The AI Race Enters the Physical World

A German Marshall Fund analysis argues that the AI race has entered the physical world. The analysis states that for the United States the binding constraint is deployment, and that producing more frontier models matters less if the operational substrate for embodied systems is not at hand. The frame is national rather than global: the question it poses is not whether frontier models will keep improving, but whether a physical and operational base exists to receive them.

The analysis names four companies as having made meaningful progress building that base domestically: Figure, 1X, Physical Intelligence and Waymo. It characterises their efforts as remaining in early stages. The excerpt provides no deployment figures, funding amounts, timelines or performance benchmarks for any of the four, and does not rank them against one another.

The central claim is about constraint rather than capability. Under this framing, further gains in frontier model production do not by themselves determine the outcome of the race, because embodied systems require an operational base that model development alone does not supply. The analysis presents the domestic build-out as underway but incomplete, and does not indicate how quickly it might progress or what would count as sufficient. The excerpt contains no policy recommendations and does not address how the deployment constraint relates to regulation or liability.