On Tuesday, Vice President JD Vance declared to the Paris AI summit that he would work to “ensure that the most powerful AI systems are built in the US with American-designed and manufactured chips.” It was just the latest indication that the Trump White House views AI as a crucial new arena of geopolitical competition, especially with China. In the background of his speech was the recent launch of the Chinese AI model DeepSeek-V3, widely framed as a stunning blow to American technological dominance. DeepSeek’s groundbreaking efficiency, which promised to make AI more accessible and scalable, wiped billions off the value of US tech stocks. It was AI’s “Sputnik moment.” Tech investors panicked, and industry gurus tried to dismiss the breakthrough as “hype.”
But the real hype is the talk of a new AI Cold War. It’s just the latest distraction from what’s really at stake: unleashing the full potential of the industrial revolution. The real crisis is not which country, or company, dominates AI. It is that AI, like so many technological leaps before it, is being held back by economic structures that keep people toiling while tech giants try to figure out how to make a profit, and eventually corner the market.
Since OpenAI released ChatGPT in 2022, tech companies have poured billions into data centers, propelling the stock price of Nvidia—the top designer of chips used for AI—skyward. China has been accused of profiting from American investments, bypassing export controls to train its model on Nvidia chips. OpenAI is in talks with US government officials about a potential breach that allowed DeepSeek developers to access large amounts of OpenAI data. Their suspicions are probably well-founded, as DeepSeek seems to confuse itself with ChatGPT on occasion.
In the end, all of this amounts to schoolyard bullies squabbling over lunch money. AI is not a question of national security; it is a much older question of technological innovation held hostage by the drive for profit.
Already in the 19th century, political economists were troubled by a strange paradox: While the expansion of productive capacity undoubtedly made society richer, it seemed everywhere accompanied by poverty and toil. Observing this, John Stuart Mill wondered whether, “all the mechanical inventions yet made have lightened the day’s toil of any human being.” He concluded that while “they have enabled a greater population to live the same life of drudgery and imprisonment,” they have enabled a great “number of manufacturers and others to make fortunes.”
Likewise, David Ricardo remarked, “No one could suffer in consequence of abundance, but as long as society is constituted as it now is, abundance will often be injurious to producers, and scarcity beneficial to them.” Ricardo recognized that industrial technology could create incredible prosperity, but only if society adapted its economic structure to match. If it didn’t, workers would be tossed out of the process while wealth concentrated in fewer and fewer hands.
AI presents a stark reminder of this problem: It could vastly increase production, streamline distribution, free us from monotonous work, and rapidly improve living standards. Instead, it’s being throttled by corporate monopolies, made artificially scarce by paywalls, and treated as a national-security risk. Its true potential is lost as it becomes one more tool in the surveillance arsenal, used to make us work harder and compete with machines for jobs.
Sure, society might find new ways to employ us, and one can only hope it does. But we will only find ourselves lucky enough to be working harder and longer once more, like our 19th-century forebears, as technology spurs our drudgery. By that century’s end, governments were already shifting toward protectionist policies aimed at maintaining their market share. Extreme wealth became not a boon for humanity, but an “oversupply problem.” Today, as noted by scholars Michael Pettis and Matthew Klein, politics is still characterized by the “perverse coincidence” of having plenty of idle resources while many material needs remain unmet.
Already in the age of Mill and Ricardo, technological advancements were enabling human beings to step outside the production process. Little by little, we find ourselves not simply appendages of machines, but their “watchmen and regulators.” In theory, AI enables us to harness society’s general intellect and make things run with less and less intervention.
One might assume that if this process continued indefinitely, we would soon be living in a Jetsons world. But while we changed how we produced, we have maintained the old “measuring rod” of labor-time—in a world in which human labor seems to matter less and less. So freedom from toil becomes “unemployment” and the risk of being disposed of by society.
The Large Language Models (LLMs) produced by AI companies are only the latest iteration of this centuries-long struggle. That DeepSeek has thought it is ChatGPT is unsurprising, because the models all rely on the same material. They are more powerful the more fruits of society—our discussions, literature, science, philosophy—they can gobble up. At the risk of oversimplifying, the “intelligence” we respond to when chatting with these models is in essence our own history, made to speak through one large mathematical equation. AI is our collective intellect, repackaged as proprietary software. But it is getting harder and harder to contain it.
“The AI arms race isn’t about technological progress.”
Almost as soon as DeepSeek launched, Alibaba released a superior model and two American universities followed suit, replicating DeepSeek at astonishingly low costs. AI isn’t scarce, and the techniques used to create and train models are well-known. But those who can best streamline semiconductor supply chains and dominate inputs like rare-earth minerals will gain the upper hand. The AI arms race isn’t about technological progress, but about who will control it and reap its benefits—benefits that scale with the amount of human knowledge the AI models incorporate.
Instead of heralding this as a step toward a post-scarcity world, we are watching as the United States tries to beat out China’s world-leading production in rare-earth minerals with land grabs for Greenland and a proposed Ukraine-Russia peace plan that secures access to Ukraine’s mineral deposits. The AI revolution is unfolding under the same monopolistic logic that has held progress in check for so long: control resources, limit supply, extract the profits.
Indeed, while financial analysts emphasized the cost-cutting benefit of DeepSeek, it’s likely to be used for an ever greater range of applications as tech giants seek to justify their front-loaded capital expenditures. If anything, AI has only advanced labor discipline. It has reduced menial work, but filled that time with greater expectations for productivity.
DeepSeek CEO Liang Wengfeng, might claim that his company doesn’t seek “excessive profits,” but both he and Sam Altman know that these models don’t run on goodwill. The AI Cold War keeps monopolies on profits at the center of technological development. US lawmakers seek to outlaw DeepSeek as a “national security threat” not because they fear China, but because they fear losing control of the forces that have constrained technological innovation for so long.
Both China and the United States falsely claim to be leading the AI revolution. In reality, AI is being locked into an economic model in which potential is used only where it is profitable. This is not inevitable. The real war isn’t between Washington and Beijing but between those who want to realize the potential of technology and those who want to contain both it and all of us.