Palantir sells software that helps large organizations make sense of complicated data. It has four main platforms: Gotham, which helps military and intelligence agencies see what is happening in real time; Foundry, which acts as a central data operating system for hospitals, manufacturers, and governments; Apollo, which keeps all the software updated and running in any environment; and AIP, an artificial intelligence platform launched in 2023 that connects advanced AI language models to a customer's own data and workflows. Customers pay for access to these platforms through fixed-term contracts, typically one to five years long, and revenue is recognized steadily across the contract period. In 2025, 54% of revenue came from government customers and 46% from commercial customers, spread across 954 customers worldwide. The diagram below traces where the money goes.

Five years of financial data tell a clear story about direction. Revenue has climbed every single year, from $1.5 billion in 2021 to $4.5 billion in 2025. That is not slow, steady growth. Revenue jumped 56% in 2025 alone. What makes this more interesting is that the business is becoming more efficient as it grows, not less.

Gross margin, which measures how much revenue is left after the direct cost of delivering the software, has held above 78% every year and reached 82% in 2025. That level of margin is a sign that delivering more software to more customers does not require spending proportionally more money to do it. More telling is what happened to cash generation. In 2021 and 2022, the business produced about $200 to $300 million in free cash flow. By 2025, that number had grown to $2.1 billion. The business went from covering its bills to generating serious cash.

The balance sheet is also in a strong position. Palantir carries more cash than debt, reflected in a negative net debt figure across all five years. The company is not borrowing to survive. It is sitting on a cash cushion while the business accelerates. The pipeline of future revenue also looks substantial. As of December 31, 2025, total remaining deal value, meaning the value of contracts already signed but not yet recognized as revenue, stood at $11.2 billion. That figure doubled in a single year.

The risks are real and specific, not generic. Government contracts make up 54% of revenue, and those contracts come with unusual uncertainty. The U.S. federal government cannot commit to exercising contract options more than one year in advance. Budget delays, shifting spending priorities, and policy changes can slow or shrink revenue from government agencies with little warning. The top three customers alone accounted for 16% of revenue in 2025, so losing even one of them would leave a visible gap.

Sales cycles add another layer of unpredictability. It typically takes six to nine months or longer to close a deal. Palantir also runs pilot programs and AIP bootcamps at its own expense, with no guarantee that the customer will sign a paying contract afterward. Large deals often close in the final weeks of a quarter, which makes quarterly revenue hard to predict. A deal that slips by a few weeks can make a quarter look much worse than it actually is. Finally, Palantir's work with military agencies and sensitive government programs draws repeated criticism and public activism. That scrutiny can affect its reputation with commercial customers and with potential employees.

The AI platform, AIP, is the most important new variable in the story. Palantir is spending heavily to build it out and to train customers on it through bootcamps and conferences. U.S. commercial revenue grew 109% in 2025, and much of that acceleration is tied to AIP adoption. But fast growth in a new product does not automatically mean the product is winning the market permanently. Competitors, including large enterprise software companies with much bigger sales forces, are also building AI tools. Whether AIP holds its ground in a crowded field is not yet settled.