Firefly Aerospace will run NVIDIA Jetson AI in lunar orbit for the first time on Blue Ghost Mission 2

NVIDIA and Firefly Aerospace announced on June 29 that Blue Ghost Mission 2, targeted for late 2026, will carry the first operational NVIDIA Jetson edge AI deployment in lunar orbit. The mission will run AI inference directly on NVIDIA hardware in space, processing imagery on-orbit rather than transmitting raw data back to Earth for analysis.

What's new

Firefly's SciTec subsidiary is developing the Ocula moon imaging service, which runs AI algorithms on NVIDIA Jetson modules aboard the Blue Ghost Mission 2 spacecraft. The system processes ultraviolet and visible spectrum imagery onboard, identifying and transmitting only relevant data rather than the full raw feed.

The contrast with Mission 1 is stark. Blue Ghost Mission 1, which landed on the Moon in March 2025, downlinked approximately 120 gigabytes of raw imaging data that required weeks of ground-based processing. Mission 2 eliminates that bottleneck by moving the compute to the spacecraft.

Primary applications for the on-orbit AI system include:

• High-resolution landing site mapping for future missions
• Mineral detection, specifically ilmenite extraction candidates
• Infrastructure monitoring
• Object tracking throughout cislunar space

The Blue Ghost Mission 2 architecture splits into two spacecraft: a lander that will descend to the Moon's far side, and the Elytra orbital platform that will continue a five-year mission in lunar orbit. The NVIDIA Jetson systems will operate aboard Elytra, combining solar power with AI processing to enable continuous autonomous analysis.

Context

NVIDIA's Jetson platform is designed for edge AI inference — running trained models in resource-constrained, power-sensitive environments. It has found broad adoption in terrestrial robotics, autonomous vehicles, and embedded systems. Deploying Jetson in lunar orbit represents a significant extension of edge AI into space infrastructure.

Firefly Aerospace has established itself as a serious commercial lunar player through its NASA Commercial Lunar Payload Services (CLPS) contracts. Blue Ghost Mission 1 was a technical success, delivering NASA payloads and returning lunar surface imagery. Mission 2 continues that work while pushing further into AI-enabled autonomous operation.

The broader industry context is a growing push for on-orbit compute. As commercial lunar activity increases — from NASA's Artemis program to private landers from multiple companies — the ability to process and act on data locally, without the 1.3-second round-trip light delay to Earth, becomes increasingly valuable for time-sensitive tasks like hazard avoidance and resource identification.

Why it matters

This is the first planned operational deployment of commercial edge AI inference hardware in lunar orbit. If Blue Ghost Mission 2 launches and operates as planned, NVIDIA Jetson becomes the first commercially available AI accelerator to run sustained workloads beyond Earth orbit.

For the AI hardware space, it validates that modern edge inference silicon can operate reliably in deep-space radiation environments at lunar distances — a meaningful certification for future space infrastructure bids. For the commercial space sector, it demonstrates that AI processing is mature enough to replace ground-based analysis pipelines that add days or weeks of latency to mission data.

The mission also serves as a template: if Ocula's on-orbit inference proves reliable over Elytra's five-year orbital life, similar architectures could extend to asteroid survey, Mars orbit, and beyond.