AstroForge plans a spacecraft that can fly itself using onboard AI

In short: AstroForge says its next spacecraft, planned for 2027, will use onboard AI to handle many decisions that mission teams usually make from Earth.

What happened

AstroForge, a startup working on asteroid mining, says it has built an autonomous control system for spacecraft called Solo. The company says Solo uses a small “transformer” AI model (the same general family of AI used in many chatbots) along with standard flight control software.

Space missions are often run by large teams on the ground who send commands to the spacecraft. TechCrunch notes that NASA’s OSIRIS-REx asteroid mission had about 100 operators per eight-hour shift. AstroForge does not have resources like that, and it also faced problems communicating with its Odin spacecraft in 2025 because deep-space communication depends on a limited number of very large antennas and short contact windows.

AstroForge’s plan is to put more decision-making onboard, so the spacecraft can notice problems and try fixes on its own. For example, the company describes a case where the AI could connect a power problem to a faulty “star tracker” (a camera-like sensor that helps a spacecraft figure out where it is) and then try a simple reset (like turning a device off and back on).

Before it hands over control, AstroForge plans to fly Solo in “shadow mode” on its DeepSpace-2 vehicle, expected to launch by the end of 2026. Shadow mode means the AI watches and suggests actions, but humans still make the final call.

Why it matters

If this works, it could make deep-space missions cheaper to operate, because the spacecraft would rely less on constant help from Earth. It also raises a key question: how much control should we give software when there is no easy way to step in if something goes wrong.

Source: TechCrunch AI