For years, the conversation around AI in space has centered on what’s possible. Concepts and research papers have explored the idea of turning satellites into compute platforms capable of interpreting sensor data directly on orbit. What has been missing is the transition from theoretical architectures to flying the processors that can make this possible.
Launched in November 2025, YAM-9 is one of the first commercial satellites to take that step.
Rather than only modeling high-performance processing architectures, Loft is now flying hardware in space to demonstrate scalable, multi-node edge compute under real orbital constraints. This demonstration will collect the performance data needed to shape the next generation of AI-enabled constellations.
“Loft is turning satellites from data collectors into decision engines with one of the most powerful AI-capable onboard processing architectures ever flown,” said Pieter Van Duijn, CTO at Loft. “Our goal has always been to separate aspiration from implementation. YAM-9 gives us the on-orbit feedback loop we need to design systems that are not only powerful, but usable by real customers.”
From Data Collectors to Decision Engines
YAM-9 represents a fundamental shift toward treating satellites as decision engines, where AI models run directly on orbit to interpret sensor data in real time. This approach reduces bandwidth bottlenecks and shortens the time from “something happened” to “we know what to do about it,” especially for time-critical domains like disaster response, maritime security, and RF signal monitoring.
YAM-9 serves as an on-orbit testbed for the architectures that will underpin Loft’s future multi-sensor, AI-enabled constellations to enable this type of decision-making. Data and lessons from YAM-9’s performance will also support the development of a broader ecosystem of sovereign-ready, AI-driven constellations, where governments, primes, and analytics companies can operate on shared infrastructure while preserving control over their applications and data.
The YAM-9 rideshare program is introducing a new class of on-orbit compute capabilities, representing a significant change from single-box demos to a scalable, mission-aligned compute platform that can perform complex AI applications and data fusion algorithms.
One of the Most Capable, Mission-oriented AI Environments in Orbit
YAM-9 is flying a demonstration four-node processing architecture. Each node brings a different mix of CPUs, NVIDIA GPUs, and AI accelerators spanning both x86 and ARM architectures so that diverse workloads can run on the hardware best suited to each task.
“This unique processing architecture makes YAM-9 one of the first commercial satellites of its kind to fly four networked compute units in a single, heterogeneous environment,” said Van Duijn. “Together, these nodes deliver nearly four times the processing performance of previously flown commercial AI processors, creating one of the most powerful AI-capable onboard environments in orbit today.”
This multi-node configuration, the future evolution of Loft’s Hub Compute, forms a secure, distributed processing platform where multiple AI and data-processing applications can run in parallel on the same spacecraft. Rather than a single “black box” processor, this design is a scalable compute layer that can grow with mission complexity and demand, from single-satellite experiments to full, multi-sensor constellations.
“Loft is taking an edge-first approach, distinct from orbital data centers, that exists to serve mission-specific applications, reduce downlink bottlenecks, and improve responsiveness, as opposed to supporting general-purpose workloads, “ according to Van Duijin. “The YAM-9 demo architecture is embedded directly alongside sensors to enable real-time inference, filtering, and prioritization of data at the point of collection.
Making Space-Based AI Accessible
By demonstrating advanced compute capabilities on operational satellites, Loft is building an infrastructure platform where developers, governments, and enterprises can deploy AI applications in space as easily as they do on the ground. Customers don’t have to wait on high-volume data packets to downlink when they can quickly receive reports that answer their questions. AI workloads can run at the edge, on orbit, on a flexible, secure compute architecture built for rapid iteration, real-world operations, and future constellations at scale.
As AI continues to reshape how decisions are made on Earth, YAM-9 will demonstrate how Loft’s infrastructure is helping reshape how decisions are made from space, turning satellites into a global, AI-enabled infrastructure that delivers answers, not just data and images.
