Palantir Q3 2026 Results: What 148% US Commercial Growth Says About Enterprise AI

Palantir reported $2.313 billion in third-quarter revenue on November 2, up 96% from a year earlier. US commercial revenue rose 148% to $984 million, passing US government revenue for the first time. For enterprise AI builders, that shift deserves more attention than another model leaderboard.

Palantir Q3 2026 results at a glance

The earnings release filed with the SEC covers the quarter ended September 30, 2026.

Metric	Q3 2026 result
Total revenue	$2.313 billion, up 96% year over year
US commercial revenue	$984 million, up 148% year over year
US government revenue	$953 million, up 96% year over year
GAAP operating margin	47%
Adjusted operating margin	62%
Full-year revenue guidance	$8.530 billion to $8.534 billion

The guidance is management’s forecast, not revenue already earned. The two operating margins also measure different things: the adjusted figure excludes costs including stock-based compensation and related employer payroll taxes.

What changed in the commercial business

The crossover is about the United States, not a claim that government work has disappeared. Both US businesses grew sharply. Commercial revenue simply grew faster.

Palantir’s quarterly presentation reports net dollar retention of 165%, eight percentage points higher than the previous quarter. That metric compares trailing-twelve-month revenue from an existing customer cohort with the same cohort’s earlier revenue. It is evidence of expansion inside customer accounts, not a count of newly acquired customers.

For founders selling enterprise software, this is the more useful distinction. Winning a pilot and becoming something a customer buys more of are different achievements. A product can generate enthusiastic demos without becoming important enough to expand across a business.

What Palantir AIP actually does

Palantir’s AIP documentation describes a platform for connecting AI to company data and operations. Its tools support workflows, agents and functions built on the Ontology, alongside security controls, evaluations and auditing.

The Ontology provides a representation of business objects and their relationships. In practical terms, an enterprise application needs more than an answer to a prompt. It needs to know which customer, order or asset an answer concerns, what the current state is, and what actions the user is permitted to take.

Consider a delayed shipment. A chatbot can explain why deliveries run late. An operational application needs to identify the affected order, check available alternatives and record an approved decision. The difficult work sits in those connections and controls as much as in the model’s response.

That is why these results matter beyond Palantir. There is a difference between selling access to intelligence and getting intelligence into a workflow someone already depends on. The latter creates more integration work, but also a clearer reason for a customer to keep using the product.

What these results do not prove

The reported revenue is Palantir revenue. The release does not provide a standalone AIP revenue figure or isolate how much growth each AI capability caused. Treating the entire quarter as pure generative-AI revenue would overstate what the numbers establish.

Nor does supplier growth measure a customer’s return on investment. Buyers still need to compare deployment cost, ongoing operating cost and the actual change in their business process. A strong vendor result cannot answer that calculation for them.

For builders, the useful next question is whether a deployment earns its expansion. Can the first workflow make the second one easier to implement? Can a customer see the benefit without relying on a persuasive demo? Palantir’s quarter makes those questions harder to ignore. It does not remove the need to answer them.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help founders explain complex enterprise AI products through clear technical and executive stories.

Book a 30-minute call.