Palantir's 93% growth is now a procurement argument, not a manifesto

Palantir reported quarterly revenue of roughly $1.94 billion, up 93% year over year, and its CEO used the release to escalate a direct attack on OpenAI and Anthropic, arguing their models strip enterprises of data sovereignty ^[1][3]. That shift matters: Alex Karp is no longer making a values argument about control of data, he is offering an income statement as evidence that buyers pay for it, which turns rhetoric into a thesis a procurement committee can test against its own vendors.

The reported figures, per Cryptobriefing citing 247wallst, come from a Q2 2026 release dated August 3-4 ^[2][19]. GAAP net income was $1.062 billion on that $1.94 billion of revenue ^[4], a GAAP net margin of about 55% ^[15]. The 93% growth rate implies a prior-year comparable of roughly $1.01 billion ^[16]. US commercial revenue growth accelerated to approximately 149% year over year ^[5], about 1.6 times the company-wide rate ^[17].

The product claim behind the sales claim is specific enough to be checked. Palantir says its platform is model-agnostic, so customers are not locked to a single provider's models ^[6], and it advocates open-weight models deployed inside secure, customer-owned environments ^[7]. A partnership with NVIDIA supports sovereign deployments on client infrastructure ^[8]. Karp accuses frontier labs of trying to "drug addict" enterprises into controlled AI futures ^[11] and coined "tokenmaxxing" for what he describes as pushing enterprises toward oversimplified, controlled interactions with their own data ^[12]. He has been making versions of this argument since at least June 2026, with the post-earnings commentary an intensification rather than a new position ^[13].

Here is where a buyer should push back. Growth of 93% is evidence that Palantir sold more software, not evidence that sovereignty commanded a premium. The company's roots are in defense and intelligence, where data sovereignty is a requirement rather than a preference ^[9], so a large share of the demand curve is mandated by contract terms and classification rules, not chosen on the merits by a CIO weighing model quality. Karp's assertion that every enterprise customer has expressed dissatisfaction with frontier-lab dynamics is his own, unaudited and unfalsifiable from outside ^[10]. And the material contains no pricing, no customer counts, and no comparable enterprise revenue figures for OpenAI or Anthropic ^[18], which means the word "premium" is doing work that nobody in this story has priced.

There is also a definitional gap worth naming in an RFP. Running open weights on your own hardware is sovereignty over model weights and data residency. It is not sovereignty over the compute supply chain, and the NVIDIA dependency sits inside the same pitch ^[8]. The national security framing helps Palantir here, because rising concern about foreign exploitation of US technology has made data control a policy question rather than a preference ^[14], and an on-premise, model-agnostic posture sidesteps a lot of that argument ^[14].

What to watch: whether US commercial growth of about 149% holds as comparables harden ^[5], since that segment is the only part of the mix that tests demand outside the mandated-sovereignty base ^[9]; whether the frontier labs answer with credible on-premise or open-weight deployment options, which would collapse the differentiation Karp is selling ^[7]; and whether Palantir ever attaches a number to the premium, because until it does, the 93% is a growth figure being asked to carry an argument about pricing power ^[1].