AstroForge is putting AI in command of its next spacecraft

AstroForge is developing an onboard autonomous control stack called Solo — a transformer-based AI that will command upcoming spacecraft. The company plans to fly its first fully autonomous probe, Autonomy-1, in 2027 aboard Stoke Space's inaugural rocket, after testing Solo in shadow mode on an earlier mission slated for late 2026.

By AI Newsroom· Reviewed by Pranav, Founder & Editor-in-ChiefPublished about 1 hour agoUpdated about 1 hour ago0 views

Why It Matters

If successful, AstroForge’s approach would reduce dependence on large, expensive ground networks and human flight-controller teams by placing more decision-making on the spacecraft itself, a significant shift from traditional, algorithmic spacecraft control. The effort follows earlier communications problems the company experienced with prototype vehicles, which helped motivate the move toward onboard intelligence.

Key Facts

Company: AstroForge
AI stack name: Solo
AI architecture: Transformer-based model, developed in-house
Planned fully autonomous flight: Autonomy-1 in 2027 on Stoke Space's first rocket
Shadow-mode test: Solo to fly in shadow mode on DeepSpace-2 alongside Intuitive Machines' third moon mission expected by end of 2026 (launch target)

AstroForge, a startup focused on asteroid mining, has built an onboard autonomy system called Solo that uses a transformer-based AI to control spacecraft operations. The stack pairs conventional control algorithms and subsystem-specific models with an overarching intelligence layer trained on readings from about 2,500 onboard sensors. The company says Solo is intended to resolve anomalies and perform tasks that normally require ground intervention. The decision to pursue onboard autonomy followed difficulties communicating with AstroForge’s earlier prototypes. In 2025 the company’s Odin spacecraft reached deep space but the team was unable to regain control, in part due to limited availability of large ground antennas and narrow communication windows. Those constraints prompted AstroForge to consider embedding more problem-solving capability on the vehicle itself rather than building an extensive global ground network. AstroForge plans to test Solo in “shadow mode” on its third vehicle, DeepSpace-2, which is slated to fly with Intuitive Machines’ third lunar mission expected by the end of 2026. The company’s goal is to have Solo in charge of its first intentionally autonomous flight, Autonomy-1, on Stoke Space’s inaugural rocket in 2027; that mission will have NASA support and is expected to collect solar science data. Leadership has discussed minimizing or omitting radios on Autonomy-1 to force onboard independence, though the CEO indicated the team might change that decision before launch. Company executives characterize Solo as a constrained autonomy system rather than a general-purpose AI. Solo’s architecture combines transformer-driven decision-making with trained models for specific subsystems like power and navigation, and conventional control loops. AstroForge frames the approach as an alternative to spending hundreds of millions of dollars to deploy and operate a dedicated global ground antenna network to conduct real-time operations from Earth.