The palantíri are the seeing-stones of The Lord of the Rings, objects that allow one to see across great distances and through time. Peter Thiel, co-founder of PayPal and one of Silicon Valley’s most influential investors, chose that name for the company he founded in 2003 alongside Alex Karp, a philosopher with a doctorate from the University of Frankfurt who had never written a line of code. The premise was simple and ambitious: after September 11, America’s intelligence agencies were drowning in data they could not process. The CIA, the NSA, the FBI, all had mountains of information scattered across incompatible systems. What was missing was not more information but the ability to integrate it, cross-reference it, and find patterns. Palantir set out to build the software that did exactly that.
Twenty years later, Palantir revenues approach three billion dollars a year. It is publicly traded at a valuation that defies any conventional metric. It holds a contract with the United States Army worth ten billion dollars, possibly more. It operates the Pentagon’s most important artificial intelligence system. And its CEO claims his software is “responsible for the majority of targeting in Ukraine.” It is, by a wide margin, the most important company most people have never heard of. And it represents something beyond its own business: the transformation of war from a matter of hardware to a matter of software.
From Silicon Valley Startup to Defence Contractor
Palantir was born of CIA money, literally. In-Q-Tel, the agency’s venture capital arm, was one of the first investors. This was not unusual in the Silicon Valley ecosystem, which owes its original existence to the Department of Defense and the American intelligence community, but what was unusual was the degree of integration Palantir sought with its government clients. The company did not sell a packaged product. It sent engineers who sat alongside intelligence analysts, understood their problems, adapted the software to their specific needs, and stayed. It was a technology consulting model disguised as a software company, and it proved extraordinarily effective at generating dependency.
The first product was Gotham, a platform designed for intelligence analysts and military operators. Gotham takes data from disparate sources, signals intelligence, satellite imagery, human intelligence, financial data, social media, and integrates them into an interface where an analyst can visualise connections that would otherwise remain invisible. If a terrorist network moves money through three different countries using intermediaries who appear in separate databases, Gotham can connect the dots. If a military convoy moves at night and shows up only in a thermal image captured by a drone, Gotham can cross that image with intercepted signals data and locate the convoy in real time.
Then came Foundry, aimed at the commercial sector and non-military government agencies. And more recently, the Artificial Intelligence Platform, which integrates generative AI models with clients’ operational data. But the core business was always the United States government. In 2024, fifty-five per cent of Palantir’s revenue came from government clients, the vast majority of them American.
The growth trajectory tells the story. From seven hundred million in revenue in 2019 to nearly three billion in 2024. In 2025, the Army awarded a framework contract of up to ten billion dollars over ten years, consolidating seventy-five prior contracts into a single agreement. The Pentagon raised the ceiling on the Maven Smart System contract to one point three billion. More than twenty thousand active users across more than thirty-five Department of Defense tools. Palantir stopped being a startup a long time ago. It is an institution of the American security apparatus, so deeply integrated into Pentagon operations that replacing it would be extraordinarily difficult.
The Controversy: Tech Workers Against Military Contracts
There is an episode worth pausing over because it illuminates a cultural tension that still defines the relationship between Silicon Valley and the Pentagon, even if that tension is being resolved more and more clearly in favour of collaboration.
In 2017, Google signed a contract with the Department of Defense to participate in Project Maven, a programme that used artificial intelligence to analyse drone imagery. The contract was relatively modest. But when Google’s employees found out, more than four thousand signed an internal petition demanding the company cancel it. Dozens resigned. The open letter to CEO Sundar Pichai read: “We believe that Google should not be in the business of war.” Google gave in. It did not renew the contract when it expired in 2019.
What followed was instructive. The contract Google rejected eventually found its way to Palantir. Project Maven became the Maven Smart System, and Palantir became its principal contractor. Alex Karp, who has no qualms about publicly declaring that his mission is to “defend the West” and “terrorise our enemies,” took the work Google found morally unacceptable.
The episode revealed two things. First, that the moral refusal of a technology company to work with the military is a gesture, not a policy: if Google won’t do it, someone else will, and the Pentagon is not going without software because a handful of engineers in Mountain View have scruples. Second, that Silicon Valley’s internal culture was split between a current that saw military collaboration as incompatible with the industry’s values and another that saw it as a strategic and patriotic obligation. Seven years later, the second current won. Microsoft, Amazon, Google itself through subsequent contracts, and of course Palantir and Anduril, all work with the Pentagon. The era of “Don’t be evil” as a defence policy is over.
Karp is a difficult man to classify. Someone with a doctorate in neo-Kantian philosophy who runs a mass surveillance company. He describes himself as progressive on some issues and hawkish on defence. He backed Kamala Harris in 2024 but also Elon Musk’s DOGE policies. He is eccentric, lives in a renovated barn in New Hampshire, practises tai chi, and at the same time is the CEO of the company that provides the targeting tools for American military operations. The contradiction, if it is one, does not appear to trouble him.
Ukraine: The Laboratory
If there is one place where Palantir proved what it can do, it is Ukraine. Karp was the first CEO of a major Western technology company to visit Kyiv after the Russian invasion, and Palantir provided its software to the Ukrainian government for free initially, as a strategic investment whose return would come later, in the form of combat data and reputation.
What Palantir’s software does in Ukraine is, in essence, compress the targeting cycle. The process that runs from detecting an objective to destroying it, which in a conventional army can take hours or days, in an army running Palantir can take minutes. The software integrates drone imagery, commercial satellite data, signals intelligence, field reports, open-source information from channels like Telegram, and cross-references all of it in real time to present the commander with an updated operational picture and attack options prioritised by algorithms. The models learn from every strike: they feed results back to refine future target identification.
Ukraine is smaller than Russia, has fewer soldiers, fewer tanks, less artillery, fewer aircraft. But thanks to software like Palantir’s, and the Ukrainian Delta command-and-control system that integrates with these platforms, Ukraine fights with a qualitative advantage in information processing that partially offsets its quantitative inferiority. When Karp says Palantir’s software amounts to “having tactical nuclear weapons against an adversary with only conventional arms,” he exaggerates, but not as much as it sounds. Information asymmetry is a real form of military power.
In January 2026, Palantir and the Ukrainian Ministry of Defence launched Brave1 Dataroom, a secure environment for training AI models with real battlefield data. The initial focus is on detecting and intercepting Shahed-type drones, the Iranian drones Russia launches nearly every night against Ukrainian cities. The idea is for Ukrainian engineers, using Palantir’s platform, to develop defence algorithms trained on actual combat data, something no laboratory can simulate.
But the most significant aspect of Palantir’s presence in Ukraine is not technical but strategic. As a Carnegie analysis noted, Palantir and similar companies are creating something that might be called a “software-defined alliance”: a form of security guarantee that requires neither treaties nor troops, but technological integration. Countries that operate on Palantir’s platform fight at NATO speed without being NATO members. Technological dependency functions as a de facto alliance. It is an updated version of what Lockheed Martin does with the F-35, only faster, cheaper, and harder to see.
The Future of War Is Software, Not Hardware
There is a phrase circulating in American defence circles that Palantir repeats like a mantra: the future of war is software, not hardware. The idea is that the decisive military advantage of the twenty-first century will not be who has more tanks or better aircraft but who can process information faster and make better decisions in less time.
This does not mean hardware ceases to matter. Tanks are still relevant. Aircraft are still necessary. But hardware without software is increasingly a depreciated asset. A tank without the ability to receive real-time intelligence, to integrate into a digital battle network, to be directed by a system that processes satellite and drone data simultaneously, is a twentieth-century tank in a twenty-first-century war. Russia has thousands of tanks. The informational advantage Ukraine gets from software turns many of those tanks into targets, not threats.
The Pentagon appears to have understood this. The Department of Defense’s artificial intelligence budget grew a hundred and seventy per cent between 2022 and 2024. The Maven Smart System went from a handful of users to more than twenty thousand in little over a year. Venture capital investment in defence technology has multiplied since 2021: nearly a hundred and thirty billion dollars in four years.
Palantir is the most visible beneficiary of this trend but not the only one. Anduril, founded by Palmer Luckey, the creator of Oculus VR, builds autonomous surveillance towers and drone systems. Shield AI develops autonomous pilots for unmanned aircraft. Scale AI provides training data for military AI models. An entire generation of defence technology companies is growing at the intersection of Silicon Valley and the Pentagon, and this time without the moral scruples that characterised the previous decade.
Big Tech and the Pentagon: An Inevitable Marriage
For years, the relationship was uncomfortable. The Pentagon needed Silicon Valley’s technology but did not know how to buy it. Its procurement processes, designed to contract aircraft carriers and fighter jets, were not equipped to purchase software that updates every week. Silicon Valley, for its part, found the Pentagon bureaucratic, slow, and morally awkward. The engineers building the platforms did not want to know that their code was being used to locate drone targets.
All of that changed. It changed because China showed it has no qualms about integrating its technology sector with its military apparatus. It changed because Ukraine demonstrated that commercial technology can be decisive on a real battlefield. And it changed because money is money: defence contracts represent a stable, predictable, and growing source of revenue at a time when the commercial tech sector faces uncertainty.
Palantir understood this before anyone else, and that is probably its greatest competitive advantage. It does not have the best technology in the world, though its technology is very good. What it has is twenty years of experience working with the most demanding and paranoid agencies in the American government, a corporate culture that is not embarrassed to work with the military, and a demonstrated capacity to integrate its systems into actual combat operations. When the Army consolidated seventy-five contracts into a single agreement with Palantir, it was not buying software. It was acknowledging that Palantir had become part of its operational infrastructure.
There are risks in this, and they are worth mentioning. The concentration of intelligence and targeting capabilities in a single private company raises legitimate questions about oversight, accountability, and the power a for-profit corporation exercises over life-and-death decisions. When Karp says Palantir is “responsible for the majority of targeting in Ukraine,” that claim, if true, should generate at least as much reflection as admiration. Who audits the algorithms? Who oversees the targeting criteria? What happens when an AI model optimised for efficiency makes a mistake a human analyst would have caught?
These questions have no easy answers, and they are probably not going to slow the trend. The integration between Big Tech and the Pentagon will deepen because the alternatives are worse. A Pentagon that does not adopt artificial intelligence is a Pentagon that loses its edge against China. A Silicon Valley that refuses to work with the military is a Silicon Valley that cedes that ground to competitors with fewer scruples and less capability. War, as Clausewitz would have recognised, adapts to the means available. And the means of the twenty-first century are data, algorithms, and software.
Palantir, with all its contradictions, its philosopher CEO who talks about “defending the West,” its origins in CIA capital, its name taken from a fantasy novel, is the company that best embodies this new reality. The future of war does not look like what we imagined. It lacks the glamour of a stealth fighter or the spectacle of a bombing. It looks more like an engineer staring at a screen in a grey office, cross-referencing satellite data with signals intercepts while an algorithm suggests coordinates. Less cinematic. But increasingly, it is what decides who wins and who loses.