By: Dawn Zoldi
The space economy is a present-day battleground of commerce, competition and conflict. By 2035, analysts project the global space economy will exceed $1.8 trillion. And the scramble to claim that value has already begun. Nations and startups clamor for position in low Earth orbit (LEO), the Moon has attracted its first commercial tenants and U.S. Space Command has been rethinking the most fundamental doctrine of how to fight and survive in a contested domain.
Against that backdrop, Autonomy Global’s Full Crew Episode 79 brought together three space sector authorities to cut through the hype and deliver grounded analysis: Moderator Juan Plaza, Autonomy Global Ambassador for Operations Safety; Adrian Thompson, Chief of AI and Data Science at Slingshot Aerospace; Jen Splaingard, Chief Space and Sensors Officer atTomorrow.io; and Sahil Desai, Vice President of Product atFortastra. Together they painted a portrait of an industry at an inflection point, sovereignty is at stake in LEO and the old doctrine of treating satellites as untouchable fixed assets seems dangerously obsolete.
Edge AI Reaches the Moon: Commercial Processors Fly Deep Space
Thompson opened the episode with the Firefly Aerospace announcement that its Elytra orbital transfer vehicle will embed an NVIDIA Jetson module in lunar orbit to power the Ocula commercial Moon imaging service, billed as the first of its kind. The article discussed, “Firefly Aerospace Enables On-Orbit Processing for Moon Imaging Service with NVIDIA Jetson,” describes how Elytra will process lunar imagery on board using Firefly’s SciTec AI software rather than streaming raw data back to Earth. It will transmit only distilled intelligence to customers during a planned five-year operational lifespan following Blue Ghost Mission 2, targeted for no earlier than late 2026.
For Thompson, the significance of this runs deeper than a hardware upgrade. The Jetson module is a consumer-grade commercial processor operating in one of the harshest radiation environments imaginable. This would have been unthinkable a decade ago, when missions routinely relied on $80,000 Field-Programmable Gate Arrays (FPGAs) to perform a fraction of the same processing.
“We’re simultaneously increasing the capability on the software side and we’re reducing the hardware requirements,” Thompson said. “We’re now seeing commercial grade hardware evolving all the way up to space. It’s opening up a door where the barriers to software and algorithms that we can run in space are rapidly falling.”
The reason edge AI matters has everything to do with the physics of deep-space communication. With the Moon a quarter of a million miles away, available bandwidth cannot keep pace with the volume of sensor data an orbiter collects. An edge processor can discriminate between meaningful data and noise in situ. For example, it can flag a mineral signature or track a surface feature rather than transmitting entire frames of blank lunar terrain.
Splaingard reinforced the point by connecting it to her own work in weather intelligence. With AI on Earth now capable of modeling data far faster than before, the demand for lower-latency observations has risen across every domain. “Now with AI on Earth, we’re able to compute, process and model things so quickly that the consumption rate for data has certainly increased, causing the need for latency to continue to improve,” she observed.
What Thompson finds most compelling about the Firefly story is not the hardware itself but what it portends for the software accountability gap that follows. “We finally have Moore’s Law for space hardware. We can finally keep up, we can finally accelerate. But what can we do now about the quality of the software to keep up?”
LEO Broadband Race: Crowded Orbit, Competition No One Can Ignore
Splaingard brought the discussion down from the Moon and into the increasingly crowded band of low Earth orbit (LEO). She centered her segment on “The Coming Wave of Competition in LEO Constellations” from Via Satellite. The piece argues that 2026 marks the first year LEO broadband has begun to look like a genuine race. Amazon Kuiper is on track to reach roughly 700 satellites by mid-year. A new generation of entrants now also target specialized niches that Starlink’s consumer model does not serve.
As the operator of a LEO weather observation constellation, Splaingard occupies a unique vantage point. She does not compete directly with Starlink-style broadband, but her operations are directly affected by everything that comes with it, including congestion, spectrum pressure and collision risk.
“Starlink to date has about 10,000 satellites in LEO,” she noted, “and their full version of this constellation approaches 550,000 satellites.”
These numbers are staggering even to insiders. The article’s core thesis is that the total addressable market is expanding fast enough to sustain multiple serious players, but only through differentiation rather than volume competition.
That differentiation is increasingly about niche and sovereignty rather than raw coverage. Governments treat LEO access as a strategic imperative, not a commercial commodity. China alone has more than 25,000 satellite authorizations, a trajectory that raises hard questions about congestion, spectrum management and whether de-orbit compliance will be globally enforced.
“There is a call and a demand for space policy that is more global than country-based or even allied-based,” Splaingard said. “We have a long way to go to get there.”
Thompson amplified the risk dimension. He argued that the sheer scale of constellation growth makes automated space traffic control inevitable. “It pretty much already heralds that we need to make a transition to much more automated space traffic control,” he said. “There’s a tremendous amount of space junk out there already. We’re going to generate a lot more.”
The collision avoidance problem is compounded by the fact that many smaller cubesats (6U, 8U, even 12U configurations) simply do not carry propulsion systems at all. This makes them passive debris once their missions end.
Space Command’s Warfighting Reset: Maneuver, Agentic AI and the Cost of Standing Still
Desai anchored the episode with a SpaceNews piece “Space Command Pushes New Warfighting Model Built on Moving Satellites.” The article details how Gen. Stephen Whiting and U.S. Space Command have pushed a doctrinal shift away from large, static, exquisite satellite platforms toward proliferated, maneuverable spacecraft designed to survive in a genuinely contested environment. The assumption that space is a peaceful domain where satellites sit unmolested has been invalidated by China’s and Russia’s growing anti-satellite weapon (ASAT) capabilities and aggressive orbital maneuvering.
Desai, a former intelligence community officer who most recently scaled munitions manufacturing before joining Fortastra, which he describes as the first private security force in space, framed the shift. The Space Force’s transition mirrors what the Air Force has been working through with unmanned combat aircraft. It moved from billion-dollar exquisite platforms, to collaborative combat aircraft (CCAs) in the $40-50 million range toward genuinely low-cost attritable systems that conflict in Ukraine has proven necessary. “We’ve gone from extremely exquisite to this middle tier of capability,” Desai said. “But there’s still this broader question of how do we get extremely low-cost systems into space so that if we were in a state of conflict, we had the ability to respond asymmetrically with some kind of advantage?”
The urgency is reinforced by lessons from terrestrial conflicts that blurred the line between commercial and military targeting. Iran attacked Amazon warehouses. ISIS adapted commercial drones into weapons within months of recognizing the technology. “There’s no questioning that the next conflict between major nations will have a space component to it,” Desai said. “The warfare we anticipated has evolved. There’s a mix of commercial and defense assets that are held at risk in times of conflict.”
Fortastra’s model draws on a simple analogy. “When you go build a bank, the first thing you do is create fences, get cameras, hire a guard force. When we’re looking at hundreds of millions of dollars of space infrastructure, we need a similar mindset,” Desai explained.
Thompson synthesized the hardware and software dimensions across all three articles in his comments. The convergence of cheaper commercial hardware has democratized access to space in ways that create as many risks as opportunities. The critical underexplored layer is mission-critical software. “The hardware is getting there rapidly, but we do need to switch a bit of our focus and start looking at what it takes to put agentic systems into space, where the cost of errors is very high?” he asked.
Desai’s vision of the near-term future is agentic. It includes satellites that can reason and act without constant ground intervention, operating in degraded communications environments the way autonomous systems in Ukraine already must. “These platforms need to think, they need to operate agentically,” he said. “They have to be able to rationalize and reason and act with context on edge, compute and make decisions.” The leap from task-based autonomy to true agentic reasoning in space is not a technology problem. Both speakers agreed that it has become an engineering rigor and verification problem the industry has not yet fully reckoned with.
These three experts made clear, collectively, that space no longer operates by its old rules, but instead it is now a domain that reflects every tension shaping the broader autonomous systems world:
- Commercial democratization versus national security risk
- Hardware advancement versus software accountability and
- Speed of innovation versus the maturity of standard.
The burgeoning space economy will generate enormous value. The question the industry must answer, urgently, is whether its infrastructure, doctrine, and governance can mature fast enough to protect it.Watch the Full Crew Episode 79: Space.