Loft Orbital and UAE-based Marlan Space are committing $1 billion to build a 50-satellite artificial intelligence infrastructure platform designed to analyze Earth-observation data directly in orbit. The Altair-Next Gen program will combine radar and optical sensors with onboard computing and AI models from Mistral, allowing satellites to identify events and send alerts within seconds rather than waiting for imagery to reach ground-based processing systems.
The next stage of AI infrastructure may not be confined to data centers.
Loft Orbital and Marlan Space are building an AI platform in orbit, announcing a $1 billion program to expand their Altair satellite initiative from a 10-satellite demonstration into a 50-satellite constellation.
Announced at the International Space Summit in Paris in the presence of French President Emmanuel Macron, the Altair-Next Gen program is intended to turn satellites from passive data collectors into distributed computing systems capable of interpreting sensor data before it reaches Earth.
The initial fleet will carry optical, radar and other sensors. Instead of transmitting large volumes of raw imagery to ground stations for subsequent analysis, the satellites will process observations onboard and transmit detected events or alerts.
The companies say this could reduce the time between observation and actionable information from hours to seconds.
That shift is significant for applications where the value of satellite intelligence declines rapidly with time, including wildfire detection, maritime monitoring, disaster response and protection of ports and critical infrastructure.
Satellites Become AI Edge Devices
The underlying technology is part of a broader move toward edge AI in space.
Traditional Earth-observation architectures collect imagery and send it back to Earth, where powerful ground-based systems perform the computationally intensive analysis. That model remains effective for many applications, but it creates a bottleneck when enormous amounts of sensor data must be transmitted before an algorithm can determine what is actually important.
Onboard AI reverses that sequence.
A satellite can filter, classify and interpret information in orbit, sending only relevant results to operators on the ground.
NASA’s Jet Propulsion Laboratory has been testing a similar concept. Its Dynamic Targeting technology used onboard AI to analyze imagery and help an Earth-observing spacecraft determine where to point its instruments, with the demonstration completing its processing cycle in less than 90 seconds. NASA has also been exploring ways for multiple spacecraft to share AI-derived observations and coordinate measurements.
ESA has been pursuing the same architectural shift. Its Φsat-2 mission processes Earth-observation imagery onboard and runs applications for tasks including wildfire detection, vessel identification and disaster monitoring.
Altair-Next Gen is attempting to take that concept from individual technology demonstrations to a commercial, multi-satellite infrastructure.
Mistral Models Move AI Into Orbit
The program’s software layer is another important component.
French AI company Mistral is joining the consortium to provide models running directly onboard the satellites. According to the companies, those models will provide reasoning and natural-language tasking capabilities and will also be available through an AI application store.
The proposed application-store model is particularly interesting.
Rather than creating a dedicated satellite constellation for every individual AI application, the infrastructure is designed to let customers select models and deploy them across shared satellite resources.
In effect, the companies are trying to create a compute-and-sensing platform in orbit.
That resembles the evolution of terrestrial cloud infrastructure, where customers increasingly consume computing, storage and specialized AI capabilities without owning the underlying hardware.
The difference is that space-based computing operates under much harsher constraints. Satellites face limits involving power, thermal management, communications bandwidth, radiation exposure and hardware availability. AI models therefore need to be optimized for the environment rather than simply transferred from a terrestrial data center.
NASA has noted that active spacecraft typically cannot accommodate large software updates because of bandwidth constraints, making lightweight and specialized onboard AI models particularly important.
BlackSky Adds High-Resolution Imaging
The constellation will also incorporate high-resolution optical satellites from BlackSky.
BlackSky has been selected as the exclusive provider of very-high-resolution electro-optical satellites for the program, with its Gen-3 technology forming part of the 50-satellite architecture.
That creates a combination of sensing technologies rather than a uniform fleet.
Radar can provide observations in conditions where optical imaging is limited, while high-resolution electro-optical sensors can provide detailed visual information. Combining different sensors with onboard AI creates the possibility of using one data source to trigger another or to prioritize observations.
This is where the constellation could become more than a faster Earth-imaging service.
The long-term objective is closer to a distributed autonomous sensing system in which satellites detect events, interpret them and potentially coordinate subsequent observations.
Abu Dhabi Becomes the Manufacturing Hub
The physical infrastructure will be produced by Orbitworks, the joint venture established by Marlan Space and Loft Orbital.
The satellites are being manufactured at Orbitworks’ facility in Abu Dhabi, with the first 10 already in production. The first launch is planned for October 2026, according to the companies.
The manufacturing strategy is strategically important for the UAE.
Instead of simply purchasing satellite capacity from international providers, the program is intended to build domestic engineering expertise, manufacturing capabilities and intellectual property around commercial space infrastructure.
Orbitworks was formed in 2024 by Marlan Space and Loft Orbital, and the facility is designed to support higher-volume satellite production.
The resulting ecosystem connects UAE capital and manufacturing with French and European space capabilities, Mistral’s AI technology and BlackSky’s Earth-observation systems.
Sovereign AI Meets Commercial Space
The program also reflects a growing convergence between AI infrastructure and national security.
Altair-Next Gen is intended to serve government users in France, Europe and the UAE as well as commercial and sovereign customers globally. Initial applications include maritime domain awareness, disaster response, wildfire detection and monitoring of ports and critical infrastructure.
That makes sovereignty a significant part of the business proposition.
Countries increasingly want control over critical data and AI capabilities, particularly when satellite imagery can reveal information about military assets, infrastructure, shipping and environmental events.
A constellation capable of processing sensitive observations before transmitting them could reduce dependence on centralized ground-processing infrastructure and potentially improve response times.
At the same time, the architecture introduces new governance and cybersecurity requirements. AI models running in orbit become part of the operational satellite system, meaning model integrity, software updates, authentication and resilience have to be treated as space-infrastructure concerns.
From Satellite Imagery to Space-Based Intelligence
The larger technology trend is clear: AI is moving the point of computation closer to where data is generated.
That transition is already underway in terrestrial industries, from manufacturing and autonomous vehicles to telecommunications and defense. Space presents a particularly compelling case because bandwidth between satellites and Earth is constrained and latency can matter.
ESA has described onboard AI as a way to make satellites more reactive, agile and autonomous, while its Earth-observation programs have demonstrated that processing information in orbit can reduce the amount of data that needs to be transmitted.
The challenge for Loft Orbital, Marlan Space and their partners will be scaling that concept reliably across 50 satellites.
The program is therefore not simply another Earth-observation constellation. Its bigger ambition is to create a reusable AI infrastructure layer in orbit, where sensors, compute, connectivity and software applications operate as a single system.
If the model works at commercial scale, satellite customers may increasingly buy access not to imagery itself, but to real-time intelligence generated from imagery.
That distinction could reshape the economics of Earth observation—and push AI infrastructure into a new frontier: space.
Market Landscape
The Altair-Next Gen announcement comes as governments and commercial space companies move AI processing closer to the satellite sensor.
- Onboard AI is already operational: ESA’s Φsat-2 processes imagery in orbit and supports applications including wildfire, vessel and disaster detection.
- NASA is pursuing autonomous observation: JPL’s Dynamic Targeting demonstrates how onboard AI can analyze observations and influence what a spacecraft observes next.
- Space is becoming an edge-computing environment: ESA has been funding research into making satellites more autonomous, reactive and capable of processing data onboard.
- AI applications are becoming modular: Altair-Next Gen’s proposed application-store approach could allow different customers and models to share satellite infrastructure rather than building dedicated constellations.
- Sovereignty is increasingly important: France, Europe and the UAE are positioning the infrastructure for government use alongside commercial services, connecting AI sovereignty with satellite capabilities.
Top Insights
- Loft Orbital and Marlan Space are expanding Altair from 10 satellites to 50 under a $1 billion AI-in-space infrastructure program.
- Onboard AI could turn satellites from imagery collectors into real-time intelligence platforms capable of identifying events before data reaches Earth.
- Mistral will provide AI models designed to operate directly alongside satellite sensors and support natural-language tasking.
- BlackSky will provide high-resolution optical satellites, complementing radar and other sensing technologies across the constellation.
- Manufacturing in Abu Dhabi gives the UAE a larger role in satellite production, engineering, intellectual property and space services.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
