In a few years, Anduril Industries, a startup founded by Palmer Luckey, has helped shift the boundaries of American military innovation. From a model structured around large industrial programs, the sector is evolving towards software architectures that are faster to design and more agile in their deployment.
Long perceived as an exception, Anduril now sees new players following suit, starting with Shield AI, which has just announced a funding round of 2 billion dollars with a valuation of 12.7 billion.
This round, led notably by Advent International with the support of JPMorgan, also signals an expansion of capital engaged in defense, now at the intersection of venture capital, private equity, and systemic finance.
The question is no longer just about the emergence of new players, but about where military innovation is now located: in the physical platforms or in the software architectures that pilot them?
At Shield AI, the answer revolves around Hivemind, an autonomy system designed to operate without GPS or communications. The V-BAT drones or the X-BAT project are the visible vectors, but the value lies in the software, designed to be deployed on third-party systems. It is no longer just the equipment that structures the operations, but the systems that orchestrate them.
The acquisition of Aechelon Technology extends this logic, by internalizing advanced simulation capabilities, Shield AI strengthens its control over the training of its systems. In an environment where real data is limited, simulation becomes a central lever, no longer just for testing, but to produce the uniform conditions of learning.
Around these players, a broader ecosystem is emerging, with Palantir Technologies occupying a strategic position on data and command layers, while other companies intervene in data preparation or associated infrastructure. Gradually, a coherent value chain is being established, from sensor to decision.
This restructuring is not limited to the emergence of new entrants; it also redefines the role of historical industrial companies. Groups like Lockheed Martin, Northrop Grumman, or RTX Corporation still hold a central position on complex systems, production, and integration, but their control over the entire chain is gradually eroding.
Faced with more modular architectures, some critical components (software, autonomy, simulation) partially escape them. Industrial players are adapting by developing their own capabilities, forging partnerships, or proceeding with acquisitions, but this evolution is constrained by organizations designed for long cycles, not aligned with software logics.
The result is a hybrid configuration, where historical industrial players retain control over physical platforms, while new players take positions on software and decision-making layers. Production remains essential, but system management, and thus a portion of strategic control, tends to shift.
In this context, the prospect of a new round of Anduril’s funding is being closely watched, as the startup has already received over 6.4 billion dollars in financing. It could confirm the trend where American military innovation is no longer structured solely around industrial capabilities, but at the intersection of software, capital, and strategy. The question remains whether this evolution will lead to a concentration around a few dominant platforms, or to a lasting coexistence between old and new players.
What the American case suggests for Europe
First lesson: value is shifting towards software architectures. Physical systems remain decisive, but value, and some control, are now concentrated in the architectures that pilot, train, and interconnect them.
Second lesson: innovation is organized at the intersection of public and private. States retain control over uses but delegate an increasing part of the design to technological actors, financed by hybrid capital. This articulation redefines traditional decision circuits.
Third lesson: speed becomes a strategic factor. Shortened development cycles, modular architectures, rapid iterations – elements that contrast with historically dominant long program logics.
Fourth lesson: the platform logic is gradually gaining ground. Beyond products, complete environments (software, data, simulation) structure the ecosystem, with potential standardization effects.
For Europe, these elements pose less a question of catching up than one of arbitrage. Integrating these changes requires adapting financing methods, facilitating bridges between industrial and technological actors, and recognizing the structuring role of software in the value chain.
