Scout AI raises $100M to train AI models for military autonomous vehicles
The AMW Read
Scout AI is a new entrant in defense-oriented physical AI (segment 10), with a novel VLA approach and ties to Figure; significant for military use case although early stage.
Scout AI raises $100M to train AI models for military autonomous vehicles
Scout AI, a defense-focused AI startup founded by Colby Adcock and Collin Otis, announced a $100 million Series A round led by Align Ventures and Draper Associates, following a $15 million seed in early 2025. The company is building a Vision Language Action (VLA) model named "Fury" to control military ATVs and drones, starting with logistics and moving toward autonomous weapons. It has already secured $11 million in contracts from DARPA, the Army Applications Laboratory, and other DoD customers. Scout is one of 20 autonomy companies being evaluated by the Army's 1st Cavalry Division for deployment in 2027.
Why it matters: This round signals the deepening of a recurring pattern where AI labs, rather than traditional defense contractors, build frontier models for military use. Scout's VLA approach—adapted from Google DeepMind's 2023 research and used by robotics startups like Physical Intelligence and Figure.AI—represents a structural shift toward using generalist AI architectures in contested environments. The company's close ties to Figure (Adcock sits on the board; his brother founded it) illustrate how foundation-model expertise is being ported into defense, creating a new distribution channel through military procurement rather than enterprise sales. This raises open questions about safety constraints in autonomous weapons and how the military's "fail-safe" requirements interact with probabilistic models.
Grounded expert take: Scout's explicit goal of building "military AGI" and its reliance on VLA models—which require vast amounts of real-world data—push it into a high-risk, high-reward territory. While civilian autonomy relies on structured environments, military operations demand robustness in unstructured, adversarial terrain. The $100 million at Series A suggests investors see defense AI as a large, undersupplied market, but the technology is still early: Scout's ATVs were only six weeks into training at the time of reporting. The outcome will hinge on whether VLA-based models can achieve the reliability needed for lethal autonomy before regulatory backlash or technical failure intervenes.
