Eliza Ward: Hey — before we get going, I want to just sit in one number for a second. Actually, wait, not a number — the absence of one. Palantir and Armada put out a full partnership announcement October 1st and disclosed nothing. No contract value. No customer. No volume.
Brian Reed: Nothing at all?
Eliza Ward: Zero commercial disclosures. What they did say: Palantir named Armada its inaugural Certified Modular Data Center Partner. Armada's Galleon data centers carry Palantir's Sovereign AI OS — AIP, Ontology, the full stack. And Alex Karp is quoted: the infrastructure is manufactured in the United States and allied nations and delivered in months. That's the announcement.
Brian Reed: Let me see if I have this right. Sovereign AI — so this is essentially governments or enterprises saying they want to own the whole stack. Not renting compute from Amazon or Microsoft, but actually running their own hardware, their own models, walls up around all of it.
Eliza Ward: That's the core idea, yes. Own the building, don't lease the desk.
Brian Reed: So Karp's quote is doing a lot of work. Because manufactured in months, allied nations — that sounds like a selling point. But if there's no customer confirmed, who is he selling it to right now?
Eliza Ward: That's — yeah. That's exactly the silence I can't get past.
Brian Reed: And then a week later — October 8th — Goldman upgrades to Buy, $230 target. Barclays initiates at Overweight, $265. Both citing sovereign-AI demand. Both on the same day. Based on... the announcement with nothing in it.
Eliza Ward: The $265 is what gets me. That's not a rounding error above where the stock is trading.
Brian Reed: Stock's at $208, roughly, as of October 9th. So Barclays is pricing in another 27% from here. And then — I mean, this is the part that actually stops me — Simply Wall St has a price target of $90.50.
Eliza Ward: Wait. $90.50? That's — okay, that's not a different opinion. That's $175 below Barclays on the same stock, reading the same public filings.
Brian Reed: Right — and that's what I'm trying to separate out. When the spread is $175 wide, someone isn't disagreeing about fundamentals. Someone is pricing the story. The question is which camp is anchoring to an actual number and which one is anchoring to, let me see, the idea that sovereign AI is inevitably enormous.
Eliza Ward: There's a William Blair note that shows exactly how this works. They flagged Palantir as a beneficiary of Anduril's $1.8 billion, five-year Army NGC2 contract — but Palantir's role is in the cloud data layer. That's inferred. That's not a disclosed contract. An upgrade built on an inference about a contract that isn't confirmed as Palantir's.
Brian Reed: That's the pattern exactly. And honestly — there's another piece of this that I think is getting blurred together in ways that make the whole thing even messier, but we'll get to that in a minute because it involves a separate company that's being folded into the same sovereign-AI thesis and probably shouldn't be.
Eliza Ward: That separate company — that's the exact problem. Because on October 12th, eleven days after the Armada deal, Ondas Holdings and Palantir launch Regna AI. Standalone sovereign-AI spinout, initially wholly owned by Ondas, built on Palantir AIP. And the take already circulating is: Palantir has two sovereign-AI wins in one week. That's wrong.
Brian Reed: Two separate structures, though — is the conflation innocent?
Eliza Ward: Probably both? I mean — Armada is a partner certification. Regna AI is a spinout. Different partners, different legal structures, different target customers. One is Ondas's company. The other is Palantir certifying a hardware vendor. Neither discloses a paying customer. But if you're a defense procurement officer sitting down October 15th and you read both press releases over coffee, you're seeing two sovereign-AI moves from Palantir in the same morning news digest. You might — wait, actually you'd be forgiven for assuming Palantir already has two deployments underway.
Brian Reed: And the confirmed number of paying sovereign-AI customers across both deals is zero.
Eliza Ward: Zero disclosed. Right. AIP as a licensing layer embedded in partner-led vehicles — that's a real pattern across both deals. It's genuinely interesting architecture. But the ecosystem narrative is aspirational until someone names a contract.
Brian Reed: So the pattern is real. The revenue is just — not yet confirmed. And the conflation is doing work on the valuation whether anyone intends it to or not.
Eliza Ward: And that's — I mean, the pattern closes on itself there. Zero disclosed paying customers, two partnership structures, two analyst upgrades. The thing that actually settles it isn't another press release. It's Q4 2026 earnings. Q1 2027. Does Palantir report actual sovereign-AI customer wins? Actual contract values? Because that $175 spread between Barclays at $265 and Simply Wall St at $90.50 — that doesn't compress on another announcement. It compresses on bookings.
Brian Reed: And if those earnings land and sovereign-AI bookings are still zero — or still undisclosed — Goldman and Barclays are pointing the wrong direction. Fast.
Eliza Ward: Fast and not gently. Right.
Brian Reed: The part I still don't have an answer to — and I'm not sure the research does either — is whether the manufactured-in-the-US constraint is actually a competitive edge or a drag. Karp is selling it as a feature. Galleon ships in months, not years, allied-nation supply chain, full sovereign control. But, hang on — months is still slower than a hyperscaler spinning up capacity in days. So the question procurement offices are going to ask is: is the sovereignty premium worth the wait and the cost? And no partnership announcement can answer that. That's a procurement cycle question. Could be two years before we know.
