NASA and Loft Orbital have started testing AI satellites that can spot fires, pollution, and other fast-moving events on their own, then tip off other spacecraft without waiting for a ground team to sift through the data first. The experiment is part of NASA-funded FAME, a push toward distributed, autonomous Earth observation that could make satellite networks much faster at noticing what actually matters.
The first test is already running on an active Loft Orbital satellite, with software from NASA’s Jet Propulsion Laboratory executing directly on the spacecraft’s onboard computers. More trials are planned for 2027 and 2028 using newer satellites, and the bigger prize is obvious: less raw data dumped to Earth, fewer delays, and less dependence on operators playing traffic cop from the ground.
How FAME changes tip-and-cue
The system is designed to automate tip-and-cue, the old-school workflow where one satellite spots something, sends the image home, humans review it, and only then another spacecraft is told to take a closer look. That chain is fine if the target is a glacier. It is terrible if the target is a wildfire moving fast or a slick spreading across the ocean.
FAME pushes most of that decision-making into orbit. The onboard AI analyses images as soon as they are captured, flags events of interest, and broadcasts the result through inter-satellite links so another sensor can jump in immediately. That architecture is the real shift here: not smarter pictures, but smarter timing.
Why onboard AI is suddenly practical
This kind of autonomy used to be a nice demo idea that fell apart the moment you looked at satellite compute budgets. The hardware was too limited, and the AI models were too heavy. That equation has changed only recently, thanks to smaller multimodal models that can compare different data sources while using far less processing power.
Loft Orbital says that combination of compact models and modern onboard computers makes the project workable now, not five years from now. That also helps explain why satellite operators are suddenly talking about edge AI with straight faces: the industry has moved from ”maybe” to ”finally” because the chips and models caught up at roughly the same time.
Altair and the push for autonomous constellations
Loft Orbital plans to build Altair, a 10-satellite constellation built around multiple sensors, onboard AI, and inter-satellite communications. The company’s pitch is simple: if spacecraft can constantly watch a region, classify what they see, and redirect other satellites on their own, the system becomes much more valuable for governments and commercial customers alike.
- Onboard AI analyses images in orbit instead of sending everything to Earth first.
- Inter-satellite links let one spacecraft alert another for immediate follow-up observations.
- Planned uses include wildfire detection, marine pollution monitoring, environmental events, security, and reconnaissance.
If the tests go well, the winners are anyone who needs faster decisions from space data. The losers are the old workflows built around human review and delayed tasking. The open question is how far this autonomy can go before operators insist on a human in the loop for the riskiest calls.