Digital Sovereignty

Palantir's Sovereignty Pitch Is 19 Percent Foreign

Palantir made AI Sovereignty the organising idea of its Q2 investor deck. Subtract the disclosed US figures from the disclosed global ones and its entire commercial book outside America comes to about 181 million dollars.

Digital SovereigntyBy Servola Tech Desk2026-08-044 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Palantir's Sovereignty Pitch Is 19 Percent Foreign

Key takeaways

Palantir grew 93 percent last quarter, but 81 percent of its revenue came from the United States, and its entire commercial book outside America works out at roughly 181 million dollars, under a tenth of the total.
That 181 million figure is not a disclosure. It is what remains when you subtract the US commercial revenue Palantir published, 764 million dollars, from the global commercial revenue it published, 945 million dollars.
The deck that produced those numbers makes AI Sovereignty its central idea, arguing institutions should not have to choose between sovereignty and capability, which is an argument aimed squarely at European buyers.
US commercial revenue alone grew 149 percent to 764 million dollars, more than four times the entire non-US commercial book, so the gap is widening rather than closing.
When you shortlist a sovereign AI platform, ask what share of the vendor's commercial revenue comes from customers under your own legal regime, because concentration decides roadmap priority and support depth long before any contract clause does.

The quarter, in the numbers Palantir published

Palantir reported second-quarter revenue of 1.935 billion dollars, up 93 percent year on year, with GAAP earnings of 0.41 dollars per share. US revenue reached 1.57 billion dollars, up 115 percent. US commercial revenue, the line the company leads with, grew 149 percent to 764 million dollars. Net dollar retention was 157 percent across 1,049 customers, and full-year guidance went up to between 8.15 and 8.158 billion dollars.

Split by type rather than geography, global commercial revenue was 945 million dollars, up 110 percent, and global government revenue was 990 million dollars, up 79 percent. The company posted a Rule of 40 score of 155. The shares rose about 12 percent on the print, though they remain down roughly 29 percent for the year.

These are extraordinary numbers for a company of this size, and nothing that follows disputes them. The interesting part is not the growth rate. It is what the geography of that growth says about who the product is actually built for.

Subtract, and the sovereignty market appears

Palantir's investor deck makes AI Sovereignty its organising idea. It argues that institutions should own, protect and compound their competitive advantages rather than converting them into training data for external AI labs, and that they should not have to choose between sovereignty and capability. Alex Karp reinforced the theme on the call, attacking frontier labs in characteristically blunt terms, saying they are trying to drug addict us.

That argument has one obvious audience. Sovereignty is not a pressing commercial anxiety in Kansas. It is the organising anxiety of European procurement, where SecNumCloud in France, the C5 catalogue in Germany, the national security framework in Spain and the state cloud in Italy all exist to answer the same question about where data sits and who can reach it.

So it is worth doing the subtraction the company did not do for you. Global commercial revenue of 945 million dollars minus US commercial revenue of 764 million dollars leaves about 181 million dollars of commercial business in the entire rest of the world. That is roughly 143 million pounds, and it is under a tenth of group revenue. Total non-US revenue, by the same method, is about 365 million dollars, or 19 percent. The vendor making the loudest sovereignty argument in enterprise software earns four dollars in five from a single country.

Yes, but: what the arithmetic does not prove

Two honest caveats. First, these are derived numbers, not disclosed segments. Palantir published the global and US lines; the rest-of-world figures here are what remains after subtraction, and a company can allocate revenue in ways that make a clean subtraction slightly wrong at the edges. Second, a US book growing 115 percent mechanically shrinks everyone else's share even if the international business is growing perfectly well in absolute terms. Nineteen percent of 1.935 billion dollars is not a rounding error.

What the arithmetic does establish is a ratio, and ratios are what procurement should care about. US commercial revenue alone is more than four times the entire non-US commercial book. Whatever the international growth rate turns out to be, it is starting from a base small enough that the American commercial customer will keep setting the roadmap for the foreseeable future.

The question to put to a sovereign AI vendor

Sovereignty has become a word vendors say rather than a property buyers verify. Every major platform now has a sovereign tier, a regional boundary and a compliance page. None of that tells you what you actually need to know, which is whether your requirements will survive a roadmap conflict with a much larger customer base somewhere else.

So make the question quantitative. Ask what share of the vendor's commercial revenue comes from customers under your legal regime, ask how many engineers work on the deployment model you are buying rather than the flagship one, and ask which customer segment last caused a published feature to be delayed. In Britain the pattern is already visible in the shape of the book itself: Palantir's largest UK footprint is a public-sector data platform, not a commercial base. A vendor can be excellent and still be built for somebody else.

Frequently asked questions

Did Palantir disclose its non-US revenue?

Not as a separate line in the material reviewed here. Palantir published total revenue of 1.935 billion dollars, US revenue of 1.57 billion dollars, global commercial revenue of 945 million dollars and US commercial revenue of 764 million dollars. The non-US figures in this article are derived by subtracting the US lines from the global ones, which gives roughly 365 million dollars of non-US revenue and about 181 million dollars of non-US commercial revenue.

What does Palantir mean by AI Sovereignty?

In its own framing, that institutions should own, protect and compound their competitive advantages rather than converting them into training data for external AI labs, and that they should not have to choose between sovereignty and capability. The deck sets out a sovereign AI operating system deployable where security demands it, a data plane that works across storage environments, and interoperability through open standards.

Does a small international book make a vendor a bad choice?

No, and it should not be read that way. It is a concentration fact, not a quality judgement. What it changes is your leverage and your position in the queue when priorities conflict. A vendor earning four dollars in five from one country will resolve roadmap disputes in favour of that country, so the practical question is whether your requirements are close enough to the majority customer's that the conflict never arises.

Sovereignty is not a feature a vendor can declare. It is a question about whose money pays for the roadmap, and that answer is published every quarter for anyone willing to do the subtraction.

PalantirAI sovereigntySovereign cloudVendor concentrationEnterprise AIProcurementEarnings

Ten Groups Built a Free Chatbot in Seven Weeks

Current AI holds 400 million dollars in committed funding, seeded by the French government. Its open Alpha Chat took seven weeks to build. Here is what that does to your next AI quote.

5 min read

Digital Sovereignty

Your Business Data Just Got a European AI Lab

SAP has completed its acquisition of Germany's Prior Labs for more than 1 billion euros and pledged another 1 billion to build a European frontier AI lab for structured business data. Why the enterprise AI prize is your tables, not a chatbot.

3 min read

Law & Regulation

The EA Buyout Now Hinges on Brussels

The largest take-private in history has cleared EA shareholders and US antitrust. The last real gate is the EU - and not classic merger review, but the Foreign Subsidies Regulation, facing its first blockbuster test.