# AstroForge puts AI in command of spacecraft

> A startup is replacing ground control teams with an onboard AI system to autonomously manage deep space missions.
- Title: AstroForge puts AI in command of spacecraft · TechCrunch
- Summary: A startup is replacing ground control teams with an onboard AI system to autonomously manage deep space missions. AstroForge is building an AI control system…
- Keywords: astroforge, spacecraft autonomy, deep space, transformer models, space startups, nasa, stoke space, intuitive machines, autonomous control, technology, startups, puts
- Source: TechCrunch — https://techcrunch.com/2026/09/22/astroforge-is-putting-ai-in-command-of-its-next-spacecraft
- Author: Tim Fernholz
- Published: 2026-09-22T15:00:00+00:00
- Read time: 1 min
- Topics: astroforge, ai, spacecraft autonomy, deep space, transformer models, space startups, nasa, stoke space, intuitive machines, autonomous control, technology, startups
## NASA's ground control approach
NASA's OSIRIS-REx mission to an asteroid in 2018 required 100 operators per eight-hour shift to control the spacecraft from Earth. This level of staffing and infrastructure isn't available to startups.
## AstroForge's autonomy solution
The startup developed Solo, a transformer-based AI control system built in-house to handle spacecraft operations without constant Earth communication. This lets smaller companies operate deep space missions independently.
## First mission planned for 2027
AstroForge plans to launch its first autonomous spacecraft in 2027 on a Stoke Space rocket, backed by NASA, to gather scientific data about the sun. This would be a first for neural network spacecraft control.
## Learning from past failures
AstroForge's 2025 Odin spacecraft lost communication in deep space because Earth has limited antennas for distant spacecraft. The company couldn't regain control, prompting the shift to onboard AI autonomy.

> 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.
## The cost trade-off
Building a global ground network with five dishes would cost around two hundred million dollars. Using onboard AI removes that infrastructure requirement entirely.

> Do I go build my own ground network, which is going to cost [around] $200 million to put up five dishes around the world?
## How Solo actually works
Solo combines traditional control algorithms with transformer models trained on about 2,500 spacecraft sensors. It's designed to detect and fix specific problems like power anomalies or navigation errors.
## Testing before full autonomy
AstroForge's third spacecraft, DeepSpace-2, will launch with Solo in shadow mode alongside Intuitive Machines' moon mission by end of 2026. Engineers will monitor the AI before the fully autonomous Autonomy-1 mission.
## Going all-in on autonomy
For the Autonomy-1 mission, AstroForge plans to remove Earth communication radios entirely, forcing the spacecraft to operate completely independently. This is a bold commitment to onboard AI control.

> I don't plan on flying radios that can receive from Earth on Autonomy-1. We have to go all in right now.
## Key takeaway

AstroForge is building an AI control system for spacecraft because startups can't afford the massive ground infrastructure that NASA uses, and the approach could solve communication problems that disabled their previous missions.