The shortlist Germany will choose from next year for a military data platform has one conspicuous absence: the American company whose software is already installed across allied headquarters.
Berlin has been assembling candidates for an AI platform that sorts and distributes military intelligence, and Palantir Technologies Inc. is not among them, according to a person familiar with the deliberations who spoke on condition of anonymity because the process is not public. The systems that did make the cut are less capable, this person said. They are also sufficient, they keep the data inside German control, and they can cover about 80% of the requirement within a year.
That trade, capability for control, is being made in several European capitals at once.
Defense ministries across the continent have begun looking for alternatives to American suppliers, Reuters reported Monday, and a group of domestic companies has started competing for the work. The shift is not a rejection of the technology. It is a decision about who holds the switch.
Germany’s own candidate arrived last week. Helsing, the Munich-based defense AI company, released a military intelligence product called Clarity, built for the kind of analysis that turns intercepted signals and satellite imagery into something a commander can act on. Helsing has become one of the most valuable defense startups in Europe by selling software rather than hardware, and Clarity gives it a product aimed at the exact layer Palantir occupies.
France is taking a slower route to the same destination. It is developing a sovereign military AI platform called Arcadia, with some contracts scheduled to be awarded before the end of the year and deployable capability targeted for the middle of 2027. The Netherlands has said it wants a domestic substitute within two years.
Palantir’s answer is that the fear is misplaced. Louis Mosley, who runs the company’s international business, said the systems NATO uses have no off switch. The point is technical and contractual: the software runs on customer infrastructure, under customer control, and the vendor cannot reach in and disable it.
That answer addresses the wrong question. The concern in European capitals is not that Washington would flip a switch in the middle of a conflict. It is that a targeting decision, a logistics plan or a piece of analysis would sit inside a system governed by American law, maintained by an American company, and priced by an American shareholder base.
Displacement is hard for reasons that have nothing to do with politics. Palantir’s product is not a database. It is an integration layer that maps an organization’s data onto a model of its operations, and the work of building that map is done by engineers who sit inside the customer’s building for months. Once a ministry’s intelligence flows through that map, replacing it means rebuilding the map, retraining the analysts, and re-certifying the workflows.
That is why the 80% figure matters more than any comparison of features. European buyers are not trying to match the incumbent capability for capability. They are asking what share of daily work a domestic system can absorb, and whether the remaining fraction can be handled by keeping both systems running.
Dual running is the likely outcome in Germany, and it is expensive. Two platforms, two integration efforts, two sets of clearances. Ministries accept the cost because the alternative is a single point of dependence that their parliaments have started asking about.
Procurement is the constraint on the European side. National programs are separate, budgets are annual, and certification of software that touches classified networks takes years. Helsing’s advantage is that it has already cleared some of those hurdles for other products. France’s advantage is that it can write the requirement around its own industrial base.
Palantir has spent several years expanding into European health and defense ministries, and its contracts have drawn scrutiny in Britain, where its work for the National Health Service became a political argument about patient data. Each of those episodes gave European procurement officers a preview of the objections they would face at home.
The commercial question is what a smaller European footprint costs the company. Its international business has been the part of the story investors watch for growth beyond the American government, and defense ministries are the customers most willing to pay for integration work. Losing the German platform would not remove Palantir from Europe. It would remove the assumption that the data layer of European defense settles on one vendor.
There is a version of this that ends well for the incumbent. Domestic systems cover the routine 80%, budgets run over, and the hard problems come back to the vendor that has already solved them. Palantir would then sell into a market with more buyers and more political cover for buying American.
There is a version that ends worse. The 80% turns out to be the part that matters, the remaining 20% gets absorbed by the domestic system in the second release, and the ministry discovers it no longer needs the second contract.
Either way, the next twelve months decide it. France awards its first Arcadia contracts by the end of the year. Germany publishes a shortlist. The Netherlands starts evaluating.
Somewhere in that sequence a colonel will open a screen, look at a map of what his analysts have collected, and use whatever is on it. Whichever system he learns first is the one that will be hardest to take away from him.


