Europe talks about AI driven weapons and home grown AI. The harder problem is the physical stack underneath: sensors, semiconductors, and jam proof navigation (the hardware layer that keeps platforms working when satellite signals are cut off), much of which still comes from…
A one-way drone enters a corridor where GPS has been jammed or spoofed. It has good software. It has a clear mission. It still has to know where it is, orient without satellites, avoid obstacles in the dark, and stay functional when the signal it was relying on has disappeared. That gap, between what an algorithm can do and what the physical stack underneath it can survive, is where European defense autonomy is actually failing, and almost nobody in the public debate is naming it.
The case for treating that gap as the real story is laid out in a recent EE Times analysis of Europe's drive for defense technological independence. The piece argues, in effect, that autonomy is not an AI problem. It is a sensors, chips, and jam-proof navigation problem, and Europe is short on all three. The frame matters because the conversation about military AI is still dominated by what models can do, not by whether the hardware they run on will keep working when the electromagnetic environment turns hostile.
The war in Ukraine has done more than any white paper to make the inversion visible. Commercial kit, including Wi-Fi networks, smartphones, publicly available messaging apps, and low-cost drones, is now baked into combat coordination on both sides. Widespread GPS jamming and spoofing have exposed, on the other side, the soft underbelly of any autonomous system that depends on unprotected satellite navigation. Resilient navigation is, in the language of the EE Times piece, strategically indispensable. Without it, the drone, the autonomous naval vessel, the robotic rover, and the intelligent satellite are all running on a single point of failure.
So what does "the stack under the AI" actually contain? It is the layer an autonomous platform cannot do without: cameras and radar for perception in degraded conditions; LiDAR, or laser-based depth sensing, for three-dimensional mapping; GNSS receivers, the Global Navigation Satellite System family that includes GPS, for position, velocity, and time; inertial measurement units to carry that position forward when GNSS drops out; embedded processors and edge AI accelerators (machine-learning chips designed to run perception and decision-making onboard, without round-trips to a remote data center) for real-time decisions; secure radio links; and resilient data infrastructure on the ground. None of this is glamorous. All of it is where autonomy actually breaks or survives.
The strategic point, per the EE Times argument, is that defense autonomy is not created by software alone. The under-the-stack hardware, components, and resilient data plumbing are the gating bottleneck. Europe talks about AI-powered weapons and sovereign cloud. It talks much less about who builds the GNSS receiver, the anti-spoofing module, the radiation-hardened processor, or the foundry that fabricates any of it. Septentrio, a Belgian specialist in jam-resistant GNSS receivers and one of the worked examples in the EE Times analysis, sits inside exactly that problem. Its work on multi-frequency, anti-spoofing receivers, often fused with inertial sensors to keep a platform oriented when satellite signals are jammed, is the kind of unglamorous capability that decides whether an autonomous system survives its first minute in a contested environment. It is one example of a layer Europe already has. The wider question is how many more it needs.
That question is also where the rhetoric outruns the supply chain. The argument for technological independence is built on reducing dependence on large foreign tech ecosystems. The reality, as the same analysis concedes, is that European defense still leans on non-European foundries, intellectual property, software platforms, and rare-earth processing. "Technological independence" in the current decade does not mean autarky. It means enough depth across the layers of the stack that an adversary cannot turn off the lights by choking two or three suppliers. Whether Europe is buying, building, or partnering its way to that depth, and on what timeline, is the open question the policy debate has not yet answered.
What to watch next is concrete. The first signal will be procurement: whether European defense ministries begin writing jam-resistant positioning, navigation, and timing (PNT) into platform requirements as a baseline, rather than treating it as an add-on. The second is industrial: whether more European champions at the layer below AI, the chip designer, the sensor maker, the secure-comms vendor, attract sovereign or pan-European capital at a scale that matches the political rhetoric. The third is operational: whether the lessons from Ukraine about commercial, low-cost, GPS-fragile drones get translated into doctrine, or whether European militaries continue to procure for the network they wished they had rather than the electromagnetic environment they will actually fly in.