By Branka Marijan
Published in The Ploughshares Monitor Spring 2025
For years, the contest to be dominant in military applications of artificial intelligence (AI) has been framed as a race between the United States and China. The prevailing wisdom among analysts tracking the development of autonomous and intelligent military systems held that the United States enjoyed a clear lead, with China trailing by several years. This conclusion was in part due to US investments and the significant breakthroughs in AI that had come from American technology firms. This head start was expected to translate into a decisive advantage for the Pentagon in applying AI to defence and warfare.
However, a seismic shift occurred this past January, when DeepSeek AI, a Chinese startup, released its R1 generative AI model, claiming it rivaled OpenAI’s ChatGPT but at a fraction of the cost. The Chinese company asserted that it had developed this model for a mere $6 million, utilizing just 2,000 chips (semiconductors) — an astonishing claim when compared to the 16,000 chips reportedly used by OpenAI for its latest model.
A closer examination of these figures revealed that the quoted cost reflected only the marginal expenses of training the model, omitting broader financial considerations such as the initial procurement of high-performance chips, infrastructure for data centres, and the salaries of a highly skilled research team.
Nevertheless, the news that a Chinese startup had managed to develop a generative AI product approaching the sophistication of OpenAI’s offerings while seemingly more efficient sent shockwaves through both the defence community and financial markets. Investors reacted swiftly in a sell-off that erased nearly $1 trillion in market value across the AI sector. More significantly, the revelation forced a fundamental reassessment of China’s standing in AI research and development. The notion that Beijing was years behind Washington in cutting-edge AI innovation suddenly seemed outdated.
Can China lead in AI development?
The key question now: can China surpass the United States in AI development, or will it remain a fast follower?
Beijing’s AI push has been methodical and well funded. The Chinese government has invested billions in AI research, aligning its technological ambitions with national security priorities. Both state-backed entities and private firms have benefited from a concerted strategy to close the gap with the West by acquiring advanced semiconductor manufacturing capabilities, expanding computational resources, and accelerating AI talent development. Moreover, a new generation of innovators or “disruptors,” such as DeepSeek founder Liang Wengfeng, seems to be emerging. As Charles Mok notes in a commentary piece published on the Stanford Cyber Policy Centre website, Wengfeng has claimed that China needs to innovate more, rather than copying and following other countries.
If AI competition continues unchecked, the world risks entering an era in which autonomous weapons become as ubiquitous as conventional arms but with no clear legal frameworks to govern their use. Without international agreements to establish standards, the likelihood of accidental escalation or unintended conflict due to AI misjudgments increases dramatically.
But China still faces obstacles. The United States holds a critical advantage in semiconductor technology, while China faces supply chain vulnerabilities from export restrictions on advanced AI chips, such as Nvidia’s A100 and H100. These chips are indispensable for training large-scale AI models.
While China is rapidly developing alternatives, it is not clear that domestically produced chips can match the performance of their US counterparts in the near term. Although China has made much of the homegrown innovation exemplified by DeepSeek, tech researchers Tye Graham and Peter W. Singer note that DeepSeek is using Nvidia H800 chips, which are not subject to US trade restrictions. Moreover, DeepSeek had access to Nvidia A100 chips that were purchased before restrictions were put in place. Thus, the reality is that recently revealed advances had been made with sophisticated chips, some of which are no longer accessible.
As well, China’s AI ecosystem operates within a regulatory framework that, while supportive of state objectives, also imposes constraints. While research in the United States thrives on openness and collaboration, China’s AI industry is subject to tighter government oversight, particularly in politically sensitive domains, which can limit engagement by Chinese researchers with global AI communities and fundamental research.
Nonetheless, DeepSeek AI’s achievement might be seen to exemplify the age-old adage that “necessity is the mother of invention” — or innovation. If Chinese firms continue to develop sophisticated AI models at lower costs, they could gain a significant advantage, particularly in military applications in which cost efficiency and rapid deployment matter as much as raw capability.
The militarization of AI and its consequences
The DeepSeek AI breakthrough raises urgent questions about the trajectory of military use of AI. As AI systems become increasingly sophisticated, their role in warfare is growing, in applications that span logistics optimization and support in broader decision-making and targeting. Indeed, Chinese military analysts have claimed that DeepSeek’s AI models can already perform many military functions.
The implications of this technological acceleration are profound.
The global AI arms race could become more intense. Washington and Beijing have long recognized AI as a strategic frontier, but the realization that China is closer to parity than previously thought may prompt the United States to double down on its AI investments.
The Pentagon has already established partnerships with private tech firms to integrate AI into defence systems. Recent policy shifts indicate a growing willingness to streamline AI adoption within the military. The Trump administration, on the second day after taking office, announced that the budget of Stargate, an AI infrastructure project, is expected to reach $500 billion before the end of the current Trump presidency. Investment is focused on building data centres that will then be used by OpenAI for a variety of research and development purposes. Another announcement shed more light on the type of work, including a partnership with the US National Laboratories on scientific and nuclear security.
Google recently dropped its promise not to use AI for weapons and surveillance. Company leadership expressed the view that “democracies should lead in AI development.” What appears to be happening in the United States, pushed by tech companies, is a mirroring of the military-civil fusion strategy that China is pursuing. If China’s AI firms continue to advance, the pace of US military AI development could quicken in response, setting off a feedback loop of heightened competition.
Other countries are also making major investments in AI technologies for defence. If they also make important strides forward, the world will see a greater use and testing of these technologies, likely without much disclosure. As retired Australian Army Major General Mick Ryan notes in his post on X, “the fact that DeepSeek-R1 is such a surprise proves again that humans will continue to surprise each other despite the presence of advanced technology.” In practice, many states see revelations about DeepSeek as the signal to double down on military AI investments, adding loops to the Great Power competition.
This escalation carries risks. AI-enabled weapons systems introduce new dimensions of unpredictability to conflict scenarios. Autonomous drones and missile systems, for example, could reduce human decision-making in warfare, raising ethical and legal concerns about accountability. An AI-driven escalation, in which autonomous and intelligent systems misinterpret intent and trigger conflicts, adds another layer of instability to an already volatile geopolitical environment.
Governing military AI: A diplomatic minefield
Rapid advancements in AI-driven military technologies have reignited debates over AI governance. In the last few years, there have been more efforts to establish international norms for responsible military AI development, but little consensus among major powers. International forums, such as United Nations gatherings on lethal autonomous weapons systems, have made limited headway in crafting binding regulations.
It is not clear if existing processes, including the Responsible Military AI summits and the political declaration on responsible military AI and autonomy led by the United States, have the backing of the new US administration. Given the apparent instability of alliances in the Euro-Atlantic relationship, the path forward remains uncertain.
If AI competition continues unchecked, the world risks entering an era in which autonomous weapons become as ubiquitous as conventional arms but with no clear legal frameworks to govern their use. Without international agreements to establish standards, the likelihood of accidental escalation or unintended conflict due to AI misjudgments increases dramatically.
What comes next?
The revelation of DeepSeek AI’s capabilities reminds us that technological superiority is neither static nor guaranteed. While the United States still enjoys key advantages in AI infrastructure and semiconductor manufacturing, China’s ability to develop competitive AI systems at lower costs should not be underestimated.
For policymakers, the imperative is clear: the AI arms race must be met by robust diplomatic efforts to mitigate the risks of unfettered militarization. Discussions on AI governance need to evolve from broad guidelines to concrete and enforceable agreements. Achieving this end will require engagement not just between Washington and Beijing but among a broader coalition of nations that want to prevent AI from becoming an unregulated force multiplier in military conflicts.
The era of AI-powered warfare is no longer a distant prospect. It has arrived. Whether it leads to greater stability or deeper insecurity will depend on how states navigate the challenges of AI governance, technological competition, and military strategy.
