AstroForge Deploys AI to Control Next Spacecraft

Reaching an asteroid, landing safely, and executing flawless maneuvers sounds simple—but it’s far from easy.

NASA spacecraft exploring the solar system rely on redundancy and automation, yet flight controllers on Earth still manage most operations. The 2018 Osiris-Rex asteroid mission, for instance, required 100 operators per eight-hour shift.

Startups lack such resources, so AstroForge, an asteroid mining tech developer, is turning to AI. The company has built “Solo,” an in-house transformer-based autonomous control stack for its spacecraft.

AstroForge plans to launch its first autonomous spacecraft in 2027 via Stock Space’s inaugural rocket. Backed by NASA, this mission will collect scientific data about the Sun.

This would be a significant milestone: Most spacecraft rely on traditional control algorithms due to neural network reliability concerns. The first use of a neural network to control satellite orbital positioning occurred only last year.

Founded in 2022 with $56 million in venture funding, AstroForge has launched two prototype spacecraft, both experiencing anomalies that prevented them from meeting most mission objectives. In 2025, its Odin spacecraft entered deep space, but communication proved difficult. Earth has limited large antennas capable of transmitting to spacecraft hundreds of thousands of miles away, and communication windows are brief.

Although AstroForge ultimately lost control of Odin, the experience prompted a key question: Could sufficient intelligence be embedded onboard to allow the spacecraft to resolve its own issues?

“Would that have been recoverable with all the data on the spacecraft? I don’t know, but I can tell you nothing onboard tried it, and I would love something onboard to try if the spacecraft is unrecoverable at launch,” said Matthew Gialich, AstroForge’s co-founder and CEO.

“The trade for me is: Do I go build my own ground network, which is going to cost [around] $200 million to put up five dishes around the world and then do operations on it, or do I try to remove it with a model?”

Armand Awad, AstroForge’s head of flight software, explained that the company leveraged advances in transformer models from frontier labs. They created a stack combining traditional control algorithms, models trained on test data for specific subsystems like power generation or navigation, and an overall intelligence layer trained on approximately 2,500 spacecraft sensors.

“I’m not saying I’m going to make general spacecraft autonomy or general autonomy for the world,” Gialich stated. “I’m making a constrained autonomy at a very low sensor input, following the basic training of a transformer model.”

Theoretically, the controlling agent will handle tasks like anomaly resolution. Awad envisions it recognizing a loss of positional tracking, correlating a power anomaly with star tracker issues, and resolving the problem—“probably by turning it off and on in this case.”

AstroForge’s third vehicle, DeepSpace-2, is scheduled to launch alongside Intuitive Machines’ third lunar mission, expected to depart by late 2026. Solo will operate in “shadow mode” on that vehicle, allowing AstroForge’s engineers to test it before the Autonomy-1 mission.

Will this truly be an AI agent operating an independent spacecraft?

“I don’t plan on flying radios that can receive from Earth on Autonomy-1,” Gialich said. “We have to go all in right now. The team’s probably going to talk me into it by the time we fly it. But right now, I’m telling them no radios.”