AI Demand Ignites Data Center Boom, Open Models, and Global Tensions

As Chinese President Xi Jinping stood beside Vladimir Putin and Kim Jong-un at a September military parade in Beijing, the world watched autonomous drones streak across the sky in formation with fighter jets. The display was more than spectacle. It signaled a new phase in an accelerating global artificial intelligence arms race that now draws direct comparisons to the dawn of the nuclear age. Pentagon officials, alarmed by what they saw as China’s lead in unmanned combat systems and Russia’s drone production capacity, pressed American defense contractors to respond. Last month, Anduril Industries began production of its AI-backed Fury autonomous air vehicle at a new factory outside Columbus, Ohio, three months ahead of schedule.

This is not abstract technological competition. It is a contest over lethal autonomous weapons that can select and engage targets with minimal human oversight. Sheera Frenkel, Paul Mozur and Adam Satariano reported in The New York Times that the buildup has prompted urgent reassessments inside the U.S. defense and intelligence apparatus. The risks are obvious: once multiple states deploy swarms of AI-enabled killing machines, escalation becomes faster, miscalculation more likely, and accountability harder to assign. Critics have warned for years that the world is sleepwalking into an arms race without meaningful guardrails, much as it did with nuclear weapons in the 1940s and 1950s. Those warnings now appear prescient.

At the same time, the civilian AI sector is undergoing its own quiet but significant shift. In recent weeks Google, Microsoft, Alibaba and Nvidia have released new open-weights models that analysts say have crossed an important threshold. No longer mere research curiosities, models such as Qwen 3.5, Gemma 4, and Microsoft’s specialized MAI systems are being described as credible enterprise platforms. Andrew Buss, senior research director at IDC, told The Register that the industry has moved “from interesting to now serious enterprise platforms.”

This development highlights a growing divide. Frontier models from OpenAI, Anthropic and Google’s most advanced systems remain extraordinarily expensive to run and require users to feed potentially sensitive corporate data into external APIs. Enterprises are increasingly unwilling to take that risk. The same companies promising that enterprise data will not be used for training have faced repeated copyright lawsuits and public scandals over data handling. For corporations guarding intellectual property, customer information or trade secrets, the privacy calculus is straightforward: why hand your most valuable data to organizations with a documented record of pushing legal and ethical boundaries?

Open-weights models offer a partial alternative. Organizations can run them on their own infrastructure, customize them for specific tasks, and avoid constant data exfiltration to Silicon Valley or Chinese cloud providers. The Register notes that this split, between massive “everything to everyone” frontier systems and smaller, more specialized open models, reflects a maturing market. Yet even these open models largely come from the same dominant players, Google, Microsoft, Alibaba, raising questions about how genuinely open the ecosystem truly is.

The infrastructure demands of this dual military-civilian AI surge are enormous. Training and running advanced models requires vast data centers that consume staggering amounts of electricity and water. That reality has triggered a fierce political backlash. Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez have called for a federal moratorium on new data-center construction, arguing that Congress has a “moral obligation” to address the existential risks AI poses to society, from labor displacement to unchecked corporate power. Maine’s Democratic-led House has already voted for its own moratorium. Progressives argue that humanity should not sacrifice democratic oversight, environmental stability, or economic justice at the altar of technological acceleration.

Free-market advocates counter that such restrictions are modern Luddism. Writing in National Review, Andrew Follett argues that data centers will ultimately benefit American families and businesses, and that Republican-led states should welcome them with lighter regulation. The debate echoes older fights over automation: whether machines liberate human potential or simply concentrate wealth and power. Libertarian voices at the Cato Institute insist that, just as tractors and computers did not impoverish society, AI will free workers for higher-value tasks. Yet this optimism rarely accounts for the concentrated market power of the firms building these systems or the geopolitical incentives pushing autonomous weapons forward.

What emerges is a picture of AI development increasingly detached from meaningful public control. Military planners race to match Chinese capabilities. Tech giants release ever-more-capable models while fighting copyright claims and privacy concerns. Data-center construction becomes a proxy battle between those who see existential danger and those who see limitless profit. The open-weights trend may give some enterprises more sovereignty over their data, but it does not resolve the deeper questions of accountability in an era of lethal autonomous systems and unprecedented computing demands.

As the AI arms race intensifies, the window for deliberate, democratic governance narrows. History shows that technological races framed as existential competitions rarely pause for ethical reflection. The question now is whether governments, particularly in the United States, will treat AI’s military and commercial expansion with the gravity it demands, or whether profit, power, and panic will dictate the terms. The drones flying in formation over Beijing and the data centers sprouting across the American heartland suggest the latter trajectory is currently winning.