Report contents
At a glance
- Business model: Palantir sells hosted and customer-controlled software subscriptions, plus implementation and operations services, mainly to U.S. government and allied agencies and to large commercial customers. The Ontology links data, decisions and actions, making Palantir hard to remove once workflows are encoded. Expansion depends on converting pilots into production use and on founder-led sales becoming repeatable; commercial AIP is the swing factor.
- Financial quality: Palantir converted $2.31 billion of September 2026 revenue into $1.21 billion of operating cash flow and $1.20 billion of free cash flow, so earnings are cash-backed. It holds $10.4 billion of cash and Treasury securities against $2.07 billion of liabilities and no long-term debt. Cost of revenue is growing faster than revenue, below-the-line items help, and dilution trims per-share cash flow.
Business model
Business model: expansion economics with government ballast
Palantir sells software subscriptions and the operating support that makes them useful inside large institutions. Its customers are government agencies, chiefly U.S. defense and intelligence, and commercial enterprises with complex, regulated operations. The company hosts some environments and installs others in customer-controlled infrastructure, then sells implementation and operations services around both. Revenue is mostly recognized ratably over contract terms, so the model rewards installed-base expansion more than one-off license sales. This is a business-to-government and business-to-business software platform in an expansion stage, not a consumer brand or a foundation-model developer.
Revenue mix: subscriptions set the pace, services open doors
Palantir’s 2025 Form 10-K describes three contractual forms: hosted subscriptions, customer-controlled software subscriptions with ongoing operations and maintenance, and professional services. The core platforms are Gotham, Foundry, Apollo and AIP (Artificial Intelligence Platform), with the Ontology (a governed map of an organization’s data, decisions and actions) as the common layer. The company does not publish a complete revenue split across those products, hosted versus on-premises delivery, or subscriptions versus services. That gap matters. Investors can see total growth but cannot tell how much comes from AIP, how much from services attached to deployments, or how much of the installed base remains on older contracts.
The three forms carry different economics. Hosted subscriptions can scale with limited incremental delivery cost. Customer-controlled deployments require ongoing operations and maintenance, which supports retention but consumes engineering and support labor. Professional services are project-based and can be lumpy. Because revenue is reported in aggregate, a mix shift toward services or on-premises delivery would be hard to see before it affects margins.
Ratable recognition creates a second gap between activity and reported revenue. A large new deployment may take quarters to show up, and a slowdown in bookings may not appear immediately. The economic engine is expansion within existing accounts: land on a high-value operational problem, connect data and decision workflows through the Ontology, then widen use across functions and missions. Services help embed the software and can be strategically valuable, but they are likely lower-margin than hosted subscriptions. Palantir does not break out margin by contract form, so the size of that drag is not public. The inference is that services serve customer acquisition and retention as much as they generate revenue.
The moat is workflow lock-in, not model ownership
Customers do not buy a model from Palantir. They buy a governed environment in which data, permissions, analytical models, human decisions and real-world actions are linked. Once an agency or company encodes its operating logic in the Ontology, replacing Palantir means rebuilding workflows, audit trails, security reviews and user training. That is the strongest source of pricing power, and it explains why the company can sell multiyear subscriptions with expansion options. The 2025 Form 10-K nonetheless flags contract optionality and termination rights, so the installed base is stickier in daily use than in legal commitment.
The Ontology also creates a data-model standard inside the customer. As more departments build on it, the cost of removal rises with each new use case. That makes expansion more important than initial win rates. Standardization can cut both ways, however: if a customer’s cloud provider or model vendor offers a credible substitute for the orchestration layer, switching costs fall at the next architectural refresh.
The sales motion reinforces the same tradeoff. Founder-led, high-touch selling has won complex initial deployments, but it is difficult to industrialize. Palantir has warned that sales hires, reorganizations and compensation changes can temporarily reduce productivity. Expansion therefore depends on converting pilots into production use and on account teams that can sell new modules without the founder in the room. A pilot that stays a pilot generates services revenue without the software leverage that makes the model attractive.
Government is ballast; commercial AIP is the swing factor
The geographic market is the United States and allied countries. Government work provides durability because missions are long-lived and switching costs are high, but procurement cycles, recompetes and termination rights make it less annuity-like than recurring revenue suggests. Allied sovereign programs are an extension of the same model, not a separate consumer business; they add procurement and security-review cycles that can delay revenue. Commercial expansion is the swing factor. It can dilute customer concentration and widen the addressable market, but it also pushes Palantir toward buyers with less mission urgency and more alternatives, including large software vendors and internal development teams.
AIP, deployed in 2023, is the main commercial vector. It lets Palantir sit above foundation models: customers can change models without replacing the orchestration layer. If model vendors bundle that layer into their own suites, Palantir must defend price with services and customization, which weakens software economics. The judgment would weaken if commercial customer additions slow while revenue per customer falls, if public awards shrink at recompete, or if filings begin disclosing early terminations or a rising top-customer share. Sustained growth in remaining deal value (contracted work not yet recognized) and in production deployments would show the operating layer is spreading beyond initial pilots.
Questions worth following
- How exposed is Palantir to early termination of its largest contracts?
The 2025 Form 10-K flags contract optionality and termination rights as risks. If agencies or large enterprises can cut scope before renewal, reported backlog and ratable revenue overstate durability, and the installed-base expansion that supports the growth case becomes less predictable.
Financial quality
Financial quality: real cash, rising cost of revenue
Palantir's earnings are now backed by cash. The company reported $2.313 billion of revenue in the September 2026 quarter, up 96% year over year, $1.085 billion of GAAP operating income (47% margin) and $1.205 billion of operating cash flow (52%), alongside $1.216 billion of adjusted free cash flow. Cash, equivalents and U.S. Treasury securities came to $10.4 billion against $2.07 billion of total liabilities and no long-term debt. The balance sheet is the least ambiguous part of the story. What is less settled is what holds the margin: a cost line now growing faster than revenue, and a bottom line that leans on items below the operating line. Financial quality: Strong, not exceptional.
| USD bn unless noted | FY2021 | FY2022 | FY2023 | FY2024 | FY2025 | Q3 FY2026 | Assessment |
|---|---|---|---|---|---|---|---|
| Revenue | 1.54 | 1.91 | 2.23 | 2.87 | 4.47 | 2.31 | Strong |
| Revenue growth y/y | n/a | 24% | 17% | 29% | 56% | 96% | Exceptional |
| Gross margin | 78.0% | 78.6% | 80.6% | 80.2% | 82.4% | 83.4% | Strong |
| Operating margin | -27% | -8% | 5% | 11% | 31% | 47% | Exceptional |
| Net margin | -34% | -20% | 9% | 16% | 36% | 49% | Exceptional |
| ROE | -27.3% | -15.1% | 6.8% | 10.8% | 26.2% | 10.7% | Strong |
| Liabilities/assets | 30% | 24% | 21% | 20% | 16% | 15% | Strong |
| Current ratio (x) | 4.3 | 5.2 | 5.5 | 6.0 | 7.1 | 7.1 | Strong |
| Receivables turnover (x) | 8.9 | 8.5 | 7.1 | 6.1 | 5.5 | 1.4 | Adequate |
| Operating cash flow | 0.33 | 0.22 | 0.71 | 1.15 | 2.13 | 1.21 | Exceptional |
| Capex (derived) | 0.01 | 0.04 | 0.01 | 0.01 | 0.03 | 0.01 | Exceptional |
| Free cash flow | 0.32 | 0.18 | 0.70 | 1.14 | 2.10 | 1.20 | Exceptional |
| Operating cash flow/net income | n/m | n/m | 3.4 | 2.5 | 1.31 | 1.07 | Adequate |
*Quarter ended September 30, 2026. ROE and receivables turnover are single-quarter figures, not annualized.
Where the margin actually comes from
Cost of revenue, derived from the gross profit line, went from $0.21 billion in the March 2026 quarter to $0.30 billion in June and $0.38 billion in September. That is an 81% increase over two quarters against 42% revenue growth, and gross margin fell from 86.8% to 83.4% over the same span. Operating margin did not follow because operating expenses excluding cost of revenue grew about 9% sequentially against 19% revenue growth. The margin engine is now scale in sales and engineering, not unit economics, and that source is finite. A 3.4 percentage-point slide in two quarters needs a cause: usage-based hosting and inference for AIP workloads would make it structural, while mix toward larger deployments would make it temporary. The company does not break out the composition of cost of revenue, so the level to watch is whether gross margin finds a floor above 81%.
The quarterly pattern shows the same tension. Adjusted operating margin held near 47% while gross margin fell for two straight quarters, then net margin dropped 6 percentage points sequentially to 48.8% even though operating margin was unchanged at 46.9%.
| Quarter | Revenue | Net income | Gross margin | Net margin | Total assets | Liabilities | Net PP&E | OCF | FCF | Assessment | Trend |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 2025 Q1 | 0.88 | 0.21 | 80.4% | 24.2% | 6.74 | 1.22 | 0.25 | 0.31 | 0.30 | Adequate | Improving |
| 2025 Q2 | 1.00 | 0.33 | 80.8% | 32.6% | 7.37 | 1.34 | 0.25 | 0.54 | 0.53 | Strong | Improving |
| 2025 Q3 | 1.18 | 0.48 | 82.4% | 40.3% | 8.11 | 1.43 | 0.25 | 0.51 | 0.50 | Strong | Improving |
| 2025 Q4 | 1.41 | 0.61 | 84.6% | 43.3% | 8.90 | 1.41 | 0.25 | 0.78 | 0.76 | Strong | Improving |
| 2026 Q1 | 1.63 | 0.87 | 86.8% | 53.3% | 10.20 | 1.64 | 0.06 | 0.90 | 0.89 | Exceptional | Improving |
| 2026 Q2 | 1.94 | 1.06 | 84.7% | 54.9% | 11.68 | 1.79 | Not disclosed | 1.22 | 1.20 | Exceptional | Stable |
| 2026 Q3 | 2.31 | 1.13 | 83.4% | 48.8% | 13.41 | 2.07 | Not disclosed | 1.21 | 1.20 | Strong | Mixed |
All values are single-quarter figures in USD bn, not year-to-date. Net PP&E is a rounding-level item for a company spending about 1% of revenue on capex.
Nearly all of that net margin decline happened below the operating line: EBIT exceeded operating income by $0.17 billion in the June quarter but only $0.07 billion in September. Tax is a second prop. The FY2025 provision was $0.02 billion on roughly $1.6 billion of pre-tax profit, a low single-digit rate that likely reflects loss carryforwards. A 20% rate would have removed about $0.3 billion from FY2025 net income with nothing changed in the business. ROE of 26.2% therefore reflects a 36% net margin and a thin tax charge, not leverage: liabilities are 16% of assets and short-term debt is $0.05 billion.
Cash conversion has converged, and dilution takes a cut
Operating cash flow was 3.4 times net income in FY2023, 2.5 times in FY2024, 1.31 times in FY2025 and about 1.1 times in the September 2026 quarter. That is what happens as loss shields and noncash addbacks stop dominating a larger earnings base. It is healthy, and it is no longer exceptional. Capex runs near $0.01 billion a quarter, so free cash flow is essentially operating cash flow and the 52% margin is not a deferral artifact. Investing cash flow swung from -$1.48 billion in June 2026 to +$0.37 billion in September, which tracks purchases and maturities of Treasury securities; the $3.62 billion cash line on the balance sheet understates liquidity by roughly $6.8 billion.
The qualification is dilution. Equity rose $2.39 billion in FY2025 against $1.62 billion of net income, and financing cash flow has been near zero (-$0.03 billion in FY2025, +$0.01 billion in the latest quarter), so nothing offsets issuance. Free cash flow per share trails free cash flow, and the company's guided $4.8 billion to $5.0 billion of FY2026 adjusted free cash flow should be read on a per-share basis.
New contract value is not keeping pace with revenue
Palantir closed $2.918 billion of total contract value in the September quarter, up 6%, and $1.463 billion of U.S. commercial contract value, up 11%, against revenue growth of 96% and U.S. commercial growth of 148%. Revenue is being drawn from backlog faster than it is being replaced: U.S. commercial remaining deal value of $6.517 billion, up 79%, shows the backlog is still deep, so this reads as faster conversion of signed work, not a demand break. It does mean the 2027 revenue path needs signings to reaccelerate, and the company's guidance of $8.53 billion for FY2026 rests on obligations signed earlier.
Collections are not the problem. Receivables of $1.88 billion are up 81% since December 2025 while revenue nearly doubled year over year, and turnover has held near 5.5 times, about 66 days, for five quarters after sliding from 8.9 times in FY2021. A current ratio of 7.1 times reflects the Treasury portfolio more than working-capital strength, but it removes refinancing from the list of concerns.
Three checks would settle the quality question. Whether cost of revenue keeps outgrowing revenue for another two quarters, which would point to usage-based hosting costs instead of one-off mix. Whether total contract value growth returns to double digits, without which 91% revenue growth becomes a peak year. And whether the tax charge and non-operating income normalize, since both flatter the current net margin without any operating achievement behind them.
Questions worth following
- Can Palantir hold gross margin above 80% as AIP workloads scale?
Cost of revenue rose from $0.21 billion to $0.38 billion over two quarters while revenue grew 42%, and gross margin fell 3.4 percentage points to 83.4%. If inference and hosting costs scale with usage, the 47% operating margin must come entirely from expense leverage that cannot keep outpacing revenue.
- Will Palantir keep converting receivables into cash as its mix shifts toward larger government contracts?
Receivables grew 58% in FY2024 against 29% revenue growth, receivables turnover fell from 7.4x to 6.1x, and the June 2025 quarterly rate annualizes near 5.4x. Slower collections would tie up cash even while reported earnings keep rising.
- Are Palantir's 50%-plus net margins durable, or do interest income and a near-zero tax rate carry too much of them?
Net margin reached 54.6% in the June 2026 quarter while EBIT exceeded operating income by about $0.17bn, and FY2025's tax provision was only $0.02bn. If that support fades, earnings power is below what the reported margin implies.
Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.
Continue reading the full report
- Key risksWhat could break the investment case
- Valuation and margin of safetyWhat the current price assumes, and how much room for error it leaves
- SummaryOur overall judgement and what to watch next
Free to sign up · 20 research points · unlocking this report uses 10
20 points cover 2 full reports. No card needed.
Research reflects its generation date. Preview the first two modules.
More on Palantir Technologies Inc.
Palantir Technologies Inc.
Business, moat and key questions
Open company page