Palantir’s controversy is the product
Palantir’s fiery rhetoric helps mystify its mostly mundane tech — propping up its share price and preserving its national security contracts
Most corporate statements of intent get little to no attention — but Palantir is evidently not most corporations. When it released a 22-point manifesto onto X on April 18, the post quickly amassed more than 35 million views, and attracted media coverage across the world.
The thread suggested that the West “must resist the shallow temptation of a vacant and hollow pluralism,” claimed some cultures “remain dysfunctional and regressive” and decried the “postwar neutering” of Germany and Japan in the wake of World War II.
As the manifesto showed, Palantir rarely shies from controversy. Indeed, both the company and its CEO seem to actively court it, almost gleefully embracing the company’s role in controversial US government programs, including ICE mass deportations. The company seems to be aggressively leaning into Trump’s agenda even as Democrats seem set to take at least one chamber of Congress at the midterms, and as European governments diverge from its course. This doesn’t just make Palantir itself controversial in key markets, but also drives opposition to the very idea of AI-powered public services — seemingly against its own business interests.
The go-to answers typically paint Alex Karp, along with Palantir founder Peter Thiel, as ideologues. That may be part of the story — but it is not all of it. The fuller explanation is that Palantir’s controversy is not a tax on the business. It is the business. The brash narrative the company puts forward is a way to mystify what is ultimately a rather mundane product. That in turn serves as a mechanism to prop up the company’s share price — and as a way to preserve its national security moat, which gives Palantir its most lucrative contracts.
What Palantir does
A casual onlooker might be forgiven for thinking that Palantir is an AI company in the same vein as OpenAI or Anthropic, or an arms company like BAE Systems or Raytheon. In reality, it is neither. Palantir does not have a foundational AI model of its own, and has made no suggestion that it is trying to develop one. Similarly, it does not build surveillance devices, drones, or weapons hardware in any conventional sense. For all the company plays a role in surveillance networks, it does not build cameras.
Palantir rarely seems upset at being credited with powerful and potentially nefarious capabilities. Perhaps most famously, Palantir is credited with playing an (unspecified) role in the operation to kill Osama bin Laden — something its executives have coyly failed to deny, but have not elaborated upon. More recently, Palantir’s software reportedly played a role in the US capture of Venezuelan president Nicolás Maduro, albeit using Anthropic’s underlying technology. And only last month, London’s Metropolitan Police revealed it had launched investigations or disciplinary action into hundreds of its own staff based on an AI tool developed by Palantir. It’s the stuff of Hollywood thrillers and action films, and helps build mystique around the company.
“If the smartest thing the devil ever did was persuading people he didn’t exist, the smartest thing Palantir ever did was persuading people they had a role in the Osama bin Laden operation without anybody ever leaking exactly what it was,” says a former top-level UK government official heavily involved in IT and security — who, like others Transformer spoke to, did not want to discuss Palantir on-the-record owing to its ferocious and controversial reputation. “The whiff of sulfur in tech never cost anybody a dime.”
Yet much of Palantir’s actual operations are more prosaic.
“They take loads of shit, hard-to-gather datasets from all over different bits of an organization — or different bits of the data ecosystem that are completely shit, pretty unreliable and take ages to get – and put them in a common, usable, aggregate format,” says the UK official. “They clean them up and they present them to you: Beautiful, mostly accurate, and incredibly user friendly.”
The official stressed that role — cleaning up existing data sources and presenting them in a more usable format — was both an incredibly difficult and valuable function, which public services across the world in particular had struggled to fill themselves. Multiple other sources from public procurement, rival businesses, and a former Palantir staffer, said the characterization of Palantir as a data middleman — albeit one using AI to enhance its work — was a fair one. In reality, Palantir operates more like Oracle, or the consulting side of Amazon Web Services, than a defense or surveillance tech business. Concretely useful, surely, but not sexy.
“Palantir is just the latest supplier of complex middle- and back-office software hoping to ride the wave of governments’ magical thinking,” said one international consultant, who advises governments across the world on digital public services.
Such software may seem mundane, but it is a good business.
Palantir’s public accounts show that it is both a growing and readily profitable business — with margins most companies would kill for. In 2022, Palantir recorded adjusted EBITDA (a key measure of profitability) of $443m from sales of $1.9b. By 2025, that number stood at $2.3b from sales of $4.5b. That represents sales more than doubling over three years, but profits increasing by just over 5x. Palantir’s revenues are just a tiny fraction of Amazon, Apple, Google or Meta, but it has a success story to tell.
Those numbers exist in a different reality, however, to Palantir’s stock price. Over the same period, the company’s market capitalization increased by far more than either sales or operating profit. On December 31, 2022, Palantir was valued at just over $15b. By the same date in 2025, it was worth $453b — almost a 30x increase. At the time of its last annual accounts, Palantir’s market capitalization was 103x its revenues, and 205x its profits. Compare that to Tesla, another much-hyped stock: even its market cap is “only” 15x its revenues and around 100x its operating profits. Palantir has a higher proportion of retail investors (generally seen as less sophisticated) than Tesla, too: around half of its shares are held by retail investors and public companies, versus 40% of Tesla, with even that smaller proportion seen as unusual among big companies.
Sustaining such a high stock price requires a narrative which goes above-and-beyond the actual business reality. Tesla, for its part, has always claimed not to be a car company. First, it was an AI company, about to launch full self-driving and a fleet of automated taxis. Now, it tells a story about being the world’s leading robotics company instead.
Palantir is arguably doing something similar. Its controversial and even nefarious reputation might well be what supports its stock price. It is selling a narrative to justify excitement, presenting itself as an indispensable AI company ready to ride a political wave to the tune of trillions.
That is not necessarily the only reason for the company’s fiery rhetoric, however. While its loud, jingoistic language may cost Palantir deals in the UK and Europe, its reputation as an American champion might be useful in securing US government contracts. Certainly that is the case while Donald Trump is president and Republicans control Congress. But it may persist even after that. Selling data integration to the Pentagon, NSA, or ICE requires more than technical competence: it requires being trusted to service the government even when contracts become politically toxic.
Viewed in this light, Palantir’s controversial positioning is the point. Each new ICE contract is a signal to other procurement officers that the company is willing to do whatever dirty work is on offer, and will not pull out when the news cycle turns. Palantir’s manifesto signals to the national security state that Palantir is the rare tech vendor that shares its worldview, or is at least willing to pretend convincingly. It may antagonize EU officials, and cost the company financially: Europe is a large market, and a potential growth area for Palantir as its governments increase their defense spending. But that is the cost of winning American business. Being less trusted in Brussels arguably makes the company more trusted in Langley and Arlington.
Palantir’s executives often seem dragged in two directions. Sometimes what would best serve the growth of the substantive business — becoming a quiet, boring company that secured multi-million dollar public sector contracts across the world — is in conflict with what will support the share price, keep retail investors excited about the stock, and secure its US government moat.
This has wider lessons and consequences, particularly as the leading AI companies deepen their own national security work and prepare to go public. As Palantir shows, this can warp business incentives — which can, in turn, shape a company’s politics.
The perverse incentives and challenges that look to have shaped Palantir’s political interventions will soon apply to far more of the AI sector than they currently do. That could materially affect the roadmap of AI development. The uncomfortable truth is that the rollout of AI will be shaped not just by what works, but by what produces a good story. The Palantir playbook may soon become an industry standard.
 | A guest post by
|