PALANTIR: War, Surveillance, Artificial Intelligence

Part 6: The Future and Implications, Pax Technica

Chapter 12. The U.S.-China AI Hegemony Competition and Geopolitical Risks

Attorney Kyungjin Kim

A. The West's Response to China's Technological Rise

(1) Civil-Military Fusion and China's Palantir: Deepexi

In the fall of 2024, a research team at Georgetown University in Washington, D.C. was poring over a massive Excel file. On the screen were 2,857 contracts. They were AI-related procurement contracts signed by the Chinese People's Liberation Army from January 2023 to December 2024. Researchers at the Center for Security and Emerging Technology (CSET) analyzed this data and discovered a pattern. Of the 1,560 contracting companies, a significant number were private firms. Three-quarters of them were not state-owned enterprises. What was even more striking was that two-thirds of these companies had been founded after 2010.

What this data conveyed was clear: China's civil-military fusion strategy was actually working.

The term civil-military fusion first appeared in China in the late 1990s. But it transformed from a mere slogan into a tangible structure only under Xi Jinping. In 2017, Xi established the Central Commission for Integrated Military and Civilian Development. The Chinese Communist Party's highest-level body would directly oversee this strategy. That same year, the Ministry of Science and Technology issued a five-year civil-military fusion plan, and the State Council released an artificial intelligence development plan. Three documents were aligned toward a single goal: converting civilian technological innovation into military capability.

There is a reason the West finds this strategy difficult to understand. In the United States and Europe, the civilian and military sectors are separated. Google employees revolt against drone projects, and Microsoft engineers send protest letters over Pentagon contracts. But in China, no such boundary exists. Alibaba's cloud computing capabilities can be directly linked to the PLA's command-and-control systems. Tencent's facial recognition technology is applied to surveillance cameras in the Uyghur region. Hikvision's security monitors are already being supplied to the Chinese military's surveillance and video processing systems.

Former U.S. Assistant Secretary of State Christopher Ashley Ford summarized the situation this way: "Every technology that is accessible in China, and that Beijing deems useful for military and national security purposes, will be used for those purposes. Come what may."

In this context, the name Deepexi emerges. This company revealed itself to the world when it pursued a listing on the Hong Kong stock exchange in October 2025. According to its prospectus, Deepexi defines itself as a "provider of enterprise-scale large AI application solutions." It explains that it bundles data integration, real-time analytics, and predictive modeling into a single package to support corporate decision-making. It is remarkably similar to what Palantir started doing in the United States twenty years ago.

This is why Western intelligence analysts call Deepexi "China's Palantir." Both companies integrate data, visualize it, and use AI to accelerate decision-making. But there is a critical difference. Palantir grew under the demanding constraints of Western legal restrictions and privacy protections. Deepexi faces no such constraints. It enjoys the Chinese government's full support and unlimited data access. Under the civil-military fusion system, data collected by civilian companies can be immediately repurposed for national security.

Georgetown's CSET analysis shows how effective this strategy actually is. Among the top 15 companies in AI-related military procurement contracts, eleven were still state-owned enterprises or national defense research institutions. Names like China Electronics Technology Group Corporation (CETC), China Aerospace Science and Technology Corporation (CASC), and NORINCO. But the rest were private companies. iFlytek, a leader in voice recognition and natural language processing, won 20 military contracts over two years. They included translation software and AI-based decision-support systems. PIESAT, a satellite and geospatial analytics company, supplied combat simulation platforms and automatic target recognition technology.

This is why the West is on edge. China's defense industry ecosystem is expanding. Young, agile private companies are entering domains that were once monopolized by massive state-owned enterprises. These companies are supplying cutting-edge AI technology to the military, rapidly boosting the PLA's technological capabilities.

Under Xi Jinping's direction, China is now seeking to make civil-military fusion a core pillar of the 15th Five-Year Plan (2026-2030). This initiative, called "AI Plus," aims to deeply integrate AI across the economy and society. In March 2024, Premier Li Qiang declared at the National People's Congress that he would "develop new quality productive forces at a faster pace." This signifies the completion of a system where civilian AI advances automatically translate into military strengthening.

The West's response is proceeding on multiple levels.

The first is narrowing the gateway for semiconductors and computing power. The U.S. Bureau of Industry and Security (BIS) tightened advanced semiconductor export controls in December 2024. In January 2025, a global "AI Diffusion" rule was announced, an attempt to block circumvention through third countries.

The second is the realignment of alliances. AUKUS Pillar 2 institutionalizes joint development among the United States, the United Kingdom, and Australia in AI, autonomous systems, cyber, electronic warfare, and quantum technology. This is not a simple arms purchasing alliance. It is a technology alliance that shares software and data.

The third is the classification and sanctioning of Chinese companies. The U.S. Department of Defense maintains a "Chinese Military Company" (CMC) list under the FY2021 National Defense Authorization Act. Section 805 of the FY2024 NDAA goes a step further. Starting June 30, 2026, the Department of Defense will be unable to contract not only with companies on this list but also with any other companies that supply goods or services produced or developed by those listed companies.

But export controls and sanctions alone are not sufficient. In October 2025, China expanded export controls on rare earth elements including holmium, erbium, thulium, europium, and ytterbium. These elements are essential for F-35 fighter jets, Tomahawk missiles, radar systems, and AI semiconductors. China dominates 60% of global rare earth mining and 90% of processing. It is a strategy of converting economic use into a tool of defense power.

Palantir's Alex Karp says of this situation: "China and we are in an AI arms race. We are already in a state of war." His CTO Shyam Sankar shares the same view. He analyzes that China is stealing Western intellectual property, economically infiltrating through the Belt and Road Initiative, and pressuring the West with hybrid tactics.

Ultimately, the West's response to China's technological rise goes beyond a simple technology competition. It is a contest between systems. A system in which the state integrates and mobilizes everything versus a system that pursues innovation while preserving democratic values and market principles. Palantir stands at the center of that contest, asking the West: Can you secure technological advantage while still protecting freedom? The answer to that question has not yet been written.

(2) Taiwan Strait Crisis Simulations and AI as Deterrence

On July 31, 2025, retired military officers and security experts gathered in the auditorium of the Center for Strategic and International Studies (CSIS) in Washington, D.C. On the screen appeared the title "Lights Out?" The lights go out. The title was a metaphor for what would happen in the event of a Taiwan blockade. The research team presented results from 23 wargames. The conclusion was that a Chinese maritime blockade of Taiwan could paralyze global trade and trigger a large-scale conflict in the Western Pacific.

Retired Marine Colonel Mark Cancian looked at the audience and said: "This would be a large-scale convoy battle reminiscent of Operation Pedestal in the Mediterranean in 1942, and ONS-5 in the North Atlantic in 1943." It was a warning that the largest naval battle since World War II could occur.

The Taiwan Strait is no longer a distant sea. It is the front line where the U.S.-China AI hegemony competition could physically collide. And on that front line, AI is emerging as a new form of deterrence.

The CSIS wargame series began in 2023. The first report, "The First Battle of the Next War," simulated a Chinese amphibious invasion of Taiwan. Across 24 games, in most scenarios, the combined U.S., Taiwanese, and Japanese forces repelled the landing. But the cost was devastating. Two U.S. aircraft carriers were sunk, and hundreds of fighter jets were shot down. Japanese bases were devastated by missile strikes.

The second report in December 2024 posed an even more uncomfortable question. What is the probability that nuclear weapons would be used? In 15 simulations, nuclear catastrophe occurred in three games. These were cases where Chinese forces, on the verge of defeat, decided to use nuclear weapons to prevent regime collapse.

The researchers' conclusion was bleak: "Complete victory is impossible. The United States must wage a high-intensity conventional war while simultaneously offering the adversary an honorable exit. Otherwise, nuclear catastrophe ensues."

The third report in July 2025 addressed the blockade scenario. Across 26 games, a Chinese maritime blockade could cut off Taiwan's energy supply and paralyze its economy. But if the United States and allies sent convoy escorts, a large-scale naval battle was inevitable. The winner of that battle was unclear.

The lesson these simulations show is singular: traditional military power alone cannot make deterrence work. If nuclear deterrence was based on the terror of "assured destruction," digital deterrence operates on a different principle. It plants the conviction that "your plan has already been detected."

The way AI becomes a deterrent in the Taiwan Strait operates on multiple layers.

The first is early warning. Preparing for a Taiwan invasion would require months of military mobilization by China. Ammunition stockpile increases, troop movements, spikes in communications volume, civilian ships being converted for military use. All of these signs leave data trails. If AI integrates and analyzes satellite imagery, communications intercepts, and open-source intelligence (OSINT), it can detect invasion preparations weeks to months in advance.

Palantir's Gotham system performs exactly this role. It fuses information streaming from thousands of data sources in real time to find patterns. "Fuel output from missile bases in Fujian Province was three times the usual amount last week." "Communications volume from the Eastern Theater Command increased 40% compared to last month." These subtle signals come together to form a picture. The United States can pre-position aircraft carriers or send warnings through diplomatic channels based on this information.

The second is the speed of the battlefield. In modern warfare, the kill chain is becoming ever faster. The time from target detection to strike has been reduced to minutes. In the Ukraine war, Palantir's system shortened this process from hours to minutes. The same principle applies in the Taiwan Strait. The moment Chinese landing craft approach the shore, AI analyzes satellite data, transmits coordinates, and Taiwanese missiles are launched.

This is why the U.S. Department of Defense increased the Maven Smart System contract ceiling with Palantir to $1.3 billion in May 2025. This system is deployed across five combatant commands, supporting operations in real time. Pacific Command, European Command, Central Command, Northern Command, and Transportation Command. In a Taiwan crisis, the Pacific Command commander would grasp the battlefield situation in near-real-time through this system.

The third is masses of autonomous weapons. The U.S. Department of Defense's "Replicator" initiative aims to deploy thousands of low-cost autonomous drones to counter China's numerical superiority. But humans cannot individually control thousands of drones. AI must become the brain of those drones. Palantir's AIP (Artificial Intelligence Platform) applies large language models (LLMs) to military operations, enabling command and control of autonomous weapons systems.

According to the Department of Defense annual report, China also invested in AI for various military applications in 2024, including unmanned systems, intelligence, surveillance, and reconnaissance (ISR) analysis, decision support, cyber operations, and information warfare. Civil-military fusion is broadening the supply lines for that investment. Both sides are strengthening their AI capabilities.

Here a paradox emerges.

The way AI strengthens deterrence simultaneously accelerates the speed of crises. When both sides gain "faster decision-making," both sides reduce the other's reaction time.

This is precisely the risk participants at the 2025 UN Military AI Dialogue pointed out: "AI tools increase the speed of interpreting complex information, but rapid decision-making cycles reduce human deliberation time and weaken commanders' comprehension. This raises the possibility of errors, escalation, and miscalculation in stressful crisis situations."

In August 2025, 25 international experts gathered at Syracuse University to conduct an unusual wargame. They designed not America's defense, but China's attack. It was an attempt to enter Beijing's strategic mindset. The results were uncomfortable. The insight was that the basic assumptions of U.S. military planning , large-scale amphibious operations and preemptive strikes on U.S. bases , might not align with Beijing's actual calculations. Some participants argued China had to act within ten years; others predicted that Xi Jinping, at 72, would seek to achieve unification before his death.

Taiwan's government is also preparing. In December 2024, Taiwan's Presidential Office conducted its first-ever tabletop wargame. Vice President Hsiao Bi-khim and National Security Council Secretary-General Wu Zhao-xie led the exercise, which tested the government's response capabilities in a scenario of rising military tensions with China. Civilian movements are underway as well. Taiwan's Mizo Games released a board game called "2045." Players become military commanders, secret agents, and civilian resistance fighters battling a hypothetical Chinese invasion. It raised NT$4 million (about $120,000) through crowdfunding.

AI as deterrence ultimately delivers a single message: "Your attack will not succeed." The moment China decides to invade Taiwan, it shows data proving that decision is wrong. What percentage of landing craft will be sunk before reaching the shore, how long it will take to achieve air superiority, how long supply lines can hold. AI can push all these calculations in the adversary's face.

But this is a double-edged sword. A Department of Defense report warns that China may use generative AI for information operations. In May and October 2024, China timed military exercises around Taiwan to combine official accounts with proxy accounts impersonating Taiwanese citizens, exaggerating PLA capabilities and spreading disinformation about U.S. willingness to defend Taiwan. Evidence that AI can serve not only as a tool of deterrence but also of deception.

Whether AI in the Taiwan Strait becomes a shield preserving peace or a spark igniting war remains unknown. The only certainty is this: an invisible war of data is already underway beneath those waters.

B. Digital Deterrence Theory

(1) Gaining Superiority in the AI Arms Race Prevents War

"If you want peace, have smarter software than your enemy."

Alex Karp said this at the 2024 Munich Security Conference. He was offering a modern reinterpretation of the Roman strategist Vegetius's maxim. The original was "If you want peace, prepare for war (Si vis pacem, para bellum)." In Karp's version, war preparation is not tanks and missiles but algorithms and data.

This is the core proposition of digital deterrence theory. The logic is that achieving overwhelming superiority in the AI arms race paradoxically prevents war from occurring.

Traditional nuclear deterrence was based on "Mutually Assured Destruction (MAD)." If you attack me, I will destroy you. Because both sides perish, nobody strikes first. This logic has prevented nuclear war for 70 years. But deterrence in the AI era operates on a different principle.

Digital deterrence is based on "information asymmetry." If one side can track all of the opponent's movements in real time, predict their intentions, and instantly execute optimal responses, the opponent will not even attempt an attack. They know the probability of victory approaches zero.

The case supporting this logic is the Ukraine war. Russia possessed thousands of tanks and overwhelming artillery power. But when the war began, that massive armored force was nothing more than large, expensive targets before Ukrainian drones and precision strikes. Palantir's system identified Russian forces' positions in real time and transmitted that information to Ukrainian artillery units. The results are what we witnessed.

The U.S. Department of Defense embraced this lesson. In August 2025, the U.S. Army signed a $10 billion Enterprise Software Agreement (ESA) with Palantir. It consolidated 75 individual software contracts into one. Daniel Moyer, director of the Army Contracting Command, explained: "It's not a bulk purchase, but rather an à la carte menu. You can pick and choose the software you need." This was a declaration that software had become a core national defense asset.

That same May, the Department of Defense increased the Maven Smart System contract ceiling with Palantir to $1.3 billion. According to officials, this was because "combatant commands have increased their use of MSS for commanding and controlling dynamic operations and activities in their respective theaters." Military commanders in the Pacific, Europe, and the Middle East had begun to rely on AI-based systems.

NATO is moving in the same direction. In August 2025, NATO's Joint Warfare Centre (JWC) integrated the Maven Smart System NATO. More than 100 participants joined the exercise. JWC Commander Major General Ruprecht von Butler said: "Our warfighters demand a faster, more agile, and more realistic training environment punctuated by technological innovation." MSS NATO is NATO's first AI-based battle command-and-control system, deployed across Allied Command Operations (ACO).

The logic behind all these moves is the same: lead in AI, and the enemy will not dare to provoke. Former Google Chairman Eric Schmidt has warned that the slope of AI development is so steep that even a six-month lead can create a permanent gap. If the democratic bloc wins the AI development race, it can perpetuate economic and military hegemony. Conversely, if China gets there first, digital totalitarianism will expand worldwide.

But this logic harbors dangerous traps.

First, if AI superiority is opaque to both sides, the adversary may become more dangerous. A 2024 RAND report warned that AI military systems could increase the risk of unintended conflict without human judgment. What AI provides is not prophecy but calculation. If that calculation is wrong, it becomes catastrophe.

Second, as the AI arms race accelerates, decision-making speed exceeds human cognitive capacity. If the time to determine whether an enemy missile has been launched and issue an intercept order shrinks to within seconds, humans may ultimately have to delegate trigger authority to AI. This gives rise to the danger of "Flash War." Just as algorithmic collisions in stock markets cause instantaneous crashes known as "flash crashes," an accidental data error could lead to an unwanted war.

Third, AI technology spreads rapidly. In November 2024, Reuters reported that Chinese military-affiliated researchers had developed military tools based on the open-source Llama model. The open-source ecosystem makes it difficult to monopolize deterrence. When one side locks the technology, the other enters through a different door.

Participants at the 2025 UN Military AI Dialogue pointed out this dilemma: "Arms race dynamics are an additional risk factor. Perceptions of strategic competition and the psychological drive to gain an edge over an adversary create pressure to rapidly deploy capabilities, and this can outpace assurance, testing, and governance processes."

Nevertheless, proponents of digital deterrence argue there is no alternative. If the West sits idle while China strengthens its AI capabilities, the balance of power collapses. When balance collapses, miscalculation occurs. Miscalculation invites war. Therefore, "Responsible Acceleration" is the only path. They argue that slowing development out of concern over AI risks amounts to "unilateral disarmament," ceding the initiative to adversaries who do not follow regulations.

Alex Karp says: "Our technology is sometimes used to kill people." An uncomfortable truth. But according to his logic, the possibility of that killing paradoxically prevents greater death , war. The cold conclusion of digital deterrence.

(2) The Future of Defense Determined by Software

In March 2024, the U.S. Army awarded Palantir a $178.4 million contract. It was to develop a next-generation targeting system called TITAN (Tactical Intelligence Targeting Access Node). This system connects data collected from high-altitude and space-based sensors to ground forces, supporting beyond-line-of-sight targeting. Bryant Cheung, Palantir's senior vice president of defense solutions, said: "TITAN greatly improves capabilities, but we want software and AI to reduce complexity."

That sentence captures the essence of future defense capability. Hardware provides the ability, but software makes that ability real.

Past arms races were numbers games about who had more tanks and fighter jets. In World War II, the Soviet Union produced 80,000 T-34 tanks to overwhelm Germany's armored forces. During the Cold War, the U.S. and the Soviet Union stockpiled tens of thousands of nuclear warheads to maintain a balance of terror. But on the 21st-century battlefield, that equation has broken down.

The lesson the Ukraine war revealed is clear. Russia's massive armored forces were helpless before low-cost drones and precision-guided weapons. The era has arrived when a $100 million fighter jet can be suppressed by a swarm of $5,000 drones. Palantir CTO Shyam Sankar expressed this by saying, "Hardware has become the accessory of software."

This change is fundamentally reshaping the defense industry's structure. In the past, giant manufacturers like Lockheed Martin, Boeing, and Raytheon were the masters of defense. They built aircraft carriers and fighters over decades, and the government paid the costs. It was a practice called "Cost-Plus" contracting. When the manufacturer submitted costs, the government paid those costs plus a margin. Even if costs exceeded estimates, it did not matter. The government bore the additional expense.

But now engineers who write code have taken center stage in defense. Palantir filed suit against the U.S. Army in 2016 and won. It opened the door for commercial software to participate in defense procurement. Since then, "defense tech" companies like Palantir, Anduril, and SpaceX have begun encroaching on traditional defense companies' territory.

The $10 billion contract in August 2025 was the pinnacle of this shift. The U.S. Army consolidated 75 individual software contracts into one. In 60 of those contracts, Palantir had been a subcontractor, but now it became the prime. Contracting Command Director Daniel Moyer said: "This is our big initiative. It's not just about the Palantir contract. We looked at how many times the same commercial software was purchased across the Army and whether it makes sense to bundle it all for maximum discounts."

The advantages of software-centric defense are multiple. First, upgrade speed. Upgrading hardware requires taking equipment to a factory and replacing parts. It takes months or years. But software patches can be deployed remotely. Just as Tesla updates its cars, tanks and missiles can become smarter overnight.

Second, interoperability. JADC2 (Joint All-Domain Command and Control) is an initiative to link land, sea, air, space, and cyber into a single network. Palantir's software serves as the operating system of this network. Different weapon systems share the same data and communicate in the same language. Interoperability between allied nations works the same way. This is why NATO adopted the Maven Smart System.

Third, cost efficiency. Traditional defense programs are notorious for cost overruns and schedule delays. The F-35 fighter program has exceeded $1 trillion. But software-based systems can be developed faster and cheaper. Palantir's TITAN prototype was delivered within months.

But as software determines defense capability, new vulnerabilities also emerge: cyberattacks. If code, models, and data are hacked or manipulated, the physical equipment itself is rendered useless. In January 2025, CISA, DARPA, the Office of the Under Secretary of Defense for Research and Engineering (OUSD R&E), and the NSA issued a joint report identifying the "software understanding gap." The situation where AI systems are so complex that humans cannot understand how they operate. If you do not understand it, you cannot find the vulnerabilities either.

Another risk is dependence. Excessive reliance on a specific country's or company's proprietary software and cloud deepens technological and political subordination. This is the issue of digital sovereignty. It is one of the reasons Europe is wary of Palantir.

Nevertheless, the trend cannot be reversed. In December 2025, U.S. Secretary of the Navy John Phelan and Palantir CEO Alex Karp unveiled "ShipOS," a software platform for ship operations. That same year, NATO first operationally tested the Maven Smart System in its largest-ever allied exercise, "Steadfast Dart 2025." Exercise planners said: "Maven will improve integration of cross-domain data sources, improve decision-making capability, and drive digital transformation. We will get better and faster."

The future of defense capability is no longer measured by "who has the bigger ships." It is measured by "who can run a faster OODA loop (Observe-Orient-Decide-Act)" and "who can clear the fog of war with more sophisticated algorithms." In a future where software determines defense capability, code is the weapon, and data is the ammunition.

The era of Pax Technica is approaching. If Rome's peace came from the mobility of Roman roads, and Britain's peace came from naval supremacy, then a technological peace comes from superiority in data and algorithms. Palantir positions itself as the guardian of that peace. Their logic is clear: only if the Western democratic bloc secures AI , the most powerful spear and shield , can it deter provocations from authoritarian states and maintain world order.

Whether this is an uncomfortable truth or a necessary reality is for the reader to judge. One thing is certain: 21st-century power is etched in code on silicon wafers.