Alex Karp (Alexander Caedmon Karp), Co-founder and Chief Executive Officer of Palantir Technologies

Palantir's third-quarter figures arrived this week to the kind of applause that sends analysts hunting for fresh superlatives. Revenue passed $2 billion in a quarter for the first time, up 96 per cent on a year earlier. Net dollar retention reached 165 per cent, free cash flow ran at a margin of 53 per cent, and the company reported $10.4 billion in net cash with no long-term debt. Shares edged higher in extended trading. One research house nonetheless cut its rating from buy to hold, on the sole ground that the price already assumes near-perfect execution.

Across the Atlantic, a different set of numbers was in circulation. A newspaper investigation published on Monday found that 29 NHS trusts listed as benefiting from the Federated Data Platform told reporters they did not use it, or that it was not yet live in their organisation. Palantir responded that trusts using its software had recorded 117,000 additional operations. One set of figures is produced under the disclosure rules of a listed company. The other is assembled by a health service with every incentive to show that a flagship national programme is working. That asymmetry is the story.

The relevance to Britain is direct. The £330 million contract, awarded in 2023, makes Palantir the supplier of the platform that ministers describe as a key enabler of the 10 Year Health Plan. NHS England reports 139 trusts live, 170 signed up and 137 reporting benefits. The platform is now being extended to AI tools for triage and discharge planning, and to the sharing of GP data. A programme of that reach cannot be treated as an ordinary software purchase. Its success or failure will shape how the NHS recovers waiting list performance, and how much faith the public places in the next digital reform.

Palantir's results also illuminate the commercial model behind the contract. Net dollar retention of 165 per cent means existing customers are spending far more each year than they did the year before. The pattern is land and expand: begin with a narrow use, embed the product in daily operations, then widen the scope. It has served Palantir well in the commercial market, where revenue grew 113 per cent. It is visible in the NHS too, where a platform that began with waiting list management is moving towards clinical decision support. For a supplier this is sound strategy. For a public buyer, each extension raises the cost of leaving.

The bargaining position deserves plain acknowledgement. Worth £330 million over seven years, the NHS contract is a small line in the accounts of a company that now books more than $2 billion a quarter. Palantir does not need the NHS in the way the NHS has come to depend on Palantir. A customer in that position has to compensate with rigour, because it cannot compensate with weight.

Rigour is where the evidence looks thin. Freedom of information requests found that 13 of 41 trusts using one inpatient scheduling module performed fewer procedures than before it went live. NHS officials reply that benefits accrue only where clinicians build their own use cases, and that this takes years. Both positions are plausible and neither is settled, because activity in a trust moves with industrial action, winter pressure and staffing as much as with software. The missing ingredient is a published baseline and a comparison with trusts that did not adopt. Without those, the NHS can describe what happened after go-live but cannot say what caused it.

Adoption figures carry a further problem. Trusts were directed to submit adoption plans, and NHS England commissioned a consultancy at a cost of £8.5 million to promote uptake. Leeds judged it would lose functionality by moving from its existing systems. Greater Manchester built its own data infrastructure and has described the national platform as potentially retrograde. A count of signatories measures compliance with an instruction. It says little about whether clinicians and managers find the product useful enough to choose it unprompted.

Politics compounds the difficulty. Palantir's work in defence and intelligence gives the company a reputation that many patients and campaigners find uncomfortable, and parliamentary pressure to end the contract has grown. London's mayor blocked a £50 million police contract with the firm earlier this year. Public consent for data use is the scarce resource in any digital reform, and a platform that is mandatory in practice yet contested in principle spends that resource quickly. Every disputed benefits claim makes the next conversation with the public harder.

Several practical implications follow. NHS leaders should ask for usage data, since sign-up data tells them little, and should insist on clear exit and data portability terms before any further AI module is added. Policymakers should commission an independent evaluation with trust-level access, completed before the next round of extensions is agreed. British health-tech and life sciences companies should note that regional work such as Greater Manchester's shows a credible alternative exists, and that open standards and procurement frameworks that reward interoperability are their best route into the market. Patients are entitled to indifference about who writes the software and to firm expectations about who can see their records.

The analyst's advice to wait for a better entry point is a market judgement, and the market has already priced in Palantir's future. The NHS has priced in nothing, because it has not yet measured what it is buying. It is already inside the contract, so the equivalent discipline is to establish a proper exit point, to know what each extension costs, and to be able to say, with its own evidence, whether the platform is worth keeping. Until it can, the supplier's good quarter tells the public very little about the health service's.

‍