You Cannot Govern Through Purchase Orders

In February 2025, the Prime Minister of the United Kingdom flew to Washington to meet the President of the United States. It was a significant visit. Starmer needed to reassure Trump on defence spending and to signal closer technical cooperation. The meeting at the White House went well.

The Prime Minister's next stop was not the State Department, nor the Pentagon, nor a meeting with congressional leaders. It was an office building on Thomas Jefferson Street NW, ten minutes from the Oval Office, housing the Washington headquarters of Palantir Technologies. He was accompanied by Peter Mandelson, then the British Ambassador, whose lobbying firm Global Counsel listed Palantir as a client. They toured the facilities, attended a presentation, and met the company's chief executive, Alex Karp.

No minutes were taken. No transcript was produced. The Cabinet Office later described it as an informal visit. Freedom of information requests returned nothing. The Foreign Office said it held no emails about Mandelson's role in arranging the trip.

Ten months later, in December 2025, the Ministry of Defence awarded Palantir a contract worth £240.6 million. It was granted by direct award, without competitive tender, using a defence and security exemption. The justification: only Palantir could supply the required services, and changing supplier would cause technical difficulties. Palantir had, by that point, hired four former MoD officials in a single year, including Barnaby Kistruck, who left his post as the department's director of industrial strategy nine days before joining the company as senior counsellor. He had helped write the Strategic Defence Review recommending an increased role for AI in defence.

The MoD hid a data breach behind a super-injunction. Then it handed £240 million in data contracts, without competitive tender, to a company it cannot audit.

This is not a story about corruption. It may yet become one, but the evidence so far does not support the charge. This is something more structural, more revealing, and in the long run more dangerous. It is a story about a state so hollowed of technical capacity it cannot build, specify, or even properly evaluate the systems on which it increasingly depends. And having surrendered the ability to think for itself, it has handed the work to a single company. And then arranged the machinery of government to ensure nobody else can compete.

The NHS Gets Tech For £1, But Fails To Smell A Rat

Palantir's route into the British state follows a sequence so consistent it might have been designed as a case study in vendor capture. It begins with a crisis, continues through dependency, and ends with contracts too large and too embedded to reverse.

In March 2020, at the start of the pandemic, Palantir offered its Foundry data analytics platform to the NHS for £1. There was no competitive tender. The platform was used to coordinate hospital capacity, ventilator supply, and vaccine logistics. The £1 contract became a £1 million extension. The extension became a £23 million two-year deal, again without competition. Further extensions followed. By the time NHS England opened the formal tender for its Federated Data Platform in January 2023, Palantir had accumulated approximately £60 million in non-competitive NHS contracts and was already embedded in the infrastructure the new platform would sit upon.

The tender was technically competitive. Two consortia reached the final stage. But trusts had been encouraged to adopt Palantir's Foundry software for pilot projects throughout the procurement period. The proposed contract value doubled during the process, from £240 million to £480 million. Palantir, leading a consortium with Accenture and PwC, won a seven-year contract worth up to £330 million in November 2023. The contract was published on the last working day before Christmas, with three-quarters of its 586 pages blacked out.

The MoD sequence mirrors the pattern. An initial Enterprise Agreement worth £75 million, signed in 2022. A strategic partnership announced during Trump's UK state visit in September 2025. Then the £240.6 million follow-on, awarded without tender the following December. More than three times the size of any previous Palantir contract with the department.

Across the two departments alone, the committed expenditure exceeds half a billion pounds. A Nerve investigation in February 2026 identified at least 34 current and past Palantir contracts with UK public bodies across ten government departments, local councils, and police forces, with a total value exceeding £670 million. The contracts include previously undisclosed deals with the Atomic Weapons Establishment, the agency responsible for Britain's nuclear warheads.

What Your Tax Payments Bought

The Federated Data Platform is intended to link operational data across NHS trusts, allowing managers and clinicians to coordinate care, track waiting lists, and plan services using integrated information rather than siloed local systems. The principle is sound. Data integration in the NHS is genuinely poor. Clinicians often cannot see a patient's full record across different parts of the system. The administrative waste this produces is real and measurable.

The question was never whether the NHS needed better data infrastructure. The question was whether the state had the capacity to specify, procure, and oversee it competently — or whether it would simply hand the problem to the first company through the door and hope for the best.

By the end of 2024, fewer than a quarter of England's 215 hospital trusts were actively using the platform. Leeds Teaching Hospitals told NHS England in a private letter, released under freedom of information, the adoption of certain FDP tools would mean losing functionality rather than gaining it. Some of the largest trusts in the country, such as University College London Hospitals, Guy's and St Thomas', Royal Free London, Sheffield Teaching Hospitals, had not adopted the system. The government's response was not to investigate why so many trusts found the platform inadequate. It was to pay KPMG £8 million to promote its adoption. By May 2025, Palantir claimed 72 trusts were using the platform; still under a third of the total.

Meanwhile, a Scottish NHS consultant named Matt Freer built an operating theatre efficiency system during the pandemic using five years of Scottish operations data. It increased theatre efficiency by more than 25 per cent: equivalent, by his estimate, to an extra 10,000 hip operations a year. He built it with a small team, using existing NHS data. He did not require a £330 million contract, a consortium led by a US intelligence contractor, or an eight-million-pound promotional campaign.

The comparison is instructive. Not because every problem can be solved by a lone clinician with a spreadsheet, but because it illustrates the chasm between what the state claims it cannot do without external help and what competent people with domain knowledge can accomplish when the procurement apparatus is not in the way.

Political Suppliers Selling To Political Clients

The British Medical Association, a partisan medical union, voted at its annual representative meeting in June 2025 to declare Palantir an unacceptable partner for the NHS, citing concerns about transparency, data governance, and the company's work with US Immigration and Customs Enforcement. In February 2026, the BMA's chair of council wrote to the British Medical Journal urging doctors to refuse all non-clinical use of the Federated Data Platform.

These objections deserve serious treatment, but they also require honest assessment. Much of the opposition to Palantir is ideological in character, and specifically hard-left. The BMA's resolution and the campaigns from Foxglove, the Good Law Project, and the Green Party are framed around Palantir's involvement with ICE, its co-founder Peter Thiel's politics, and broader anxieties about American surveillance technology in the NHS. These are not trivial concerns. The association between a healthcare data platform and a company whose software facilitates immigration enforcement does raise legitimate questions about institutional compatibility and public trust.

But the framing risks obscuring the more fundamental problem. If the NHS had awarded a £330 million contract to a British firm, with proper competitive process, for software most hospitals actually wanted to use — and the BMA had objected because the firm also held military contracts — the objection would look partisan. The reason the Palantir story matters is not primarily who Palantir is. It is what the contract reveals about the state's inability to do its own work.

A government buying good software from a competent supplier is not a scandal. It is, in many circumstances, preferable to the civil service attempting to build it. Britain's public sector IT record — from the NHS National Programme for IT, abandoned after billions in waste, to Universal Credit's digital infrastructure, years late and over budget — is not an advertisement for in-house development. The market can do things the state cannot. Companies have expertise the civil service does not possess. Procurement, done well, is a perfectly legitimate instrument of government.

The issue is not buying. The issue is buying without the wisdom to know what you need, the technical literacy to evaluate what you are getting, or the institutional independence to walk away when it does not work. When the state signs a contract it cannot audit, cannot replace, and cannot afford to cancel, it has not made a purchase. It has made a confession.

The Revolving Door Industrial Complex

Palantir hired four MoD officials in 2025 alone. Kistruck, the director of industrial strategy, joined nine days after leaving the department. Laurence Lee and Damian Parmenter, both senior civil servants, joined the same year. Former Conservative armed forces minister Leo Docherty, who lost his seat at the 2024 election, completed the quartet. The previous wave of Palantir hires from the defence establishment came in late 2022, around the time the first MoD Enterprise Agreement was signed.

The revolving door is not unique to Palantir. It is standard practice across defence and technology procurement. Former NHS director of AI Indra Joshi now works for Palantir. The NHS head was guest of honour at a Palantir investor event. The pattern is familiar and the constraints imposed on departing officials (do not use government contacts, do not advise on specific bids) are routinely described as robust. They are also routinely described as unenforceable.

But the revolving door is a symptom, not the disease. Officials move to companies because the companies value their knowledge of how the state works; its procurement processes, its decision-making rhythms, its institutional anxieties. The officials are not selling secrets. They are selling fluency. And the state, having lost the internal capacity to develop its own technology strategy, is in no position to object. It needs the companies more than the companies need it. The power relationship has inverted.

Technocracy Keeps Failing To Deliver

Every institution examined in this series has undergone the same displacement. Local auditors replaced by a centralised quango, then nothing. Experienced inspectors replaced by a digital framework the regulator could not operate. Caseworkers replaced by algorithms the department went to court to keep secret. In each case, the pattern is the same: distributed human judgment, accumulated over years, is replaced by centralised institutional machinery — which then fails.

The Palantir story is the same displacement applied to the state's technical brain. Government departments once employed people who understood how to build information systems, specify requirements, manage data, and evaluate contractors. These functions were not glamorous. They did not produce headlines. But they were the connective tissue between political decisions and operational reality. When a minister said "build me a system to track hospital capacity," somebody in the department knew enough to write a specification, evaluate tenders, identify risks, and manage delivery.

Those people are largely gone. Replaced not by algorithms but by consultants — and then the consultants were replaced by contractors. The sequence is visible across Whitehall. McKinsey writes the strategy. Accenture manages the programme. Palantir builds the platform. The state pays three times over for a capability it once possessed internally and now cannot recover. Because the people who understood the work have left. And the institutional knowledge they carried left with them.

This is why the £240 million MoD contract was awarded without competition. Not because of corruption (though the revolving door stinks) but because the Ministry of Defence has so thoroughly outsourced its data infrastructure to Palantir it genuinely cannot switch providers without, in its own words, "disproportionate technical difficulties in operation or maintenance." The dependency is real. The lock-in is real. And it was entirely predictable, because vendor lock-in is not a bug in outsourced procurement. It is the business model.

Denmark's intelligence services are now seeking to replace Palantir, fearing the US government can access data processed through the company's systems. Switzerland rejected a deal with Palantir after its army concluded there was a risk of US intelligence accessing shared data.

The United Kingdom, in the same period, tripled the size of its MoD contract with the company and invited it deeper into the NHS.

The Question This Series Keeps Asking

Palantir could be a British company and nobody would have cause to complain about its nationality. It could be headquartered in Shoreditch and the fundamental problem would be identical: a state so empty of technical capacity it cannot specify what it needs, cannot build what it wants, cannot evaluate what it buys, and cannot walk away from what it has.

The markets do what markets do. Companies pursue contracts, embed themselves in clients, and structure deals to maximise long-term revenue. This is not villainy. It is commerce. The villainy — if there is villainy — lies in the decades of decisions by successive governments to dismantle the state's internal capacity to do its own thinking, and then to hand the wreckage to whoever arrived first with a slide deck and a £1 contract.

You cannot govern through purchase orders. Governance requires the capacity to know what to build. It requires people inside the institution who understand the domain well enough to write specifications, evaluate proposals, manage delivery, and say no when the product does not work. Without those people, procurement is not a tool of government. It is a substitute for it.

This week's investigation opened with patients deleted from NHS waiting lists. It moved to regulators ordered to stand down by the Treasury. It revealed an algorithm profiling a million benefit claimants in secret, and a data breach concealed by the first super-injunction a government has ever obtained against itself. Each story documented a state maintaining the appearance of function while the reality collapsed behind the dashboard.

The Palantir story is the capstone of the week because it explains how the appearance is maintained.

• You cannot build your own systems when you have no engineers.
• You cannot audit your contractors when you have no technical staff.
• You cannot switch suppliers when you have allowed yourself to become dependent.

And you cannot admit any of this, because the admission would reveal the single most dangerous fact about the modern British state: it has outsourced the capacity to understand its own operations, and it does not know how to get it back.

The alarms have been off for long enough. The numbers have been faked for long enough. Now the buildings start to fall.

Next week: reality catches up. The armed forces cannot sustain combat for more than eight days. The courts are booked to 2030. The hospital ceilings are held up by metal props. And the police are deploying technology they cannot explain.