In February 2025, then-U.K. Prime Minister Keir Starmer traveled to Washington to meet U.S. President Donald Trump for the first time since the latter reentered office. They talked tariffs, Ukraine, and an invitation to visit King Charles. That was routine: The prime minister’s next stop was more unusual. Starmer headed to the showroom of software giant Palantir Technologies for a meeting arranged by Peter Mandelson, then the U.K. ambassador to the United States, whose now-shuttered lobbying firm Global Counsel counted Palantir as a client worth more than 30,000 pounds (about $40,900) a month. Footage shows Starmer flanked by American military personnel and Palantir’s chief executive, Alex Karp, as he inspects an attack drone.
Civil society groups such as the Good Law Project and Foxglove had questions about the visit, and about the 240 million pound ($327.1 million) contract that Palantir was awarded shortly afterward. I have a different set of questions: Why does an enterprise data-integration company have a showroom? Why does that showroom contain weaponized drone platforms and screens relaying satellite feeds? What, exactly, was the prime minister being shown, and what was he meant to feel?
Unlike its competitors, Palantir places a military and intelligence aesthetic front and center, and not just when selling to military clients. It self-consciously evokes the visual grammar of post-9/11 spy, war, and superhero films: dark “tactical” interfaces, dense webs of link-analysis graphs, and an “AI-defined” truck for the U.S. Army. Its London office has blacked-out windows and no signage; visitors reportedly sign nondisclosure agreements simply to enter. The showroom exists to immerse the visitor in that world of elite operators and threat feeds rather than the quotidian reality of data visualization. It’s theater, and so is the product.
This approach has paid off extremely well for Palantir in the U.K. In the London Review of Books, Peter Geoghegan and Lucas Amin describe a company with an unprecedented level of political access and fanatical institutional backing that is nonetheless quite ordinary, and often ineffective. Palantir began targeting Britain aggressively from 2019, with particular focus on the National Health Service (NHS). Its opening gambit came during the pandemic, when it supplied its services for 1 pound for four months.
That foothold became a much larger contract to build the NHS’s Federated Data Platform, which draws the data held in various trusts and general practitioners’ surgeries into a single system to power the apps that clinicians use. Yet 52 of the 139 trusts that the NHS classifies as “live” did not use a single Federated Data Platform app in the year preceding this June. Internal briefings call the interface “slow and clunky,” with queries that once took 30 seconds taking four or five minutes; one daily user described waiting 20 minutes for a dashboard to load, only for the system to crash. Further inquiry has met with delays and obfuscation, including a communications agency’s attempt to pay influencers to soothe public anxiety about the firm.
That refusal to act like a normal company extends to its contracts. Palantir creates structures that make it impossible to leave without punishment: Were the NHS to end the Federated Data Platform tomorrow, its staff would lose almost everything built on it, because most of the work lives only inside proprietary software and cannot be exported. The Defense Ministry has conceded that it is in the same trap, admitting that a change of supplier would mean rebuilding the underlying architecture from scratch.
Donald Campbell, the director of advocacy at Foxglove, describes Palantir’s model as being able to arrive cheap or free, learn the system, and then make itself hard to remove, a “land and expand” strategy.
“It’s a subscription service with a huge risk of lock-in,” Campbell said, “because you don’t really own what you’re building.” Palantir’s true talent, he added, is persuading each customer “that they’re irreplaceable and it’ll be too difficult to move on.” That is less a commercial strategy than an abusive relationship.
Palantir and ministers have leaned on a single number: that the software helped deliver 110,000 additional operations. Campbell told me how little that figure survives contact with its own small print: NHS England, he said, “will say, look, these are NHS figures,” but “even by NHS England’s own admission, very quietly, the claim doesn’t stand up at all,” because a methodology note concedes that the analysis “can’t establish cause and effect,” he added. The caveat was added quietly this spring, but the number is still cited in Parliament. The U.K. Statistics Authority has issued a warning over NHS England’s use of statistics to promote the platform.
Other governments haven’t been so willing to stay locked into the deal. France’s internal security service, the General Directorate for Internal Security, parted with Palantir this year in favor of the domestic firm ChapsVision. Like Germany and Spain, Paris cited digital sovereignty as the reason, but if the tool were indispensable, it is hard to see why a state would swap it out at all. Palantir’s U.S. position is unassailable, but its intimacy with the Trump White House and, frankly, extremely odd executive team make it toxic almost everywhere else.
Why, then, have successive British governments paid hundreds of millions for software that they could build themselves, and why do they defend one contractor among thousands with such ardor, even to the point that they may have misrepresented data for it?
Consider the circumstances of the British leaders who end up in the Palantir showroom. The people who end up in those kind of rooms have gained power only to discover that it is the power to administer decline. Governments operate with depleted state capacity, able to cut checks to contractors and nothing more. Nobody can build anything: High Speed 2 was conceived as a Y-shaped high-speed rail network linking London to Birmingham, Manchester, and Leeds; what survives is a single stub between London and Birmingham, and even that has seen its projected cost rise from an estimated 33 billion pounds ($45 billion) to as much as 100 billion pounds ($136 billion), with the National Audit Office concluding that it was impossible to say when it would open or what it would finally cost.
The state has similarly hollowed out its own security functions. When the private contractor G4S failed to supply the guards that it had promised for the 2012 London Olympics, the government had to call in 3,500 soldiers at the last minute to cover the shortfall. The contract was worth 284 million pounds ($387 million).
The response of a declining institution is rarely renewed rigor; it is more often to go all in on an emotionally satisfying and yet ineffective grand project. Nazi Germany spent a Manhattan Project-scale budget on its “vengeance weapons”—the V-2s that reached London killed roughly two people per rocket. The South African apartheid regime that shaped part of Palantir Chairman Peter Thiel’s childhood pursued a race-selective bioweapon that was based on the idea that Black and white people are biologically distinct.
A faltering Russia today has built T-14 Armata, a tank too costly to deploy in Ukraine, of which perhaps 20 to 50 exist against a planned 2,300. There are numerous digital wonder-weapons selling the idea that something as complex as a society can be understood and therefore manipulated: China’s Integrated Joint Operations Platform, which flags lawful behavior at scale and reinterprets it as a threat to the state; and the U.S. National Security Agency’s Trailblazer, which prefigured Palantir and was canceled after more than a billion dollars were spent.
The fantasy is always that one can hurt without being hurt, but the language of the wonder-weapon changes over time. Today, the core doctrine of modern militaries is light-footprint, network-centric warfare—combat operations in which small, elite special forces units and drone weapons systems are the tip of the spear and conventional forces play a secondary role. Palantir exists within this world, operationally and aesthetically.
One of the systems shown to Starmer during his 2025 visit to Palantir was the TITAN—the Tactical Intelligence Targeting Access Node, an ugly and squat truck from which one can coordinate the precision missile strikes and door-kicking raids that characterize the current generation of warfare. It is promoted through a video trailer that self-consciously employs film tropes down to the various companies collaborating on the project being listed like cast members in a Hollywood production. The aesthetics would be familiar to anyone who has watched a film since the planes hit the twin towers: Zero Dark Thirty, Blackhat, Sicario.
This is effective marketing, but there is nothing cynical about it. It directly reflects the stated beliefs of its creators.
Thiel’s book Zero to One is a document of one man’s insecurity. Its central tenet is that, as Thiel puts it, “competition is for losers” and “monopoly is the condition of every successful business.” Rivalry, he writes, is “a relic of history” that traps firms in a struggle for survival; only the monopolist is free to think beyond the next quarter. It is unacceptable to learn from mistakes or accept that maybe Thiel’s way isn’t the only way.
Thiel’s 2007 essay “The Straussian Moment,” written the year before Palantir first deployed Gotham, is somehow even more explicit about his anxieties: It argues that Enlightenment faith in human nature died on Sept. 11, 2001, and that the West must recover Carl Schmitt’s distinction between friend and enemy. Together, these ideas point to what Palantir has become and the fantasy it sells: as with the V2, it is to strike at enemies without being able to be struck in return, to run a business without needing to produce results for clients. He is also no fan of the NHS, calling Britain’s affection for it ‘Stockholm syndrome’.
His business partner shares this outlook. Karp’s The Technological Republic is declinist from its first page.The book’s argument is that the technology industry has feminized the West, that it is emasculating for a nation’s foremost companies to be “soft” platforms for gossip or fashion-forward smartphones, and that the remedy is a return to the “hard” engineering of the weapons like those Starmer was shown. This is part of a larger project of reclaiming a “national purpose” for the US and the West in general.
The book does in a few hundred pages what F.T. Martinetti was able to do in 71 words in the ninth and tenth points of the “Manifesto of Futurism,” and yet is unable to name the object of its critique. Gender is never cited as the locus of Karp’s unease, but it saturates every page. He also has merch.
Klaus Theweleit, studying the paramilitaries that became the Freikorps, the post-WW1 German far-right militias, found in them a psyche organized around terror of the formless, the fluid, the unbounded, and the feminine. The answer was armor, the self remade as a rigid, bounded machine, stripped of ambiguity and femininity, unable to lose. Though its creators are nothing like the young men returning from the trenches, this would seem to be where they have ended up psychologically, and they are not alone – this aesthetic has a broad appeal.
Consider what Palantir fears: unstructured data, the shapeless mass of unsorted information, the messy compromises of living in a plural world. Consider what it sells: an ontology of objects and links, a single fixed model in which everything sits in its assigned place and can be acted upon. It is a fantasy, and specifically a male fantasy, of total mastery and hard boundaries, cloaked in the visual language of genre fiction and sold to states that also lack true confidence.
That is what Palantir understood about its customers, in large part because they felt that fear of dissolution and irrelevance, too. They were fearful men who could be sold technological potency wrapped in operator chic. History is not kind to institutions that choose hysterical masculinity over the competence and compromise that would actually have served them.
