For most of its life, Palantir Technologies’ most dependable customer has been the government that helped create it. On Monday evening the company said, for the first time in a quarter, that this is no longer where most of its American revenue comes from.
Revenue in the third quarter reached $2.31 billion, ahead of the $2.18 billion analysts expected and close to double the figure a year earlier, according to the company’s statement. It was the first quarter above $2 billion. Net income was $1.13 billion, or $0.44 a share, and adjusted earnings per share of $0.44 compared with $0.41 expected.
The mix is what changed. Revenue from US commercial customers rose 148% to $984 million, exceeding the $953 million booked from American government agencies, which grew 96%. The crossover had been expected; the speed of it had not.
Palantir raised its forecast for the full year to about $8.53 billion and said it expects more than $3.56 billion from its American commercial business, up from its previous guidance. For the fourth quarter it guided to about $2.65 billion in revenue.
The stock fell 8% in premarket trading the next morning, which is the part of the release worth reading twice. A company that beat on revenue, beat on adjusted earnings and raised its full-year outlook does not normally lose 8% before the opening bell unless the number it reported was already in the price.
Palantir was built to sell software to intelligence agencies and militaries, work that is slow to win, hard to lose and paid in annual increments appropriated by legislatures. Its early products organized intelligence data for analysts; later ones tracked supply chains and factory output for manufacturers. The business was profitable late and polarizing throughout, and for years the argument against it was that government contracts would not scale into a software company’s margins.
The argument in favor was always the same: the software built for the hardest customers would be worth more to everyone else. Management spent the past two years testing that claim by selling access to its artificial intelligence platform through short, intensive workshops rather than through the year-long pilots that enterprise software usually requires, and the method compressed the time between a first meeting and a signed contract.
A commercial segment growing at 148% changes the character of the revenue. Government work arrives in appropriations cycles and survives changes of administration; enterprise work arrives monthly and can be cancelled with a quarter’s notice. The first is more predictable, the second compounds faster, and a company that has both is less exposed to any single budget. Whether the commercial growth is durable is the question the next four quarters will answer, because enterprise software contracts are easiest to sign in the first year and hardest to renew in the second.
The government business has not stalled, which matters for the bear case. American government revenue grew 96%, a rate no one would describe as a legacy franchise. The commercial side did not overtake it because the public side shrank; it overtook it because the private side grew faster from a smaller base.
The company’s reputation travels with the numbers. Its early contracts with intelligence and defense agencies drew protests, its name became shorthand in debates about surveillance, and its chief executive has been willing to argue in public with customers and critics alike. That history is a liability in some procurement offices and an asset in others, and it has shaped which enterprises take the first meeting.
Skeptics make three points. The valuation assumes years of growth at rates that few software companies have sustained at this size. A large share of revenue sits with a small number of customers, and the loss of one would be visible in a single quarter’s results. And the platform competes with tools that cloud providers give away as part of a larger contract, which caps what Palantir can charge for the software layer alone.
The bull case rests on the width of the deployment. A customer that starts with one use case and expands to ten is worth more each year without a new sale, and Palantir’s sales motion is designed to find the second use case before the first contract renews. That is also the motion most exposed to a slowdown in enterprise technology budgets, since expansion spending is the first line item to be deferred when a chief financial officer gets nervous.
What the quarter establishes is narrower than a verdict. The direction of the crossover is now measurable, the full-year outlook has been raised once more, and the market’s reaction suggests the bar has moved above the numbers. For a company that spent twenty years arguing it was not a consulting firm with a government contract, the proof is arriving in the segment that pays monthly, and the test is whether those customers are still there when the novelty of the first deployment wears off, still paying for the second year of a contract that no longer has a novelty to sell.
