Palantir’s US commercial revenue grew 148% year over year, while its US commercial customer count increased 29%. That difference makes expansion within existing relationships more interesting to me. It does not establish that AI agents improved customers’ work enough to explain the revenue increase: the public figures do not separate usage, pricing and customer mix. Results Customer data
The company released its third-quarter results on the US afternoon of 2026-11-02, at 05:05 on November 3 in Taipei, before this article’s 08:36 cutoff. This is new information on the article’s Taipei publication date. The underlying business activity belongs to the quarter ending in September, not to a single day of new usage. Release timestamp Independent coverage
Quarterly revenue reached $2.313 billion, up 96%. US commercial revenue of $984 million exceeded US government revenue of $953 million for the first time. Palantir also reported a 47% GAAP operating margin. These figures establish recognized revenue and operating profit for the software supplier. They neither isolate AIP revenue nor account for the integration costs incurred inside each customer’s organization. Financial results Business comparison
US commercial customers numbered 686, measured by revenue recognition over the preceding 12 months rather than new users in this quarter. Company-wide net dollar retention was 165%, comparing revenue from the same customer cohort across successive 12-month periods. It supports an expansion interpretation, but is not a US-commercial-only expansion rate. Metric definitions
Hexion gives agents business context
At September’s AIPCon 11, Hexion described agents using tank levels, consumption and delivery windows to propose delivery plans, then connecting these to raw-material procurement through Ontology. Business relationships and decision evidence support human review. This was an earlier customer account, not a new quarterly feature or an independently verified investment-return comparison. Customer presentation transcript
I find this reusable business context a promising product direction. Suppose a company has already established product definitions and customer permissions. An additional quoting application might reuse that work instead of rebuilding rules for each agent. Reuse depends on keeping information current and permissions appropriate to the task. If every new workflow requires another integration project, buying more software may also mean accepting more maintenance work.
That distinction changes how an enterprise AI product should be evaluated. Answering one query and allowing another department to reuse the same business definitions have different cost implications. Palantir’s revenue and retention results give the latter approach a stronger commercial basis. They do not show that every customer follows it, or establish equivalent returns for Taiwanese manufacturers. Differences in data maturity and existing systems still determine how much work a deployment requires.
There is also a constraint in the quarter’s figures. US commercial total contract value signed was $1.463 billion, up 11%, considerably slower than recognized revenue. Contract value includes options, and most contracts can be terminated; it cannot simply be counted as future revenue. Expansion by existing customers can support current growth, while replenishment through new contracts and renewals remains a separate question. If revenue keeps rising rapidly while new contract growth continues to slow, the durability of that expansion needs reassessment. Contract figures and limitations Financial cross-check
The cover reuses this site’s historical Open Secure AI Alliance graphic, which includes Palantir’s logo. It is neither a quarterly-results chart nor a new partnership announcement. Download-side DNS resolution failures prevented retrieval of the official earnings visuals.
Sources: