TL;DR
AstroForge is replacing a $200 million ground-control network with an in-house transformer AI called Solo, betting that onboard autonomy can do what 100-person NASA shift teams cannot afford to do for a startup.
What happened
- AstroForge unveiled Solo, a transformer-based autonomous control stack built in-house to command its spacecraft without ground intervention.
- Solo integrates traditional control algorithms, subsystem-specific models, and a top-level intelligence layer trained on roughly 2,500 onboard sensors.
- The system will fly in shadow mode on DeepSpace-2, launching alongside Intuitive Machines' third moon mission before end of 2026, to validate performance before full deployment.
- The Autonomy-1 mission is slated for 2027 on the first Stoke Space rocket, backed by NASA, with CEO Matthew Gialich stating his intent to fly with no Earth-receiving radios onboard.
- The decision was driven by the 2025 loss of the Odin spacecraft: limited antenna windows on Earth made recovery impossible once communications degraded.
Why it matters
- Building a five-dish global ground network costs around $200 million by Gialich's estimate. Solo is the startup's bet that a model is cheaper and more resilient.
- Most spacecraft autonomy still relies on classical control algorithms because neural networks are considered unreliable in flight. The first neural-network-controlled satellite positioning happened only in 2025, making Solo a significant leap.
- NASA's OSIRIS-REx asteroid mission required 100 operators per eight-hour shift. Solo targets the same class of mission with a fraction of the human infrastructure.
- If Solo works, it sets a template for commercial deep-space operations without legacy ground infrastructure, lowering the capital bar for asteroid mining and beyond.
- AstroForge has only $56 million in total venture funding and two failed prototype missions behind it. Solo is not a research project but a survival strategy.
What to watch next
- Shadow-mode results on DeepSpace-2 (late 2026) will be the first real stress test. Watch for AstroForge disclosures on anomaly detection accuracy versus human operators.
- Whether Gialich follows through on no radios for Autonomy-1 is the defining signal. A last-minute reversal would suggest the team lacks confidence in Solo under mission conditions.
- Competitor and agency responses: if Solo performs, expect NASA and defense primes to accelerate onboard AI programs and potentially acquire or partner with AstroForge before the asteroid mining market matures.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.