Balancing AI Innovation with National Security
In an industry that moves at breathtaking speed, defining what technology is at the frontier or enabling access to it is a beguiling challenge.
According to the theories of nineteenth-century economist David Ricardo, firms trade across international boundaries freely. Geographic cost differences yield specialization; exchange yields mutual benefit. In the late twentieth and early twenty-first centuries, the Pax Americana enabled the system of comparative advantage and global trade to flourish as never before. American tech companies and their shareholders have been among the greatest beneficiaries of this system. Apple—with its products famously designed in California, made in China, and sold all over—is the prime example. Chip designer Nvidia, now among the world’s most successful firms in terms of market capitalization, is another.
Nvidia’s rise from obscurity to global sector domination is a great American success story. Starting in the 1990s, its GPUs (graphics processing units) originally served the video game market. Nvidia’s parallel processing technology development made images clearer and gameplay sharper. Incidentally, it laid the technical groundwork for the global artificial intelligence wave that has crashed ashore in the last decade. Nvidia’s revenue increased from less than $5 billion in fiscal 2015 to $10 billion in 2020 and to $130 billion in 2025. Its share price has gone from under $1 in 2015 to over $100 today.
Yet the unmatched quality of its output and the unprecedented demand for its products has lured the company into a compromised geopolitical position. Nvidia’s chips are key components driving the technological rise of America’s chief rival, China. In the most recent year, Nvidia’s China revenue topped $17 billion.
Since the debut of China’s DeepSeek open-source R1 model on January 20, Nvidia’s chip sales to China—direct and indirect—have come under acute scrutiny. Rightfully so. Assuming requisite scale, technological capacity is part-and-parcel with hard power. AI is the new key to hard power—a force multiplier par excellence that will supercharge military, intelligence, and cyber capabilities.
Technical supremacy in AI is the ultimate security trump. Amid renewed great power competition—and as the power at a scale disadvantage—maintaining, indeed expanding, that technical supremacy is essential to American security. Though not as blatant as if Lockheed Martin were selling jets to the People’s Liberation Army, if Nvidia chips are enabling Chinese AI systems with military applications, then the U.S. Bureau of Industry and Security should have the authorization to proscribe sales.
While capitalism is positive-sum, geopolitics is not.
To its credit, the Biden administration established an export control thesis that struck a reasonable balance between the economic freedom of American firms and the denial of capabilities to foreign powers—what National Security Advisor Jake Sullivan called the “small yard, high fence” approach. However, this is easier said than done. In the wake of the R1 release, the Trump administration ought to pursue a better-targeted and more adaptive export control regime to fasten the fenceposts and right-size the boundaries.
First, the explicit target should be to deny China access to the hardware and software that constitute frontier technology. Export controls are no guarantee that America will maintain its edge—as a simple matter of incentives, they will spur China’s own domestic development forward—but they are a needed firebreak. They should start as far up the production chain as possible to be most effective. Indeed, today they do, with the world’s sole producer of the extreme ultraviolet lithography machines that etch top-end chips, the Netherlands’ ASML, closely bound by its home government to the U.S. policy outlook. That arrangement requires Chinese entities to look abroad to purchase the chips they need.
In an industry that moves at breathtaking speed, defining what technology is at the frontier or enabling access to it is a beguiling challenge. With the humility to acknowledge that its execution will never be perfect, the administration should retain the technical experts indispensable to stay abreast of developments and adjust policies as necessary.
For example, the most startling feat of DeepSeek’s R1 is that its “inference” efficiency revealed that restrictions on leading-edge chips might be insufficient. R1 is believed to have used Nvidia chips designed specifically to comply with existing export controls. It may be the case that denying Chinese advancements means bans even at the trailing edge, which DeepSeek proved adept at deploying.
While tightening controls on China, the administration should make other changes to facilitate more commerce with countries that do not present security threats or the same degree of smuggling risk. A “small yard, high fence” approach works only if the yard is indeed small—blocking a rival’s access to leading-edge tech while keeping American firms free to compete globally. Dean Ball of the Mercatus Center and the Foundation for American Innovation sees decluttering the complex “model weight” rules for exports and upgrading allied countries like Israel and Poland to the approved “Tier 1” list as a crucial adjustment.
Meanwhile, the enforcement of carefully designed restrictions must be tightened. That Singapore—a favorite bolthole for Chinese businesses—accounted for a quarter of Nvidia’s revenue in the most recent year suggests that a lucrative smuggling network is at work. According to a March report from The Wall Street Journal, brokers in China are fulfilling orders for export-controlled Nvidia chips in as little as six weeks. The United States should work with Singapore—a vital Indo-Pacific security partner—to identify gray-market resellers funneling chips into China.
To restrict exports of top-selling technology is to subvert the Ricardian model that has made American firms and American consumers alike better off. It is not a choice that we ought to make hastily or with any glee. With its total market reduced, Nvidia—a company at the heart of the most promising industry this century has on offer—will have its prospects shaved down somewhat. It will garner less capital. Its technical advances might slow, and some of the gains its chips would have otherwise delivered may not materialize. But in the new world of great power competition, it is incumbent upon the U.S. federal government to execute the security duties it has to all citizens and sustain the country’s flourishing as a free, commercial republic.
A calibrated export control policy that adjusts to a fluctuating technological landscape strikes the best balance between firm-level competitiveness and national security.
Jordan McGillis is the economics editor of City Journal. Follow him on X: @jordanmcgillis.
Image: Ivan Marc / Shutterstock.com.