PLTR (Trans): US commercial tops Gov. for the first time; FY guide raised again

I. Earnings Highlights Recap

1. Full-Year and Q4 Guidance

a. Full-year revenue: FY26 revenue guidance raised to $8.530–8.534 bn, implying +91% YoY at the midpoint. This is +9 pts vs. prior guidance; U.S. commercial revenue guidance lifted to $3.559 bn+, at least +143% YoY.

b. Full-year profit and cash flow: Adj. OP guidance raised to $5.245–5.249 bn, and Adj. FCF to $4.8–5.0 bn. Management continues to expect positive GAAP OP and net income every quarter for the year.

c. Q4: Revenue of $2.649–2.653 bn. Adj. OP of $1.640–1.644 bn.

2. Bookings, Backlog, and Durations

a. TCV: Total TCV of $2.98 bn this quarter, +6% YoY. U.S. commercial TCV of $1.463 bn, +11% YoY; U.S. commercial TCV over the past 12 months reached $6.114 bn, +62% YoY.

b. Shorter durations: Multiple eight-figure deals were signed this quarter with shorter-than-usual terms. The shift reflects early-stage rapid adoption of sovereign AI, and management expects both duration and size to expand as value is realized.

c. Backlog and retention: Total remaining deal value of $13.4 bn (+55% YoY, +2% QoQ), with U.S. commercial +79% YoY and +4% QoQ. RPO of $4.9 bn (+88% YoY, flat QoQ); net dollar retention at 165%, up 800 bps QoQ.

d. RPO scope: RPO mainly reflects commercial. It excludes contracts with initial terms under 12 months and obligations outside of termination-for-convenience clauses, both common in Gov. work.

3. Margins, Expenses, and Cash Flow

a. Margins: Adj. GPM 85% and Adj. OPM 62%. GAAP OPM 47% (first time above $1 bn in GAAP OP) and GAAP NPM 49%; Rule of 40 reached 158%, up 3 pts QoQ, marking the 13th straight quarterly increase.

b. Expenses: Adj. opex $887 mn, +20% QoQ and +53% YoY. The increase was driven by AIP investment and tech hiring (new grad class); SBC was $319 mn.

c. Cash flow: OCF of $1.205 bn and Adj. FCF of $1.216 bn, with margins of 52% and 53%, respectively. Cash, cash equivalents, and U.S. Treasuries stood at $10.4 bn at quarter-end.

d. Net income: Nearly 8x growth over two years. Q3 2024 net income was $144 mn.

4. Strategic commercial contracts: Revenue was about $0.4 mn this quarter, contributing less than 0.02% of total. Q4 is expected to be under $0.5 mn, with no contribution thereafter.

II. Earnings Call Details

2.1 Key Management Commentary

1. U.S. Commercial: Sovereign AI is accelerating U.S. commercial, and quarterly revenue exceeded U.S. Gov. for the first time. Revenue rose +148% YoY and +29% QoQ, adding $587 mn over the past year; customers reached 686, +29% YoY, and Avg. revenue per customer (TTM) rose +88% YoY.

a. Top-20 customers averaged $146 mn (TTM), +76% YoY. Pilot-to-scale is accelerating: an industrial materials firm scaled from a ~$3 mn pilot to ~$29 mn ACV; a sports brand from ~$1.6 mn to ~$(21) mn ACV; and an apparel firm from ~$2.5 mn to ~$11.5 mn ACV.

b. A home appliance company converted a ~$1 mn pilot to ~$3 mn ACV in the same quarter. It expanded further to ~$15 mn by Q3.

2. AIP and Sovereign AI: The basic unit of the enterprise is workflows, not knowledge workers, and AIP serves as the diffusion layer for AI at scale. Training open-source models on operational data within the Ontology can outperform frontier models on controllable weights; the company launched auto research and continues to invest in AIP Evolve to enable an automated loop for model building, deployment, and evaluation.

a. The AIG commercial underwriting automation cited in Anthropic's prospectus runs end-to-end on AIP. L3Harris fine-tuned a model in under 48 hours that surpassed its prior frontier model, reducing initial costs by ~95%; Tampa General achieved a 7% performance lift and a 22x latency reduction.

b. McCarthy is using a 4 bn-parameter Nemotron and expects to save 30,000 hours annually. In a military exercise, a 30 bn-parameter post-trained model deployed in disconnected, constrained compute outperformed an ~3 tn-parameter frontier model.

3. U.S. Gov.: Revenue increased +96% YoY and +18% QoQ, with Maven as the foundational program. Maven users exceed 100,000 and it is open to all service branches; the Maven hackathon drew 46 teams, and the developer community is 25,000 strong.

a. The Army awarded the prime contract to produce 8 TITAN systems. DORA, a disconnected embedded device, brings Maven into no-network environments.

4. CEO framing: Commercial has surpassed Gov., and the goal is to win the entire market. U.S. commercial growth has been achieved with a very lean salesforce; market value is concentrating in a few companies like Palantir, Anthropic, and OpenAI, and management sees a mission to lead in AI apps and model usage, especially training open-weight models.

2.2 Q&A

Q: How does the sovereign AI boot camp convert participants into customers, and what is the path from attendance to signing?

A: The cycle from first contact to pilot to production has fundamentally changed. Contracts are shorter, but expansion headroom is larger; previously it was a long contact–pilot–long lock-in, whereas now customers arrive because their current solutions do not work, they fear IP leakage, and they are not capturing value.

Once clients see they can train open-source models within their Ontology that beat frontier models without losing alpha, they expand quickly. This resembles the early days with U.S. Gov., but at far larger scale: entry contracts have moved from about $250,000 to roughly $25 mn, and many use cases can be delivered in weeks. The internal bottleneck is resource bandwidth, so resources must be focused on these customers.

Management added that the market is tired of 'reselling tokens and booking tokens as revenue' and is shifting toward what actually works. The boot camp emphasizes showing rather than telling, leading to rapid adoption and scale-up.

Q: Many CEOs still work directly with LLM vendors. How do you persuade unconvinced customers of the importance and risks around the diffusion layer?

A: Demand far exceeds supply, so we do not need to pursue those not yet convinced. Anthropic is expanding the market, while the company serves the portion that is live and needs delivery, and the relationship is symbiotic on the commercial side; they also collaborate in DoD and at AIG.

They rarely poach each other's customers commercially, and 'once you use Palantir, you do not go back'. The only competitive posture is educating customers on the risks of using closed-source models without the application layer; on the Gov. side, they will proactively persuade reluctant agencies.

Management noted that Wall Street, as early investors, may struggle to see operational reality, while industry recognizes Palantir is required for deployment. As a supplement, merely boosting personal productivity may not yield returns; data generated by running supply chain and S&OP on AIP can be used to post-train models that outperform frontier models on those tasks.

At a recent DevCon, a single client built a production-ready model in 48 hours.

Q: The shareholder letter calls Palantir a pioneer of a 'truly viable anti-war movement'. What does that mean, and what is the impact on Gov. business?

A: Deterrence is the core of the product. It aims to make adversaries understand that the cost of war is too high, thereby preventing conflict; management said the product has always balanced counterterrorism with civil liberties and helps bring soldiers home safely.

China is performing exceptionally well in AI, and the U.S. must maintain leadership. It should also rally Europe and other allies, and oppose regulating AI as if it were telecom. Deterrence must be layered: the front line needs 10–100x capability over adversaries, and the rear depends on industrial capacity; after 30–40 years of deindustrialization, the company is working with commercial and public–private partners to restore production using a modern U.S. manufacturing stack.

No specific impact on Gov. revenue was disclosed.

Risk Disclosure and Statement:Dolphin Research Disclaimer and General Disclosure