Defense AI as Battlefield Infrastructure
Ross Fubini, Managing Partner at XYZ Capital, on the defense tech opportunity
Ukraine turned defense AI from a niche software story into a battlefield proven infrastructure layer. The key shift was from selling analytic tools to headquarters staff, toward building the data links, autonomy software, sensors, and counterdrone systems that sit directly inside weapons and operations. Once cheap drones, jamming, and rapid iteration became visible at scale, engineers and investors started treating core defense AI like cloud infrastructure for war, not a back office add on.
-
Palantir helped prove that commercial software could force its way into defense procurement, but its model centered on software plus deployment teams. Anduril pushed the next step, using self funded R&D to wrap software into physical systems like towers, interceptors, and autonomous vehicles that operators can actually deploy in the field.
-
The Ukraine war made the demand signal concrete. Drone swarms, electronic warfare, and constant adaptation showed that military advantage now depends on fast software iteration, onboard compute, sensors, and communications resilience. That is why interest moved from legacy subcontractor work toward core AI infrastructure embedded in frontline systems.
-
This also changed who wants to build in defense. Earlier generations of defense startups looked like specialist contractors selling time and materials into primes. The newer wave pulls talent toward products that can win fixed price programs, compound across multiple missions, and become platforms that other defense products plug into.
Going forward, the winners will look less like pure software vendors and more like defense infrastructure companies. They will own the loop from sensing to decision to action, then reuse that stack across drones, counterdrone, maritime, and ground systems. Ukraine did not just validate a category, it reset the blueprint for how modern defense companies are built.