compute off-planet back into trade-press conversation.
The framing places three constraints alongside one another rather than treating in-orbit compute as a single technical problem. Chip lifespan addresses the radiation and thermal environment that current commercial silicon was not designed for; launch availability speaks to cadence and price for deploying and refreshing hardware; cybersecurity covers the attack surface of compute assets that cannot be physically reached for remediation.
Aitech upgrades S-A2300 space AI supercomputer with NVIDIA IGX Thor
On May 27, Aitech announced an upgrade to its S-A2300 COTS AI Supercomputer, integrating NVIDIA's IGX Thor platform to expand in-orbit data processing capacity for customers. The S-A2300 is positioned as a commercial-off-the-shelf AI compute platform for space deployment, and company officials frame the IGX Thor integration as a step-change in how much customer data can be processed on-orbit rather than downlinked.
This story is a compelling illustration of just how quickly the ODC field is advancing at the technological frontier. That momentum is echoed in the remarks of Aitech's General Manager, Pratish Shah, in an interview with Payload, who notes that the challenge now lies in "the speed at which you can produce."
Space Solar and Lonestar sign LOI to host sovereign data storage on orbital power platforms
Space Solar, a UK developer of space-based solar power and large in-space assembled power structures, and Lonestar, a US company that has previously stored data on the Moon, signed a letter of intent to cooperate on hosting Lonestar's sovereign data storage payloads on Space Solar's orbital power platforms.
Space Solar is building in-space assembled structures intended to supply power to other orbital infrastructure, and Lonestar has positioned its storage business around sovereign data preservation off Earth. The LOI pairs Lonestar's data payloads with Space Solar's platforms as the host power source.
AI4Space
ESA Φ-lab opens announcement of opportunity for constellr SkyBee high-resolution thermal data
On May 29, ESA's Φ-lab opened an announcement of opportunity to disseminate constellr's SkyBee high-resolution thermal data products to researchers for scientific and pre-operational development studies. The AO is related to the backing of the company through ESA's InCubed programme and routes SkyBee thermal imagery into the Φ-lab research pipeline, making the data available for use in scientific and pre-operational work.
Israeli satellite now tracks moving targets day and night
ImageSat International disclosed that its RUNNER satellite is now tracking moving targets day and night using onboard AI to detect, classify, and characterize them directly in orbit. The system processes imagery on the spacecraft, and downlinks finished intelligence rather than raw pixels, and it operates at night as well as during the day.
According to the company, the onboard AI is capable of distinguishing between military-relevant targets and background civilian movements at city scale. The implications reach well beyond military users. Border security, disaster response, maritime domain, and infrastructure protection all have genuine needs for round-the-clock overhead motion detection that functions regardless of available light.
China unveils AI system to automate satellite targeting and surveillance
The final story of today is about Chinese researchers unveiling the "Air Target Agent System," an LLM-agent collaboration tool designed to automate satellite targeting and surveillance workflows, as reported by the South China Morning Post.
The system extends satellite surveillance beyond image recognition: agents autonomously analyze imagery, draw conclusions, and issue tasking instructions. The architecture pairs large language models with specialized agents to handle the analysis-to-action loop that has traditionally required human operators in the loop. The system dramatically widens the scope of military targeting, a domain that has only recently come to see space as an ideal means to an end.
Till next time,
Meta-beat Column of this week
Read also about the AI Pipeline that sits at the core, producing this Newsletter, including its ups and downs of this week: