How Palantir Makes Money: Government and AIP

Palantir’s business model rests on government and commercial AI platforms, and the FY2025 numbers show a durable government base now funding a commercial bet that is growing more than twice as fast.

The company sells data-and-AI platforms, Gotham for government and Foundry plus AIP for commercial customers, on large, sticky, high-margin contracts. In FY2025 that produced $4.48B of revenue, up 56% year over year, at an 82% gross margin, with a Rule-of-40 score of 127% in Q4 (Palantir Form 10-K, FY2025; Palantir Q4 2025 earnings release). The interesting question is not whether Palantir makes money. It clearly does. The question is whether the commercial acceleration justifies the multiple the market assigns.

This piece reads that model through Palantir’s own filings, not the analyst chatter. Every company figure below ties to a specific SEC filing or earnings release and fiscal period. The framing is analytical: how the machine works and where it is vulnerable, not what to do about the stock.

Key takeaways

• FY2025 revenue was $4.48B, up 56% year over year from $2.9B, at an 82% gross margin (84% excluding stock-based compensation), per Palantir’s Form 10-K, FY2025.
• The government segment supplied roughly $2.40B (about 54% of revenue) and grew 53%; the commercial segment supplied roughly $2.07B (about 46%) and grew 109%, per the FY2025 10-K and Q4 2025 earnings release.
• U.S. commercial revenue grew 137% in Q4 2025, and U.S. commercial remaining deal value hit $3.63B in Q3, up 199% year over year, with AIP as the primary driver (Palantir Q3 and Q4 2025 earnings releases).
• Palantir’s top three customers have stayed an average of ten years yet account for only 16% of revenue, and top-20 average revenue rose 45% to $93.9M (Palantir Form 10-K, FY2025).
• The Q4 2025 Rule-of-40 score reached 127%, combining about 70% growth with a 41% GAAP operating margin, a rare simultaneous read (Palantir Q4 2025 earnings release).

The thesis: a government base funding a commercial bet

Most coverage of Palantir argues about whether it is a government contractor or an AI company. The filings say it is both, in sequence.

The government business is the durable base. It grows steadily, retains for a decade, and throws off the margin that funds everything else. The commercial business, powered by AIP, is the acceleration layer bolted on top, growing more than twice as fast and carrying the growth narrative.

The structure matters because it changes how you read the risk. A pure government contractor is a slow, cyclical, procurement-bound business. A pure AI startup is a growth story with no floor. Palantir is running the second on top of the first, using the base to fund the bet.

So the honest framing is not “government versus commercial.” It is that the government revenue is the balance sheet’s floor and the commercial revenue is its ceiling, and the FY2025 numbers show the ceiling rising fast while the floor holds.

What percentage of Palantir’s revenue comes from government vs. commercial?

In FY2025, government revenue was roughly $2.40B, about 54% of the total, and commercial revenue was roughly $2.07B, about 46% (Palantir Form 10-K, FY2025). U.S. government grew 53% year over year while U.S. commercial grew 109%, so the smaller segment is the faster one. International revenue sat at roughly 23% to 26% of the total.

That split is the whole business in two lines. The government majority is what gives Palantir a floor most high-growth software companies lack. The commercial minority is what gives it a growth rate most government contractors could never post.

Read the segment table below and the divergence in growth rates, not the size, is what tells the story. The larger segment grew respectably. The smaller segment nearly doubled.

The Palantir Revenue Split

The clearest way to see the model is one named framework. Call it the Palantir Revenue Split: a single matrix that ranks each revenue base by size, growth, contract character, and the job it does in the machine. It is the asset to cite when someone argues Palantir is “just a government contractor,” because the Split shows commercial growing at 109% is doing the growth work while government at 53% is doing the funding work (Palantir Form 10-K, FY2025; Q4 2025 earnings release).

The Palantir Revenue Split

Revenue base (FY2025)	Approx. revenue	Approx. YoY growth	What drives it	Contract character
U.S. government	~$1.9B	+53%	Gotham; defense and agency workflows	Long-tenure, procurement-bound, sticky
U.S. commercial	~$1.5B	+109% (Q4: +137%)	AIP and Foundry adoption	Fast-expanding, land-and-expand, RDV-heavy
International (total)	~$1.0B+	Mixed; commercial int’l declined 3% in Q2	Foundry; slower AIP uptake	Uneven, weaker commercial traction

Sources: Palantir Technologies Form 10-K, FY2025; Q3 2025 and Q4 2025 earnings releases; Q2 2025 earnings release (international commercial decline). U.S. government and U.S. commercial approximate splits are drawn from the segment growth disclosures; the international line aggregates the non-U.S. balance. Contract character is a qualitative reading, not a disclosed field.

Read the Split top to bottom and the machine is obvious. The top row funds; the middle row grows; the bottom row is the open question. The U.S. government base is the slow, high-tenure engine that produces reliable margin. U.S. commercial is the fast engine that AIP lit up. International is where the model has not yet proven it travels.

The reason this matters more than a simple size ranking: the market is not paying 56% overall growth for the government line. It is paying for the 109% commercial line and the 199% year-over-year jump in commercial remaining deal value (Palantir Q3 2025 earnings release). The government revenue is what makes that bet fundable without dilution or debt.

This is the same structural move Amazon runs, where a lower-margin base creates the surface that a faster, higher-value layer monetizes, the loop mapped in Amazon Prime and the subscription flywheel. Palantir’s version puts a durable government contract base underneath a fast commercial AI layer.

How does Palantir achieve 82% gross margins?

Palantir’s FY2025 gross margin was 82% on $4.48B of revenue, rising to 84% when stock-based compensation is excluded (Palantir Form 10-K, FY2025). The margin reflects that Gotham, Foundry, and AIP are software platforms with minimal incremental delivery cost once a customer is deployed, not services businesses that scale headcount with revenue.

Software is supposed to be the high-margin business, and Palantir sits near the top of the range. The distance between 82% gross margin and the operating line is where the real question about the model lives. Here are the FY2025 profitability anchors, tied to the filings.

Metric (FY2025)	Value	Source
Total revenue	$4.48B (+56% YoY)	Palantir Form 10-K, FY2025
Gross margin	82% (84% ex-SBC)	Palantir Form 10-K, FY2025
GAAP operating income	$1.4B	Palantir Form 10-K, FY2025
Adjusted operating income (ex-SBC)	$2.3B	Palantir Form 10-K, FY2025
Q4 2025 GAAP operating margin	41% ($575.4M)	Palantir Q4 2025 earnings release

Sources: Palantir Technologies Form 10-K, FY2025; Q4 2025 earnings release.

The gap between $1.4B GAAP and $2.3B adjusted operating income is stock-based compensation and related payroll taxes. That is the single most important reconciliation in Palantir’s model. The 82% gross margin is real, but a meaningful slice of the profitability story runs through equity issuance rather than cash cost. Reading that gap correctly is the skill in why stock-based compensation is the number to watch in tech IPOs.

The high gross margin is also what lets Palantir carry a heavy R&D and forward-deployed engineering load while still posting GAAP operating income. The same gravity, where the margin line sets the ceiling on everything below it, is the argument in why gross margin is destiny in SaaS.

AIP as the inflection point

The reacceleration of Palantir’s commercial business is an AIP story. U.S. commercial revenue grew 121% year over year to $397M in Q3 2025 and 137% in Q4 2025, with AIP adoption cited as the primary driver alongside Foundry-driven workflows (Palantir Q3 2025 earnings release; Q4 2025 earnings release).

The pipeline signal is stronger than the revenue signal. U.S. commercial remaining deal value reached $3.63B in Q3 2025, up 199% year over year and 30% quarter over quarter (Palantir Q3 2025 earnings release). Remaining deal value is contracted business not yet recognized as revenue, so a 199% jump means the commercial acceleration is booked, not just hoped for.

What makes AIP a genuine product-led motion rather than a rebrand is the customer count. Palantir served 954 customers at year-end 2025, with customer count growing 43% to 45% year over year through Q2 and Q3 (Palantir Form 10-K, FY2025; quarterly releases). New logos plus expansion inside existing accounts is the classic land-and-expand shape.

The distinction between whether AIP is a durable platform or a sales-acceleration wrapper is the crux of the bear case below. On the filings, the forward booking (RDV) and the customer growth both point to real adoption, not a one-quarter spike.

Unit economics and stickiness

Palantir’s customer economics show two things at once: extreme retention at the top and rapid expansion within accounts. The company’s top three customers have been with it for an average of ten years, and yet they account for only 16% of FY2025 revenue, down from 17% in FY2024 (Palantir Form 10-K, FY2025).

That combination is the healthy version of concentration. Long tenure signals the platform is embedded and hard to rip out. A declining top-three share signals the base is broadening, so a single customer loss does less damage than it would have a year earlier. The framework for reading that risk straight from a filing is in customer concentration risk in SaaS filings.

The expansion is where the model compounds. Average revenue for Palantir’s top 20 customers in the trailing twelve months ended December 31, 2025 was $93.9M, up from $64.6M in 2024, a 45% year-over-year increase (Palantir Form 10-K, FY2025). Growth from selling more to accounts you already have is the cheapest growth there is, because the acquisition cost was paid in a prior year.

The difference between gross and net retention decides how much of that expansion sticks after churn. The distinction, and why it matters for a company like Palantir, is the whole subject of gross retention versus net retention in SaaS IPOs.

Rule-of-40 and the path to profitability

Palantir reported a Rule-of-40 score of 127% in Q4 2025 and 114% in Q3 2025 (Palantir Q4 2025 earnings release; Q3 2025 earnings release). The Rule of 40 adds a company’s revenue growth rate to its operating margin, and anything above 40 is considered strong. A 127% reading is exceptional.

The number is high because both inputs are high at once. Q4 growth was around 70% year over year and the Q4 GAAP operating margin was 41% (Palantir Q4 2025 earnings release). Most software companies buy growth with margin or margin with growth. Palantir posted both in the same quarter, which is what breaks the usual tradeoff.

Methodology: how to read Palantir’s Rule-of-40

• Inputs: revenue growth rate and operating margin from the same period. For Q4 2025, approximately 70% growth plus 41% GAAP operating margin equals about 111% on a GAAP basis; Palantir’s reported 127% uses its adjusted operating margin, which excludes stock-based compensation.
• Assumptions: that the growth rate is sustainable and that the margin is not being flattered by the metric choice. GAAP versus adjusted materially changes the score, so the 127% headline should always be read alongside the GAAP figures.
• Sensitivity: if growth decelerates toward guided levels, the score falls fast, because growth is the larger of the two inputs. A drop from 70% to 40% growth cuts the score by 30 points on its own.
• What this misses: the Rule of 40 says nothing about durability, dilution, or valuation. It measures the quality of a single period’s growth-plus-margin combination, not the price you pay for it. A full read of what the metric does and does not capture is in Rule of 40: what it actually tells you.

The Rule of 40 is a quality-of-growth signal, and Palantir’s is among the strongest in public software. It is also the number most exposed to deceleration, which is exactly the risk the guidance implies.

The valuation question and international headwinds

Palantir guided to about 61% year-over-year revenue growth for FY2026, including roughly 115% growth in U.S. commercial (Palantir Q4 2025 earnings release). Guiding to another year of commercial near-doubling is the company’s answer to the valuation question: the growth is expected to continue, not roll over.

The valuation debate itself sits outside the filings and outside this analysis. What the filings do let you evaluate is whether the business supports the growth narrative, and here the international line is the honest weak spot. International revenue was only about 23% to 26% of FY2025 revenue, and commercial international revenue actually declined 3% in Q2 2025 (Palantir Form 10-K, FY2025; Q2 2025 earnings release).

That matters because the AIP story is, so far, a U.S. story. U.S. revenue crossed a $1B quarterly run rate while international commercial went backward in at least one quarter. If AIP is a genuine platform, it should eventually travel. If it is a U.S.-specific procurement and enterprise phenomenon, the ceiling is lower than the guidance implies.

How much of the multiple rests on international expansion that has not yet materialized is the real open question, and it is not one the FY2025 filings resolve.

The bear case: what the skeptics get right

The strongest objection is not that Palantir’s numbers are wrong. They are audited and strong. The bear case is that the growth narrative rests on assumptions the filings do not fully confirm.

AIP could be sales acceleration, not a durable product. The 137% U.S. commercial growth and the 199% RDV jump are real, but a skeptic reads them as a one-time land grab funded by AI hype, not a platform with durable retention (Palantir Q3 and Q4 2025 earnings releases). The counter is that ten-year top-customer tenure and 954 customers point to embedding, not churn. But AIP specifically is young, and its net retention over multiple years is not yet in the filings. The bear is right that the durability is asserted, not proven.

Customer concentration is lower but government concentration is not. The top three customers are only 16% of revenue, which looks healthy (Palantir Form 10-K, FY2025). But a large slice of the base is U.S. government, which concentrates a different risk: procurement cycles, budget politics, and program cancellations that no software quality can offset. A single agency posture change can move the government line in ways a commercial customer loss cannot.

The growth guide is the fragile part. FY2026 guidance of about 61% total and 115% U.S. commercial growth is what supports the valuation (Palantir Q4 2025 earnings release). Because the Rule-of-40 score leans heavily on growth, any deceleration compresses the quality signal fast. The bear says the company has guided itself into a corner where merely strong results look like a miss.

International is a real hole. Commercial international revenue declining 3% in a quarter while U.S. commercial nearly doubled is the single most awkward line in the year (Palantir Q2 2025 earnings release). If AIP does not travel, the addressable market is smaller than the narrative assumes.

Weighing it: the bear case does not break the model, it bounds it. The government base is durable, the commercial acceleration is booked into RDV, and the margins are genuine. But the durability of AIP, the government-concentration risk, the guidance dependency, and the international weakness are all real. The thesis survives as “a durable base funding a fast but unproven-abroad commercial bet,” not as “the growth is guaranteed.”

What operators should take from this

If you build software, the transferable lesson is not “sell to the government.” It is the sequencing underneath Palantir’s model.

Palantir is demonstrating, at scale, that a durable, high-margin base can fund an aggressive growth bet without external capital, if the base is sticky enough to be predictable. The government revenue is not the exciting part. It is the part that makes the exciting part fundable. Here is the playbook.

• Separate your funding engine from your growth engine, on paper. Label each revenue line as either “funds the bet” (durable, high-retention, predictable) or “is the bet” (fast, expanding, less proven). Palantir runs government as the first and AIP as the second. Confusing them leads founders to starve a reliable base to chase a growth line, or to under-invest in the growth line because the base looks safer.
• Read the GAAP-to-adjusted gap before you trust the margin. Palantir’s $1.4B GAAP versus $2.3B adjusted operating income is stock-based compensation (Palantir Form 10-K, FY2025). When you see an “adjusted” profitability headline, find the reconciliation and decide whether the excluded cost is real to you. Equity is a cost even when it is not cash.
• Price expansion revenue as your cheapest growth. Top-20 average revenue rose 45% to $93.9M inside existing accounts (Palantir Form 10-K, FY2025). Expansion carries no new acquisition cost, so instrument net revenue retention as carefully as new-logo growth. A land-and-expand motion lives or dies on the expand.
• Treat long tenure as the moat, not the contract size. The ten-year average tenure of the top three customers is the real signal, more than their revenue share (Palantir Form 10-K, FY2025). Build for embeddedness, the switching cost that comes from being wired into a customer’s workflows, because that is what turns a customer into an annuity.
• Watch the metric choice, not just the metric. A 127% Rule-of-40 score uses adjusted margin; the GAAP version is lower. When you report a composite metric, disclose which inputs you used, because the sophisticated reader will reconstruct it and trust you more for showing your work.
• Stress-test the geography of your growth. Palantir’s growth is a U.S. story with an international question mark. Before you extrapolate a growth rate, ask whether it holds outside the market where it started. A domestic phenomenon priced as a global one is the most common overreach.

The through-line is prioritization: the durable base funds the speculative bet, and confusing which line is which is the mistake that starves one to feed the other.

Where this analysis is vulnerable

A credible read names its own holes. There are three.

The segment splits are approximate. Palantir reports growth rates and some absolute segment figures, but the precise U.S.-versus-international and government-versus-commercial dollar splits require combining several disclosures. The Palantir Revenue Split above is a faithful reading of the filings, not a line lifted verbatim from a single table, and the approximate labels reflect that.

AIP’s contribution is attributed, not itemized. Palantir credits AIP as the primary driver of commercial acceleration, but the filings do not break out AIP revenue as a separate line from Foundry (Palantir Q3 2025 earnings release). The AIP-as-inflection claim is a reasonable read of management commentary plus the commercial growth rate, not a disclosed product-level figure.

The forward numbers are guidance, not results. The FY2026 growth guide is management’s expectation, and any analysis leaning on it inherits that uncertainty (Palantir Q4 2025 earnings release). Guidance is a claim about the future; only the FY2025 actuals are audited fact.

None of these is fatal on today’s evidence. All three are why this is a strong reading of a still-accelerating business, not a settled verdict.

How the pieces fit together

Palantir’s model is a stack of reinforcing parts, read straight from the filings:

1. A durable U.S. government base (about $1.9B, +53%) provides predictable, high-tenure revenue that funds the rest (Palantir Form 10-K, FY2025).
2. A fast U.S. commercial engine (about $1.5B, +109%, Q4 +137%) driven by AIP supplies the growth narrative (Q4 2025 earnings release).
3. An 82% gross margin converts that revenue into $1.4B GAAP operating income, with the adjusted-versus-GAAP gap running through stock-based compensation (Palantir Form 10-K, FY2025).
4. Sticky, expanding accounts (ten-year top-three tenure, +45% top-20 revenue) compound the base cheaply (Palantir Form 10-K, FY2025).
5. A 127% Rule-of-40 score marks the quality of the combination, while the international weakness marks its limit (Q4 2025 earnings release; Q2 2025 earnings release).

The companies calling Palantir “just a government contractor” are reading only the floor. The companies calling it “just an AI story” are reading only the ceiling. The business is the base funding the bet, and the FY2025 filings show both working at once, with international as the unresolved test.

---

Analysis, not investment advice. Figures are drawn from Palantir Technologies Inc.’s public SEC filings (Form 10-K, FY2025, and Q1 to Q4 2025 earnings releases) and cited inline by fiscal period. Frameworks here, including the Palantir Revenue Split, are for understanding business strategy and tradeoffs, not for making buy or sell decisions.

Want the full toolkit for reading filings like this, the segment-split worksheet, the Rule-of-40 reconciliation model, and the Palantir Revenue Split template used above? It’s in the Tech Business Analysis Playbook.