Firefly Aerospace (FLY) Brings AI Lunar Vision Navigation In-House

Firefly Aerospace Inc. (FLY) said on its August 11, 2026 earnings call that an AI vision navigation system handled hazard avoidance and navigation during the last hour before Blue Ghost Mission 1 landed on the moon, and that the company has now bought that software and the team behind it in-house. Management disclosed no revenue or cost figures attached to it [1].

What Firefly sells, and what the software actually does

Firefly is in the business of getting things off the ground: its own Alpha rocket puts satellites into orbit, its Blue Ghost lander carries customer payloads to the lunar surface, and its Electra spacecraft flies missions in lunar orbit. Revenue comes mainly from launch and mission contracts with NASA and other government agencies plus commercial customers, and it is recognized in stages as mission milestones are met.

The AI application described on the call is vision navigation software that rides on Firefly's own vehicles. It reads the lunar surface imagery captured by the spacecraft's cameras, identifies terrain hazards such as boulders and craters, computes the spacecraft's position relative to the surface, and then steers the vehicle around the dangerous spots and through descent and touchdown. It also supports sustained operation in very low lunar orbit. The users are the flight control systems and mission teams of the Blue Ghost lander and the Electra spacecraft, which puts the software inside Firefly's core operations.

The software was already flying before Firefly owned it

The system came from SpaceNG, a company that builds AI vision navigation and autonomous guidance systems. Its software performed the autonomous hazard avoidance and navigation in the final hour of Blue Ghost Mission 1's descent. On the August 11, 2026 call, management presented that capability as Firefly's own: the SpaceNG team, together with flight-proven spacecraft software and camera hardware, was folded into Firefly and assigned to keep supporting the Blue Ghost lander and Electra, including keeping vehicles operating safely and persistently in very low lunar orbit [1].

Earlier calls in the period under review did not describe this software, and this is the only occasion so far on which it has been laid out in full. That first appearance marks when the capability was discussed, not when it was adopted; the software was already in service by the time it was first observed here.

What "the last hour" measures, and what it does not

The last hour refers to how long the software had autonomous control of the spacecraft, and that window happens to cover the riskiest stretch of a lunar landing. By then the vehicle is low and descending fast, ground controllers cannot intervene in time, and whether the ground below is flat enough to land on has to be judged by the spacecraft itself, which then has to change its own path. The result management reported is that Blue Ghost Mission 1 landed safely and softly [1].

That figure measures the span over which the software operated on its own. It says nothing about navigation accuracy, landing dispersion or landing success rate, and it does not separate the software's role from the rest of the landing stack, including guidance and propulsion. It is therefore not a measure of what the software contributed to this soft landing on its own.

Where the capability could reach the financial statements

The lines most likely to be affected are milestone revenue recognition on lunar delivery contracts, and the cost of a single mission. Firefly recognizes lunar delivery revenue in stages as mission milestones are reached, and landing is both the last and the hardest of those steps, so completing it decides directly whether that slice of the contract value is earned. A failed landing leaves the money already spent on the vehicle and the launch with no revenue behind it. The software acts on the final phase of descent, so as long as it stays reliable, each mission has a better chance of reaching the revenue recognition milestone. Taking the capability and the team inside the company also converts what had been an outside purchase into an internal cost, which over time could change the cost structure of a single mission.

So far, however, the company has not put any revenue, cost or mission expense figure against this software, and it has not tied this landing result to a specific contract amount. Management described the acquisition only as cost neutral. The path is directionally reasonable, but it remains unquantified and unattributed by the company.

What is confirmed today, and what would settle it

The operating change that can be confirmed is that control of the most critical step of a lunar mission has moved inside the company: the hazard avoidance and navigation software, the camera hardware and the development team all now sit within Firefly's own organization, and one real lunar mission showed the system can take a vehicle down autonomously through the most dangerous stretch. How much it has reduced landing risk, and whether that result repeats on later missions, has no comparable reading yet. Pricing this capability will take either a comparable landing or in-orbit metric from the next lunar mission, including Electra's very low lunar orbit work, or a company disclosure that maps it to the revenue and mission cost of specific contract milestones.

Application assessment

• Autonomous Lunar Vision Navigation | Business position: core operations | Deployment stage: limited production | Coverage: multiple business lines or regions | Value type: risk reduction

Source

[1] Drillr - Firefly Aerospace Inc. (FLY) - 2026-08-11 - Earnings call

"The software from our new acquisition was proven on Blue Ghost Mission 1 when it performed autonomous hazard avoidance and navigation in the last hour to ensure we safely and softly landed on the moon's surface." - Jason Kim, prepared remarks