One single conversation covered a nuclear-powered spacecraft deploying helicopters at Mars, a rotorcraft exploring Saturn’s moon Titan, a Venus probe deciding which discoveries to transmit before its operating window closes and a homeland defense network interpreting vast streams of sensor data fast enough to counter incoming threats. NASA Administrator Jared Isaacman and U.S. Space Force Gen. Mike Guetlein, director of Golden Dome for America commanded the stage as they elaborated on these ambitions during a fireside chat moderated by U.S. Senator John Hoeven of North Dakota at the recent 20th Annual UAS Summit. As they described how power, AI, autonomy and networked infrastructure could reshape both deep-space exploration and national defense from their own perspectives, it became clear that the technological challenges both leaders face increasingly intersect. At the heart of both efforts lies the need to build intelligent, interconnected systems that can operate reliably when distance, speed or scale puts direct human control out of reach. Autonomy provides the glue.
A Lunar Base, Not Another Brief Visit
Isaacman described NASA’s lunar ambitions as an effort to establish humanity’s first outpost on another world and prepare for Mars. He said NASA targets an American lunar landing in 2028, and pitted this aggressive schedule against China’s competing ambitions. “This space race is coming down to months, not years right now,” he explained.
The competition concerns more than arrival dates. Isaacman argued that access to useful terrain near south-polar water ice could determine where nations establish infrastructure and develop the ability to use local resources. Despite the moon’s size, he said, suitable operating locations in the areas of interest remain limited.
The moon also offers a relatively nearby environment for testing the systems that a Mars expedition would need. Isaacman identified habitability, mobility, power, communications and spacesuits as capabilities NASA must mature before undertaking a journey he described as six to nine months.
Building a base also means building its supporting network. Isaacman described communication relay satellites and observation capabilities that would connect surface astronauts and autonomous vehicles with Earth. He even envisioned a website showing the base’s construction in high definition. Those requirements, he argued, overlap with military needs for communications and space domain awareness, creating opportunities to avoid duplicative programs.
Mars Helicopters and Titan’s Flying Laboratory
Isaacman connected to the uncrewed systems audience when he discussed robotic aircraft operating far beyond terrestrial airspace. He described SR1 Freedom, a nuclear-powered interplanetary spacecraft NASA intends to launch in 2028. Under the plan he outlined, it would fly past Mars and release three Ingenuity-class helicopters. The aircraft would explore the surface with ground-penetrating radar to search for subsurface water ice, linking advanced space power with autonomous aerial scouting.
Farther out, Dragonfly would take rotorcraft exploration to Titan. Isaacman described the eight-rotor vehicle as roughly the size of a small car, tasked with investigating the complex chemistry of Saturn’s moon.
Distance makes autonomy fundamental. Long communications delays prevent Earth-based teams from handling every immediate flight decision. The vehicle must operate safely while preserving the scientific value of a multibillion-dollar mission.
Isaacman explicitly connected that challenge to NASA’s terrestrial work. Autonomous capabilities developed for Dragonfly can inform satellites in low Earth orbit and uncrewed aircraft on Earth. Robust communications research likewise has applications across those environments.
The exchange runs both ways. Isaacman said testing and collaboration around terrestrial uncrewed systems can help NASA prepare machines to operate on other worlds. He also described a space-to-aircraft application at home: using early wildfire observations from space-based assets to support the dispatch of uncrewed aircraft before fires become larger catastrophes.
Why Deep Space Needs More Power
More power could expand what scientific spacecraft can accomplish far from Earth. Isaacman pointed to Europa and Enceladus, discussing their oceans and the scientific promise of investigating environments that might help answer whether life exists elsewhere.
For its part, Dragonfly uses a plutonium-238 radioisotope nuclear power system, not a fission reactor. Isaacman described approximately two kilowatts of thermal output converting into roughly 100 watts of electricity. He contrasted that with a prospective 250-kilowatt fission reactor operating at approximately 40% conversion efficiency, illustrating the much greater electrical capacity NASA wants to pursue.
Isaacman argued that science missions can also mature fission technologies before NASA relies on them for human exploration. Higher-temperature materials, improved power conversion and autonomous management could benefit both spacecraft and terrestrial energy systems. “We’re not going to co-locate a control room with the spacecraft that we sent to Saturn and Jupiter’s moons,” he quipped.
For Isaacman, that is precisely the kind of difficult, initially noncommercial work NASA should undertake. The government should pursue capabilities without an obvious present-day business case while purchasing mature commercial services where industry already delivers effectively.
AI as Scientist and Onboard Decision-Maker
Isaacman identified two distinct AI opportunities: finding discoveries within accumulated data and enabling spacecraft to make better decisions during missions.
NASA has amassed approximately 150 petabytes of data, he said. That archive raises a consequential question: what have researchers missed? He recounted a Texas high school student’s use of an off-the-shelf AI application to identify a previously overlooked galaxy to illustrate how new analytical tools can extract discoveries from existing observations.
The Nancy Grace Roman Space Telescope would magnify that challenge. Isaacman described its field of view as approximately 100 times Hubble’s and emphasized its ability to survey the universe at extraordinary speed. He presented Roman’s investigations of dark matter, dark energy and distant worlds as a data-analysis task that researchers alone could not fully manage.
AI could also influence what information spacecraft collect and return. Isaacman cited DAVINCI, the planned Venus mission, describing advanced intelligence as part of its design and mission model. In a high-pressure environment with a limited operating lifespan, he said, the spacecraft could prioritize the information most valuable to scientists. “What information it chooses that it thinks is going to be most important to send back in its dying breath” becomes part of the mission challenge, Isaacman said.
That moves AI beyond post-mission analysis and into the collection process itself, particularly where distance limits traditional mission control.
Golden Dome: Integration Before Replication
Guetlein faces a different constraint: the speed and scale of threats approaching the homeland. Hoeven characterized Golden Dome as a system of systems, asking how Guetlein could integrate its components, finance development and keep pace with adversaries. Guetlein rejected simply replicating expensive military platforms designed to deploy and operate independently overseas. Instead, he described networking existing and emerging capabilities into a layered, multi-domain architecture.
Affordability and scalability guide that approach. Guetlein cited a $185 billion estimate in explaining his economics, arguing that integration could avoid the vastly higher costs of a platform-by-platform replication strategy. He said Golden Dome aims to deploy operational capability by summer 2028, then incorporate additional technologies, including NASA collaborations involving automation.
He also distinguished defense from nuclear retaliation. The nuclear triad can deter through the threat of response, he argued, but it does not defeat an incoming attack. Golden Dome must provide credible options against threats below the threshold of nuclear conflict. “If I can’t defeat it, I can’t deter it,” Guetlein said.
Guardrails, Competition and Shared Infrastructure
For Golden Dome, AI must help interpret enormous sensor volumes rapidly. But Guetlein cautioned against conflating AI, automation and autonomy. Each requires a separate judgment about acceptable use.
Hoeven pressed him on congressional guardrails. Guetlein contrasted human-intensive peacetime monitoring with a hypothetical attack involving 3,000 drones headed toward North Dakota. The scale and tempo of conflict, he said, would shape how much automated operation defenders need.
Maintaining competition presents another shared challenge. Guetlein described a command-and-control consortium of approximately 15 providers without a single integrating prime contractor. He said participants allocate work, hold one another accountable and preserve open architecture. Isaacman similarly supported a more diversified launch market.
From scouting for water on Mars to identifying threats headed toward the United States, the missions Isaacman and Guetlein outlined will depend on more than individual breakthroughs in AI or autonomous flight. They will require power, communications, sensors and decision-making capabilities that work together under unforgiving conditions. For the autonomy industry, the opportunity is not simply to build the next vehicle, but to help create the infrastructure that makes sustained exploration and layered defense possible.