The AI Arms Race Scoreboard: USA vs China vs Russia

Three nations are competing to control the future of machine warfare. The United States holds the largest budget, the most mature private sector, and the most combat-tested autonomous systems. China holds the most ambitious doctrine, the fastest institutional mobilization, and a manufacturing base that dwarfs all rivals. Russia holds a declining hand — but one that includes nuclear-armed autonomous platforms and lessons written in Ukrainian blood.

Scores based on AW Intelligence assessment across five dimensions: budget and investment, fielded autonomous systems, doctrine maturity, private sector ecosystem strength, and combat-proven operational experience. Methodology detailed below.

Scoring Methodology

Assessing relative AI military capability across three powers requires a framework that accounts for qualitatively different strengths. Raw budget figures favor the United States but obscure China's purchasing power advantage. Fielded systems counts favor China in certain drone categories but miss the capability-depth of U.S. systems. For this assessment, we evaluate five dimensions, weighted equally at 20 points each:

United States: The Budget Leader

The U.S. Department of Defense allocated over $75 billion to AI and autonomous systems programs across the fiscal years 2023 through 2025, according to DoD budget justification documents and congressional testimony. This figure encompasses direct AI research and development, autonomous systems procurement, and the software and data infrastructure programs that underpin military AI. No other nation approaches this investment level in absolute terms.

The institutional architecture for U.S. military AI is concentrated in the Chief Digital and Artificial Intelligence Office (CDAO), established in 2022 by consolidating the Joint AI Center and the Defense Digital Service. CDAO is responsible for the Maven Smart System — the successor to Project Maven, the AI-assisted drone footage analysis program that sparked controversy at Google in 2018 — and for driving AI adoption across the military services. The office has authority to direct AI integration across combatant commands, a mandate that previous institutional structures lacked.

The Replicator Initiative

In August 2023, Deputy Secretary of Defense Kathleen Hicks announced the Replicator Initiative, a two-year program designed to field thousands of small autonomous systems across multiple domains — air, sea, and ground — by August 2025. The program was an explicit response to the lesson drawn from the Ukraine conflict and from Chinese military modernization: that mass matters, and that the future of mass military capability is cheap, autonomous, and attritable rather than expensive, crewed, and irreplaceable.

Replicator's specific program details have been classified, but public reporting confirmed that it includes multiple drone platform variants, autonomous undersea vehicles, and ground-based robotic systems. Anduril, Joby, L3Harris, and Shield AI were among the contractors publicly associated with Replicator-adjacent programs. The initiative represented a significant policy signal: that the Pentagon had concluded it needed to shift procurement philosophy toward volume and autonomy rather than toward the exquisite, crewed platforms that have dominated defense budgets for decades.

DARPA Programs and the Collaborative Combat Aircraft

The Defense Advanced Research Projects Agency has been seeding the AI weapons technology pipeline since well before the current public awareness of the issue. DARPA's Air Combat Evolution (ACE) program demonstrated AI dogfighting in simulation — including the famous 2020 AlphaDogfight Trials in which an AI system defeated an F-16 pilot 5-0 — and is now transitioning those capabilities toward operational platforms. The Collaborative Combat Aircraft (CCA) program, a U.S. Air Force initiative to field autonomous wingman aircraft alongside the F-22 and F-35, represents the operationalization of ACE research. Anduril and General Atomics won the CCA development contracts in 2024; Boeing and Northrop Grumman were eliminated from competition.

The private sector ecosystem around U.S. military AI is unmatched globally. Anduril ($30.5B valuation), Palantir ($160B+ market cap), Shield AI ($5.6B valuation), Joby Aviation, Joby Defense, L3Harris, and dozens of smaller specialized firms constitute an innovation layer that no other nation has replicated. China has civilian AI firms of comparable scale, but the integration between commercial AI development and military procurement in the U.S. has accelerated dramatically since 2020.

China: The Doctrine Leader

China's approach to AI military competition is distinguished not primarily by its current capabilities but by the coherence and ambition of its conceptual framework. The People's Liberation Army has been developing a theoretical basis for what it calls "intelligentized warfare" (智能化战争) since at least 2017, when it was enshrined in official PLA doctrine documents. The concept encompasses AI-enabled command and control, autonomous weapons systems, cognitive electronic warfare, and algorithmic decision-making at strategic, operational, and tactical levels simultaneously.

This doctrinal sophistication is not accidental. China's Military-Civil Fusion (MCF) strategy — formalized under Xi Jinping and implemented through the State Administration for Science, Technology, and Industry for National Defense (SASTIND) — mandates that civilian technology development serve military requirements. Chinese AI companies including Huawei, Baidu, and DJI operate under legal obligations that require cooperation with military and intelligence agencies upon request. DJI, the world's dominant commercial drone manufacturer with an estimated 70 percent global market share, represents both a civilian commercial success and a strategic military asset that provides China with an unparalleled drone manufacturing and R&D base.

PLA Strategic Support Force and AI Infrastructure

The PLA Strategic Support Force (SSF), established in 2015 and reorganized into the Information Support Force (ISF) in 2024, is the primary organizational structure for Chinese military AI and information warfare capabilities. It consolidates cyber, space, and electronic warfare functions with information infrastructure — creating the command architecture for what Chinese doctrine describes as "system confrontation," the competition between integrated combat systems rather than between individual platforms.

Chinese military AI spending is difficult to verify independently due to the opacity of PLA budget reporting. Estimates from the Center for Strategic and International Studies, the RAND Corporation, and the U.S.-China Economic and Security Review Commission place PLA AI military investment at $15 billion or more annually, when including both direct weapons AI programs and the military-relevant portions of the civilian MCF technology base. The purchasing power parity adjustment suggests this investment goes further in China than equivalent dollar amounts in the United States.

Autonomous Submarine Swarms and Maritime AI

China's most operationally significant AI weapons advances are in the maritime domain. The PLA Navy has invested heavily in autonomous undersea vehicles and unmanned surface vessels, reflecting the strategic centrality of Taiwan Strait and South China Sea operations to Chinese military planning. The HAISHEN-1 autonomous underwater vehicle, developed by Harbin Engineering University under PLA Navy funding, represents one publicly acknowledged program in a broader portfolio of autonomous maritime systems. Chinese research institutions have published extensively on drone swarm coordination algorithms for maritime applications — research that is foundational to both offensive and defensive undersea capabilities.

The DF-ZF and Hypersonic AI Guidance

The DF-ZF hypersonic glide vehicle, which reached initial operational capability with the PLA Rocket Force, represents China's most advanced integration of AI with strategic weapons. The vehicle's maneuvering flight profile — which enables it to evade traditional missile defense intercept geometries — relies on onboard guidance algorithms that must make real-time trajectory adjustments faster than human operators could intervene. This is not AI in the command-and-control sense but AI in the guidance-and-navigation sense: the weapon's lethality depends directly on the quality of its autonomous decision-making about flight path.

Russia: The Declining Contender

Russia entered the AI arms race with genuine ambitions and some real capabilities. It exits the 2020s in a significantly weakened position, for reasons that are structural, economic, and directly related to the ongoing war in Ukraine. The combination of international sanctions, brain drain of technical talent, and the catastrophic attrition of its conventional military forces has compressed Russia's AI weapons development timeline in ways that its leadership has been unwilling to publicly acknowledge.

Uran-9: The Ground Robot That Failed

The Uran-9 unmanned ground combat vehicle was Russia's most prominently advertised AI-enabled robotic weapons system before Ukraine. Armed with a 30mm autocannon, anti-tank missiles, and flamethrowers, it was deployed to Syria in 2018 in what Russian military media described as a successful combat evaluation. Internal Russian military analysis, leaked to defense researchers and published in the Journal of Slavic Military Studies in 2018, told a different story: the Uran-9 experienced repeated communication failures, target identification errors, and unreliable weapon system performance during the Syria deployment. The system's control range was insufficient for urban combat environments; its autonomous features functioned poorly in the complex electromagnetic environment of a contested battlefield.

The Uran-9 experience illustrated a structural problem in Russian military AI development: a gap between development-phase demonstration and operational deployment. Russian defense firms, including the state-owned Kalashnikov Concern and Rostec subsidiaries, have continued developing autonomous systems — the Marker unmanned ground vehicle, the Platforma-M, and several others — but none has achieved the combination of reliability and scale that would constitute a genuine military capability.

Poseidon: The Autonomous Nuclear Torpedo

Russia's most consequential AI-adjacent weapons system is also its most extreme: the Status-6 Poseidon autonomous torpedo, an intercontinental nuclear-armed underwater vehicle reportedly capable of crossing ocean basins without surface support, locating coastal targets, and detonating a nuclear warhead of up to 100 megatons. The system was revealed in a deliberate leak of classified briefing slides in 2015 — a move widely interpreted as a strategic signaling operation rather than an inadvertent disclosure.

Poseidon represents a different conceptual approach to autonomous weapons than the U.S. or Chinese models. Rather than AI as a means to increase conventional effectiveness, Poseidon uses autonomous navigation to extend nuclear deterrence to a category of target — coastal cities and naval infrastructure — that Russia believes its current ballistic missile and submarine capabilities cannot credibly threaten in a full nuclear exchange. The system's autonomy is not about decision-making under fire; it is about operating at ranges and depths that preclude real-time human control, achieving what Russian strategists call "inevitable retaliation."

S-500 and AI-Assisted Air Defense

The S-500 Prometheus air defense system, entering limited service with Russian Aerospace Forces from 2021 onward, incorporates AI-assisted target classification and engagement sequencing. The system is designed to engage hypersonic targets — including glide vehicles like the DF-ZF — at ranges and altitudes beyond the S-400's envelope. Its AI component handles the target tracking and kill chain timing that is too fast for human operators at hypersonic engagement velocities. The S-500 represents Russia's most credible near-term AI military capability: not an offensive autonomous weapons system, but an AI-enabled defensive one that addresses a genuine strategic need.

The Ukraine Brain Drain

The single most underreported factor in Russia's AI weapons decline is the emigration of its technical talent since February 2022. Russian technology and AI researchers have left the country in large numbers — estimates from Russian sources suggest hundreds of thousands of technical and IT workers departed in 2022 alone, with continued emigration through 2023 and 2024. The destinations include Georgia, Armenia, Kazakhstan, Serbia, and a significant contingent in Israel, Germany, and the United States. The talent loss is not offset by the researchers who remained; it is compounded by the near-total severance of Russian institutions from the international research community through sanctions and voluntary withdrawal from collaboration.

Russia's Shahed-series drone program — developed with Iranian technology and deployed extensively against Ukrainian infrastructure — represents a workaround: relatively cheap, commercially available autopilot technology integrated into a weapons platform. The "AI" in these systems is basic target-seeking and navigation software, not the machine learning-based autonomy that characterizes leading U.S. or Chinese programs. They are effective as volume-production attrition weapons, but they do not represent a frontier AI weapons capability.

Gap Analysis: Where the Race Stands

The United States leads in total investment, private sector ecosystem depth, and the maturity of its AI-to-procurement pipeline. The gap between U.S. private AI capability and its conversion into deployed military systems, however, remains a genuine weakness — the Pentagon's acquisition process still imposes timelines and compliance requirements that slow the deployment of commercial AI advances.

China leads in doctrine maturity and in the coherence of its military-civil AI integration strategy. The MCF framework, whatever its limitations, means that advances in civilian AI do not remain siloed from military applications in the way they sometimes do in Western defense procurement. China also has a decisive advantage in volume manufacturing: the ability to produce autonomous systems at scale, at low cost, is a military capability in itself, and DJI's commercial drone ecosystem provides a foundation that no Western country can match.

China trails the United States in combat-proven experience. The PLA has not fought a major conflict since the 1979 Sino-Vietnamese War. Its autonomous systems have not been tested under the stress of active combat, electronic warfare jamming, and adversary countermeasures. The U.S. military has operated autonomous systems in contested environments in Syria, Iraq, Somalia, and Yemen. That operational experience — however limited — provides learning that no amount of simulation fully replicates.

Russia's position is more complex than its score suggests. It has genuine nuclear-autonomous deterrence capability in Poseidon; it has real AI-assisted air defense in S-500; and it has acquired combat experience in autonomous systems deployment in Syria and Ukraine that no Western military has yet matched in scale and duration. But its conventional AI weapons programs are behind schedule, underfunded by sanctions, and depleted of the technical talent that would accelerate them. The trajectory is downward.

Wildcard Factors

Ukraine as the Global AI Weapons Testing Ground

The Ukraine conflict has functioned as an inadvertent laboratory for autonomous weapons development that will shape every major military's programs for the next decade. FPV drone warfare, AI-assisted targeting, autonomous navigation in GPS-denied environments, counter-drone electronic warfare, and loitering munition tactics have all been tested, iterated, and counter-adapted in real combat at scale. The lessons are not restricted to the belligerents: every defense ministry with intelligence collection capabilities is analyzing Ukrainian operational data. China's military analysts, in particular, are drawing conclusions about drone swarm tactics, autonomous logistics, and counter-drone effectiveness that will directly inform PLA procurement.

Export Market Dynamics

The global autonomous weapons export market is reshaping alignment relationships in ways that budgets and doctrines do not capture. China has become the dominant supplier of commercial drone technology globally, and that commercial position is a strategic asset — nations that rely on DJI infrastructure for commercial purposes are more likely to procure Chinese military autonomous systems as relationships develop. Turkey's Bayraktar TB2, an autonomous-capable armed drone with AI-assisted targeting that achieved decisive results in the Nagorno-Karabakh conflict in 2020 and in Ukraine, has created a new tier of middle-power autonomous weapons capability that neither the U.S., China, nor Russia fully controls. The proliferation of capable autonomous weapons below the great-power tier is an arms race within the arms race — and its dynamics may prove as consequential as the three-power competition at the top.

The Trajectory to 2030

The most significant unknown in this competition is whether the United States can translate its private sector AI advantage into deployed military capability faster than China can translate its doctrinal and manufacturing advantages into operational systems. The history of major military competitions suggests that the nation with the most coherent doctrine and the fastest procurement-to-deployment cycle tends to prevail over the nation with the largest absolute investment but slower conversion speed.

China's trajectory toward parity — and in certain specific domains, superiority — in AI weapons capability is a plausible scenario by 2030. The autonomous submarine swarm domain, where China has invested heavily and where detection and countermeasure technology is immature, is the area where a decisive Chinese advantage is most plausible. The AI-enabled command and control domain, where U.S. systems have greater maturity but Chinese systems are advancing rapidly under the MCF framework, is the arena where the competition will be decided.

Russia's role in this competition is increasingly that of a spoiler rather than a peer competitor — capable of strategic disruption through nuclear-autonomous deterrence and information operations, but unable to sustain the conventional AI weapons competition that the United States and China are running at pace. The question is not whether Russia will fall further behind; it is whether its spoiler capabilities, particularly in the autonomous nuclear domain, constrain U.S. and Chinese operational planning in ways that reshape the competition's terms.

The AI arms race has no agreed rules, no verification mechanism, and no international treaty framework capable of governing its most dangerous applications. What it has is accelerating technological capability, increasing deployment, and the certainty that the systems being built today will be used in the next major conflict, wherever and whenever it occurs.