One Word, Two Meanings
On September 10, Nvidia and Palantir announced a joint AI stack for supply chain management, built from Nvidia's Nemotron models, Palantir's Foundry platform and its Ontology data layer, plus a new Sovereign AI Operating System Reference Architecture. Nvidia is running it first on its own supply chain, which spans 1.3 million parts per AI server rack and thousands of suppliers. Jensen Huang called supply chains "the operating system of the physical economy." Alex Karp said the stack delivers capabilities that exceed the frontier while protecting an operator's edge.
Both men used the word sovereign. Nvidia and Palantir mean that a customer's proprietary data stays on servers it controls, whether on premises or at a partner data centre. That is a real and useful guarantee. It is also not the guarantee that Germany's own government asked for when it walked away from Palantir this year.
What Germany Actually Objected To
The Bundeswehr excluded Palantir from its next military cloud and AI programme and shortlisted three replacements instead: Almato in Stuttgart, Orcrist in Berlin, and ChapsVision in Paris. Separately, Germany's domestic intelligence agency, the BfV, dropped Palantir for ChapsVision's ArgonOS analysis platform, the same software France's own domestic intelligence service, the DGSI, already runs.
The stated reason was not where the data physically sits. It was who controls the analytical layer, and what happens to that access if a diplomatic relationship sours. A German ministry cannot audit or replace code it does not own, no matter which data centre it lives in. That is a software-sovereignty objection, and it survives untouched by a promise about data location.
What the New Stack Does Not Change
The Nvidia-Palantir supply chain stack keeps a customer's raw data in its own environment. It does not give the customer the Ontology logic that decides what that data means, the Nemotron models that reason over it, or the chips those models run on. All three still come from two American companies, and a customer that adopts the stack for materials allocation, energy grids or pharmaceutical supply chains inherits the same dependency Germany's ministries priced in and declined.
The list of infrastructure partners supporting the reference architecture, Dell, Cisco, Rackspace and Nebius, makes the point sharper rather than softer: a buyer now has to trust the software layer, the chip layer, and a hardware and hosting layer on top of it, all outside its own jurisdiction.
Why This Still Matters for a European Buyer
Nvidia and Palantir named agriculture, manufacturing, pharmaceuticals, retail, technology, aerospace, automotive, energy, healthcare and government as target industries for the stack. Several of those, energy and healthcare especially, sit inside the same critical-infrastructure category that pushed Germany toward ChapsVision in the first place, and are exactly where NIS2-class obligations already require an operator to know who can reach its systems.
The reference deployment inside Nvidia's own operation is a genuinely strong proof point. It shows the software works on a real, complex supply chain at scale. It does not answer the question a European buyer in a regulated sector has to ask first: not whether the tool is capable, but who can be told to switch it off.
Servola Journal
We do this for everyone trying to keep up with what technology is doing to our lives. The people who build it, and the people it happens to. The Servola Journal exists so that what we learn belongs to all of them.
Nobody pays us for this. No ads, no paywall, free to everyone. We just believe that understanding what's happening to all of us shouldn't depend on who can afford to pay for it.
If it gave you something today, tell us to keep going. Follow us, leave a like, or write a positive comment. We read every one, and they are what keeps us going.
Read next: Foxconn's $78.7 Billion Quarter Is a Supply Signal | Germany's New Quantum Computer Needs No Deep Freeze
