The Terrifying Business of AI Warfare - How AI is transforming warfare by making weapons cheaper, smarter and increasingly autonomous, while giving private technology companies unprecedented influence over how wars are fought [OC]

**TL;DR: Cheap drones are disrupting the economics of warfare, AI is shortening the kill chain and driving the rise of autonomous weapons, and a new software-driven defense model is embedding private tech companies into the heart of war like never before.** # The Terrifying Business of AI Warfare Throughout modern history, military dominance belonged to nations that could afford the largest armies, the strongest fleets, and the most lethal weapons. But today, war is showing us that a drone worth only a few hundred dollars can disable a tank worth millions. Behind this revolution in how conflicts are fought is a rapidly growing industry racing to integrate AI into next-generation weapons. Companies are building drones and unmanned vehicles capable of navigating without GPS, continuing to operate under communications jamming, and seamlessly coordinating with other machines. Meanwhile, AI systems are scanning thousands of surveillance feeds, identifying suspicious patterns, and generating potential targets at a speed no human analyst can match. What happens when the cheapest weapon on the battlefield also becomes the smartest? And who stands to profit from it? ## The Economics of Traditional Defense Traditionally, the defense industry has been built around increasingly complex and expensive aircraft, missiles, warships, and armored vehicles, with military advantage closely tied to a nation's defense spending. However, in 2023—one year after Russia's full-scale invasion of Ukraine—the economic logic of war began to shift dramatically. Unmanned and remotely piloted drones have completely transformed the Ukrainian battlefield, accounting for an estimated 80% of combat casualties on both sides. ### How Cheap Weapons Destroy Expensive Equipment Soldiers strap explosives to small, disposable racing drones, converting them into precision-guided weapons—these are known as FPV drones. They are everywhere, and even the $10 million American Abrams tank—once one of the most dominant weapons in conventional warfare—was temporarily pulled back from parts of the front line because drones made it so vulnerable. Ukraine's FPV drone production surged from approximately 800,000 units in 2023 to roughly 4.5 million per year in just two years. Because these drones are so cheap to manufacture, losing them becomes an expected cost of combat; because they are assembled from readily available consumer-grade components, they can be mass-produced—and mass production itself becomes a military advantage. ### The Inverted Economics of Defense The problem is that the threat from these cheap drones forces nations to spend heavily protecting the very infrastructure these drones can destroy. In Ukraine, high-voltage transformers have been encased in concrete structures and surrounded by massive anti-drone nets designed to block FPV drones. But this defense has proven surprisingly easy to breach: Russia has demonstrated how to tear a hole in the net with one drone, allowing another to hit the transformer—or simply fly around it altogether. A $3.5 million transformer is being destroyed by drones that cost only a fraction of that. ## How AI Is Redefining the Battlefield The revolution kicked off by cheap drones is spreading across the entire battlefield, with companies developing autonomous aircraft, ground vehicles, and naval vessels. ### Autonomy vs. AI It's worth noting that not all autonomous military systems or weapons are necessarily AI-driven—some rely solely on pre-programmed rules. The distinction is that AI can push autonomy further, enabling machines to interpret their surroundings, adapt to changing conditions, and make increasingly complex decisions with less human input. Some systems can even continue their mission under communications jamming, maintaining their course even after losing contact with human operators. The commercial value lies in the fact that autonomy fundamentally reduces the number of skilled operators required: where one drone previously needed one operator, a single operator may now manage an entire fleet. The business opportunity shifts from selling individual aircraft to providing hardware, software, sensors, communications, and ongoing upgrade services to keep the entire autonomous system running. ### Shortening the Kill Chain AI can change warfare without even controlling a weapon. It can sift through massive volumes of surveillance data, identify suspicious patterns, and rank potential threats far faster than human analysts could process them manually—meaning militaries can scale their targeting processes without adding personnel. Much of this command infrastructure has existed for decades; what AI changes is the degree of automation in analysis and prioritization. The result is a dramatically shortened kill chain—the process by which militaries find a target, identify it, decide how to respond, and ultimately execute a strike. Tasks that once had to flow between multiple people and multiple systems can now be analyzed and prioritized by software. ### The Lavender System in Gaza One of the most controversial cases of AI-assisted targeting emerged in 2024 during Israel's war in Gaza, with a system called Lavender. Reports indicate it helped generate a database of up to 37,000 suspected militants, with one intelligence source stating that some individuals flagged by the system were assessed in just 20 seconds—a human life reduced to a 20-second review. Even if these systems correctly identify the intended target, it doesn't mean civilians are safe. Reports from Gaza indicate that some strikes against suspected militants were authorized even in situations where there was a high risk of civilian casualties. The Israel Defense Forces disputed this, stating that Lavender does not select targets autonomously and that intelligence analysts must independently verify its recommendations. But even so, this raises a bigger question: **If AI generates potential targets faster than humans can thoroughly evaluate them, how meaningful is human approval?** ## The Software-Driven Defense Model Once software becomes this critical to warfare, it changes how defense companies make money. Defense firms have traditionally earned revenue primarily by designing and selling physical equipment, but as software becomes increasingly important to system performance, the initial hardware sale is merely the starting point of a commercial relationship. There are enormous revenue opportunities through software licensing, updates, data processing, maintenance, integration, and entirely new features added long after the system has been delivered. Weapons are looking less like finished products and more like constantly evolving platforms. ### Procurement Can't Keep Up with Software This new model also creates a serious vulnerability: software evolves far faster than traditional military procurement processes. Currently, a new weapon can take years to go from requirements to testing, budgeting, government approval, and eventually procurement and deployment to the battlefield. By that time, the software and tactics in use on the battlefield have already moved on, rendering the system outdated before it even enters combat. > "Ukrainian technology, drone technology, becomes obsolete every two weeks. So the biggest question is: how do you improve, how do you make the procurement system more agile and responsive to evolving threats?" What's more, adversaries may not face the same constraints. Nations willing to accept more risk and bypass lengthy approval processes may be able to push a new capability from development to the battlefield significantly faster, creating a dangerously lopsided dynamic. Some argue that "if the U.S. doesn't lead in this area, authoritarian regimes will—and they won't care about our ethical standards." As a result, procurement itself must change: from purchasing finished weapons every few years to acquiring systems that can be continuously updated, tested, and improved—at a cadence closer to software development. ## The New Defense Gold Rush and Supply Chain Vulnerabilities ### The Rise of New Defense Firms This shift has already triggered a new defense gold rush. Companies like **Anduril, Helsing, Shield AI, and Palantir** are attracting billions in investment, racing to build next-generation military technology. One of their biggest advantages over traditional defense contractors is speed: instead of spending decades developing new fighter jets, tanks, or warships, they leverage off-the-shelf commercial software, COTS components, and rapid testing to deploy new technology to the front lines much faster. Helsing's HX02 is one example—a software-defined strike drone designed for mass production, continuously refined based on battlefield lessons from Ukraine. ### Supply Chains Dependent on the Adversary This software-driven approach carries a troubling dependency: many of the motors, batteries, sensors, and materials required for mass-producing these drones still come from China—precisely the West's greatest long-term competitor. A nation could theoretically develop the world's most advanced autonomous weapons yet find itself unable to manufacture enough of them without relying on an adversary's supply chain. Investors are increasingly focused on the software behind the weapons. The real commercial advantage likely lies in controlling the systems that connect drones, sensors, commanders, and weapons on the battlefield—which is why these companies are now valued more like major tech firms than traditional defense manufacturers. ## The Risks of Private Companies at the Center of the Kill Chain As software becomes ever more critical, militaries become increasingly dependent on the companies that control it. A nation may possess the physical weapons yet still rely on outside suppliers for communications, updates, and the technology needed to unleash their full potential. ### Concentrated Points of Failure Placing private companies at the center of the kill chain creates new points of failure. If a single company provides the software that drives thousands of autonomous weapons, an adversary doesn't need to hack a thousand drones individually—they can target that company or the update infrastructure that controls them all. This also means that access to critical military technology may depend on decisions made outside the military chain of command. **The Ukraine Case**: Its military relies heavily on Starlink for battlefield communications. International news agency Reuters later reported that Elon Musk ordered partial coverage shutdowns over Russian-occupied areas during a 2022 counteroffensive, which SpaceX disputed. **The Anthropic and Palantir Case**: In 2026, the Pentagon flagged AI company Anthropic as a supply chain risk stemming from a controversy over how its Claude model was being used. Claude technology had been embedded in parts of Palantir's Maven intelligence system—the platform the U.S. military uses for battlefield intelligence and weapons testing. When the Pentagon took action against Anthropic, Palantir suddenly faced the daunting task of removing Claude and rebuilding part of the system the military had come to depend on. Unlike hardware, deeply embedded software may be extremely difficult to replace. Switching to a different drone manufacturer might be relatively straightforward, but swapping out software deeply embedded in intelligence, targeting, and command systems is an entirely different matter. **The core question thus emerges**: Who truly owns the kill chain? Is it the government or military deploying the weapons, or the companies controlling the technology behind them? ## The Battlefield as a Test Bed: The Commercial Value of Data Unlike traditional weapons—which can be tested for years away from combat—many new drones and autonomous systems are being improved using data collected directly from war zones. The Ukrainian battlefield has become a real-world test bed. These experiments allow defense companies to see what survives, what gets jammed, how adversaries adapt, and where their technology fails. Ukraine has even created the **Brave One data room**, a secure environment developed in partnership with Palantir that allows approved partners to use real battlefield data to train and improve AI models. A feedback loop emerges: combat produces data, data improves systems, and improved versions go straight back to the battlefield. A company that proves its product works effectively on an active battlefield possesses something far more valuable than a successful demonstration—real-world evidence that can translate validated battlefield performance into commercial momentum far exceeding the conflict itself. For the winners, this means billions of dollars in future business. But the darkest aspect is that every failure, every destroyed vehicle, every lost life can become yet another lesson used to make the next generation of weapons more lethal. ## The Counter-Drone Industry and "Humans as the Slowest Link" The drone boom has also spawned an entire industry dedicated to intercepting them. Companies are developing counter-drone systems ranging from jammers and radar to automated guns, lasers, and even handheld devices small enough for a single soldier to carry—while the economics demand finding cheap ways to detect and destroy these drones to make the exchange ratio sustainable. > "Every advance on one side creates opportunities for the other. Better drones drive demand for better defenses, which in turn drive demand for smarter drones." ### Human Approval Becomes a Formality But as both sides race to respond faster, another major problem surfaces: **humans may quickly become the slowest link in the system.** Militaries may still require a human to approve every target, but if they're asked to review hundreds of AI-generated recommendations in just a few seconds, that approval becomes meaningless. Human oversight only truly matters when the person has the time, the evidence, and the authority to disagree with the machine. And under the immense pressure of war, the temptation to remove humans from the process will only grow stronger. ### The Emergence of "Terminator Mode" This may have already happened. In 2024, Ukraine reportedly tested 10 AI-controlled drones operating in so-called "Terminator mode," without requiring human approval for final attacks. According to the developer who later disclosed the test, the drones killed several Russian soldiers—potentially marking the first reported case of fully autonomous drones killing in combat. When one nation allows its weapons to identify and engage threats faster and with less—or no—human oversight, its adversaries may feel compelled to do the same, simply to level the playing field. Source: The Terrifying Business of AI Warfare - Blueice607 (https://www.youtube.com/watch?t=1s&v=g1VTqmiQUGE)