1,049 customers, 42% of revenue from the US government: how a customer is really counted
7 min read
On 4 August 2026 Palantir Technologies filed its 10-Q with the SEC for the quarter ended 30 June 2026 — a mandatory quarterly filing, not a set of accounts and not a press release: figures the company would have to defend in court. The same document contains two statements that are both true: that in the period Palantir served 1,049 customers, and that a huge slice of the quarter’s revenue came from a single counterparty, the government of the United States. This is not an irregularity: the convention behind the 1,049 figure is spelt out in full in the same filing, and it is defensible. But it radically changes what that number tells anyone reading it as a measure of a supplier’s solidity — the reading a procurement office ought to make before signing, not after.
Two true figures, in the same filing
The quarter’s revenue splits into two segments: government $990.0 million ($553.0 million a year earlier), commercial $945.4 million ($450.7 million). Contribution — the margin Palantir itself uses to measure each segment’s efficiency — rose to 71% on government (was 63%) and 78% on commercial (was 64%); overall, 74% against 64%.
Within the government segment, the filing isolates one line: “Revenue from U.S. government customers was $809 million for the three months ended June 30, 2026 compared to $426 million for the same period in 2025.” Total quarterly revenue, adding the two segments together, is $1,935.5 million. We do the division ourselves: 809 over 1,935.5 comes to 41.8% — nearly 42 cents of every dollar the quarter brought in, from one counterparty. And 809 over 990.0, the government segment alone, comes to 81.7%. Our own calculations on the filing’s figures, not percentages the company states.
How you get to 1,049
The filing spells out the counting convention in full: “During the period ended June 30, 2026 and 2025, we had 1,049 and 849 customers, respectively, including companies in various commercial sectors and government agencies around the world. For large government agencies, where a single institution has multiple divisions, units, or subsidiary agencies, each such division, unit, or subsidiary agency that enters into a separate contract with us and is invoiced as a separate entity is treated as a separate customer. For example, while the U.S. Food and Drug Administration, Centers for Disease Control and Prevention, and National Institutes of Health are subsidiary agencies of the U.S. Department of Health and Human Services, we treat each of those agencies as a separate customer given that the governing structures and procurement processes of each agency are independent.”
A sensible rule, not an accounting trick: the FDA, CDC and NIH really do buy independently. But anyone reading “1,049 customers” as shorthand for “diversified supplier” is using a number built to count invoices as though it measured something else: how many independent decisions it would take to hurt the supplier.
The right question isn’t how many customers
Two suppliers that both report “a thousand customers” can have opposite dependency structures: one spread across a thousand different boards, the other across 400 offices of the same apparatus that would all stop buying together if a budget or a political stance changed. Counting by agency doesn’t lie: it faithfully describes how the buyer is organised. It does not describe the supplier’s diversification — that is not the same question.
The question that matters, in a specification or in due diligence on critical suppliers, is: how many independent decisions would it take to strip this supplier of a significant share of its revenue? If the answer is “just one — a budget, a change of majority, a shift in executive policy” — the number of customers on the invoice does not reduce that risk by a single percentage point, however high it is.
Where it’s written down, and where it isn’t
The risk factor is disclosed by Palantir itself, not hidden: “existing customers have expanded their relationships with us, which has resulted in a limited number of customers accounting for a substantial portion of our revenue. If existing customers do not make subsequent purchases from us or renew their contracts with us, or if our relationships with our largest customers are impaired or terminated, our revenue could decline, and our results of operations would be adversely impacted.” A textbook line from a supplier-risk manual, written by the supplier itself — and it sits in the 10-Q, not in the quarterly press release with the growth rates. The difference is not hidden content: it is a matter of who opens the right document — the same rule seen today about a memorandum disclosed only in a photo caption: the announcement promises, the primary document commits.
The same filing measures the weight of large customers another way too: average revenue for the top twenty, in the twelve months ended 30 June 2026, was $124 million, up 67% from $75 million a year earlier. A sense of scale, without building a case on it: 20 customers at $124 million comes to $2.48 billion over twelve months — assuming, on our own assumption, an even spread across the four quarters, that is over a third of the group’s actual six-month revenue.
It applies as a mirror to you, too
There’s a point almost no procurement office sees, and it is symmetrical to everything above. If you are an agency, a division or a public body that buys separately under its own specification, you are a distinct customer in the supplier’s books — just like the FDA, CDC and NIH. You are not distinct in bargaining power: your leverage is that of your own single contract line, not that of the larger body you belong to on paper. If the supplier renegotiates terms with the central body, your line moves with it, and you neither decided that nor could negotiate as a bloc. It applies to a company with several divisions as much as to an administration with several departments: the separate count is real on the invoice, not in bargaining power.
The other side, owed in fairness
Getting US government customers to bring in $809 million in a quarter — nearly double the $426 million of a year earlier — to the point of accounting for 82% of the government segment, is the result any supplier would chase: growth concentrated in the most solvent customer on the planet, the same strength that makes Palantir, through FedStart, a trust infrastructure for other suppliers reaching the US government too. Concentration is not a fault: it is a characteristic that needs to be known, the way a company that depends on a single customer for 80% of revenue rightly discloses it as a risk in its own accounts. Palantir discloses it, without burying it in unreadable small print — public information, not an accusation of ours.
The operational lesson
- Don’t take “number of customers” as a measure of diversification without reading the methodology note: ask the supplier how it counts a customer.
- Open the filing, not just the press release: 10-Ks and 10-Qs are free and public on EDGAR; the risk factors a company discloses about itself are worth more than any guesswork of ours.
- If you buy separately from a larger body, don’t confuse contractual autonomy with bargaining power: check whether other divisions depend on the same supplier, and coordinate before an unfavourable renewal.
- Keep a register of critical dependencies updated at every filing and every announcement, not a one-off assessment done at signature: concentration shifts quarter to quarter, as these very figures show.
It is the same logic that in Europe the financial regulator applies to banks over their own critical ICT suppliers: a register — who your critical suppliers are, how much they weigh on you, how much you weigh on them, what they disclose in their own filings and how that changes over time — is the control we build on the customer’s contracts and systems, with the audit trail ready for an inspector, not a one-off due-diligence exercise filed away. It is the same set-up that ties together an organisation’s scattered data — contracts, suppliers, archives, management systems, documents — into a single operating model on which AI agents execute decisions with a human operator in command, for large enterprises, defence, public administration and healthcare: compliance is the way in, the decision-making system is what matters. On-premise on autonomous machines that do not require deep integration into your network, or a dedicated cloud with a data centre in Italy and premises staffed directly by us — always jointly managed: a first register of critical suppliers goes live within weeks.
Want to know how much your most critical supplier weighs on you — and how much you weigh on them? Half an hour to map it out together.