Third quarter 2026 earnings call, November 2, 2026
Palantir delivered the strongest growth in its history as a public company in the third quarter. Revenue rose 96% year-over-year and 20% sequentially to $2.313 billion, and quarterly revenue crossed $2 billion for the first time. The more consequential change is in the mix. U.S. commercial revenue of $984 million, up 148% year-over-year and 29% sequentially, exceeded U.S. government revenue of $953 million for the first time. Management credits the shift to a newly launched "sovereign AI" offering that it says is only one quarter old. The forward indicators are less dramatic than the revenue line, and the call offered few details on how durable the demand is.
Revenue growth is outrunning bookings, and contracts are getting shorter
The most important nuance for investors is the gap between reported revenue and forward commitments. U.S. commercial total contract value bookings were $1.463 billion, up only 11% year-over-year, against revenue growth of 148%. Total commercial TCV bookings of $1.652 billion grew 17%, and total company TCV bookings of $2.918 billion grew 6%. Total remaining deal value reached $13.4 billion, up 55% year-over-year but only 2% sequentially. Remaining performance obligations were $4.9 billion, up 88% year-over-year and flat sequentially. U.S. commercial remaining deal value grew 79% year-over-year and 4% sequentially.
CFO David Glazer offered an explanation tied to the new product. "We signed multiple 8-figure deals in the quarter that were shorter in duration than usual, as customers rapidly adopted initial sovereign AI capabilities in their enterprise," he said, adding that the company expects "both the duration and size of these contracts to increase" over time. CEO Alex Karp described the same pattern: "much quicker interaction, much quicker to deployment, shorter duration of the contract, but then an ability to expand that contract dramatically." That is a plausible land-and-expand model, and the examples given support it. Still, the trailing 12-month U.S. commercial TCV figure of $6.114 billion, up 62%, is growing far more slowly than current revenue. Investors are being asked to trust that expansion will fill the gap. Net dollar retention of 165%, up 800 basis points sequentially, is the strongest evidence that it is doing so.
What sovereign AI is, and why customers are paying for it
Palantir's pitch is that enterprises can post-train open-weight models on their own operational data inside the Ontology, retain control of the weights, and reach performance above frontier closed models on specific tasks at much lower cost. CTO Shyam Sankar framed the argument directly. "The fundamental unit of the enterprise is the workflow," he said, rejecting what he called the labs' conceit that the individual knowledge worker is the atomic unit. "You can make any set of workers more productive and generate no actual ROI."
He described the economic logic as a pipeline rather than a model. "The specific posttrained model is a commodity item. The compounding value is in the posttraining pipeline that is constantly fed from your operational exhaust captured in the Ontology." Palantir is building toward an automated loop through a new AutoResearch feature and AIP Evolve, which would continuously build, deploy and evaluate models against ROI.
Management cited several performance claims, all company-reported and several preliminary. L3Harris said a fine-tuned open-source model outperformed its frontier model in under 48 hours and cut costs by 95% in preliminary results. At a recent PACOM exercise, Palantir deployed a 30 billion parameter post-trained model on constrained edge hardware that it said outperformed GPT 5.6 Sol, described as a 3 trillion parameter model, on a specific task. NVIDIA said it trained its smallest model to beat its largest and can now retrain it on a single GPU in a couple of hours. McCarthy Building Companies expects to save 30,000 hours annually using a post-trained 4 billion parameter Nemotron model for subcontractor drafting. Sankar said the primary benefit is not cost: "the dominant variable is that more use cases can actually get to the threshold of viability."
Pilot-to-production conversions are the clearest demand signal
Chief Revenue Officer Ryan Taylor gave unusually specific conversion examples. An industrial materials company moved from a pilot of nearly $3 million in the second quarter to a $29 million ACV deal. A global athletic company converted a $1.6 million pilot that began in the first quarter into roughly a $21 million ACV deal. A global home appliance company went from a $1 million pilot to a $3 million ACV deal in the same quarter, then expanded nearly fivefold to $15 million across commercial, finance and supply chain workflows. An apparel company moved from a $2.5 million pilot after an AgentCamp to an $11.5 million ACV deal.
Karp said the sales motion has changed: deal sizes are "$25 million to start" rather than $250,000, and customers arrive after hearing from peers. "It's a little bit like a rock concert," he said of the sovereign AI bootcamps. He also acknowledged a constraint. "What we have to do internally is figure out how to re-resource our company to focus exclusively on those people because we don't even have the bandwidth to do that." Palantir sells with very few salespeople, so capacity could cap the near-term conversion rate even if demand holds.
Positioning against Anthropic and the labs
Analyst Gil Luria pressed on how Palantir persuades executives who buy directly from model labs, noting that about half of Anthropic's customers do. Karp argued the two are complementary. "They are also selling our product because people interact with them... what they are doing is increasing the size of the market. What we're doing is servicing the market that's gotten real." He said Palantir rarely tries to displace a lab's customers: "enjoy the one you're with until you find out it's not the one you want." He added that the company's only active competitive effort is educating clients on "the risks of using closed models without an application layer."
Sankar cited page 209 of Anthropic's S-1, which he said describes AIG's automation of commercial insurance underwriting, and said "that entire workflow... runs in AIP and was implemented by Palantir and AIG." Karp's claim that "once you go Palantir, you don't go back" is unverified, and his comment that Wall Street investors in the labs find it "hard to see the reality when you're getting rich" is a rhetorical jab rather than evidence. The reliance on a competitor's filing to validate Palantir's role is notable, though it also shows how much of the AI value chain Palantir is exposed to.
Government: Maven scales, but growth is outpaced by commercial
U.S. government revenue grew 96% year-over-year and 18% sequentially to $953 million, with total government revenue at $1.146 billion, up 81%. Maven now has more than 100,000 users and is available to all military services rather than only combatant commands. The Army awarded Palantir a prime agreement to produce 8 TITAN systems. Palantir also highlighted DORA, an embedded compute device that brings Maven to disconnected environments. International remains the weak spot: international commercial revenue grew only 21% year-over-year and 1% sequentially to $184 million, and international government grew 31% to $192 million. Karp deflected on Europe, saying "we should kind of avoid talking about it."
Financials and guidance
Profitability remained extraordinary. Adjusted operating income was $1.427 billion, a 62% margin, and GAAP operating income was $1.085 billion, a 47% margin. GAAP net income was $1.128 billion, nearly eight times the $144 million of the third quarter of 2024. Adjusted free cash flow was $1.216 billion, a 53% margin. Both GAAP and adjusted EPS were $0.44. The Rule of 40 score reached 158%, the 13th consecutive quarterly expansion. Adjusted expenses rose 53% year-over-year to $887 million, driven by AI platform investment and new graduate hiring. The company ended the quarter with $10.4 billion in cash and Treasuries. Net dollar retention was 165%, and the top 20 customers averaged $146 million in trailing 12-month revenue, up 76%.
For the fourth quarter, Palantir guided revenue to $2.649 billion to $2.653 billion, implying roughly 15% sequential growth, and adjusted operating income of $1.640 billion to $1.644 billion. Full-year revenue guidance was raised to $8.530 billion to $8.534 billion, a midpoint of $8.532 billion and 91% growth, a 9-point increase from last quarter's implied guidance. U.S. commercial guidance was lifted to more than $3.559 billion, or at least 143% growth. Adjusted operating income guidance rose to $5.245 billion to $5.249 billion, and adjusted free cash flow guidance to $4.8 billion to $5 billion. Revenue from strategic commercial contracts was about $400,000 and is expected to be below $500,000 in the fourth quarter, with none thereafter.
Karp's commentary: valuation and mission
Karp said the shift in the commercial business "probably justifies our current valuation," and described the company's aspiration "to take the whole market" as no longer looking "that crazy." In response to a question from Deutsche Bank's Brad Zelnick about the shareholder letter's reference to an "anti-war movement," he said Palantir's products are designed to deter conflict. "I grew up as a lefty. I don't actually like war," he said, but he argued the U.S. must lead China in AI and that deterrence comes from tools adversaries know work. Sankar added that deterrence also depends on production capacity and a "modern American manufacturing stack." Karp closed with a combative note to retail shareholders, saying he wants "the people that made fun of you and made fun of us" to be "suffering," a tone that will not help in institutional conversations.
The quarter confirms that Palantir's U.S. commercial engine has moved beyond anything the company previously delivered, with strong retention and conversion evidence. What remains unproven is whether shorter-duration contracts and 11% bookings growth will sustain revenue growth near these levels once the first wave of sovereign AI adoption is absorbed. Management also gave no quantified view of that wave's size or of the sales capacity needed to meet it.