Overly Stringent Export Controls Chip Away at American AI Leadership

The Biden administration’s export control policy for AI chips has largely been a failure since day one. Yet, year after year, it has doubled down, attempting to plug various loopholes. The policy has now expanded to an almost comical extent—but the consequences are no joke. While the U.S. government is right to prevent U.S. companies from selling advanced AI technology to the Chinese military, cutting U.S. companies off from the entire Chinese market is a cure worse than the disease. It will ultimately harm both U.S. national security and economic interests.

The Biden administration issued its first AI chip export controls in October 2022, restricting the export of AI chips to China, as well as the technology to manufacture those chips. Since then, it has expanded those restrictions to cover additional countries, less-advanced chips, and related hardware components (e.g., high-bandwidth memory).

The most recent set of restrictions, introduced in January 2025, imposes a three-tiered system for access to advanced AI hardware and models. Under this latest framework—scheduled to go into effect on May 15 unless altered by the Trump administration—120 countries, including close allies such as Israel, India, and Singapore, face a cap on the number of advanced chips they can receive. But instead of limiting China’s access to advanced AI, these sweeping restrictions risk isolating U.S. firms from the global market and creating openings for Chinese competitors to step in.

Most recently, the Trump administration has extended the export controls to Nvidia’s H20 chip. Nvidia specifically designed the H20 to comply with U.S. export controls. In terms of specifications, it is comparable to the H100 chip, a popular graphics processing unit (GPU) chip Nvidia released in 2022, but the H20 is more energy efficient and has slightly more memory. However, the H20 is significantly less advanced than Nvidia’s newest line of chips, the B100 and B200, and it has less memory than the H200.

The U.S. government is right to want to keep advanced AI chips out of China’s military systems. But effectively banning Nvidia from exporting chips like the H20—as noted, designed to meet U.S. export requirements—goes too far. Nvidia has already reported a $5.5 billion financial hit from the new restrictions. Blocking access weakens U.S. competitiveness and accelerates China’s efforts to build a self-sufficient chip industry.

The United States will not achieve AI dominance if its firms cannot access global markets. U.S. markets alone are simply not large enough to support globally competitive high-tech firms, which must have access to global markets of scale to earn sufficient revenues needed to reinvest in the expensive effort of bringing next-generation products to market.

U.S. export controls have accelerated China’s efforts to close the gap. In 2019, Huawei’s Ascend 910 chip was manufactured by TSMC in Taiwan. Following U.S. restrictions, Huawei moved production to mainland China. It took two years, but by 2022, it was producing the 910B chip, a slight improvement over the original. Now, Huawei has announced that the Ascend 910c, the next version of this chip, will be released as early as May 2025, and it continues development of its next-generation Ascend 910d chip. Analysts expect these chips to perform on par with the Nvidia H100, but at 60-70 percent of the cost.

Huawei isn’t the only player that is helping China close the gap. China has firms innovating and competing across all layers of the chip ecosystem. Companies such as Biren Technology, MetaX, and Enflame are designing AI chips. Semiconductor Manufacturing International Corporation (SMIC) competes with TSMC for fabrication and is reportedly nearing the ability to produce 5nm chips (an improvement over its current 7nm, though still well behind TSMC’s 2nm).

China’s Yangtze Memory Technology Corporation (YMTC) has produced high-density memory chips comparable to those produced by its Korean rivals, despite U.S. export restrictions. Moreover, China is innovating in how to deploy less efficient chips more effectively. Earlier this year, a Chinese research team won an award at a prestigious international conference, doing just that—using less powerful chips to outperform high-end hardware. And as DeepSeek showed when it released its highly capable AI model at a fraction of the cost of its rivals, China does not have to make the best chips if it can make the most affordable ones.

It’s also important to recognize that China’s goal is to develop high-tech products—from mobile phones to AI and legacy chips to foundry capabilities—that are good enough to replace foreign imports with domestically produced Chinese goods. Indeed, China has been quite unequivocal that it seeks to achieve at least 70 percent self-sufficiency in the semiconductor industry by the end of the decade.

Overall, China likely remains a few years behind U.S. chipmakers—although Nvidia CEO Jensen Huang recently stated, “China is right behind us. We're very, very close.” But it continues to make progress, and Chinese firms enjoy full support from their government. Last year, the Chinese government created a 344 billion yuan ($47.5 billion) investment fund to support its semiconductor industry, following similar investments in 2019 (204 billion yuan) and 2014 (139 billion yuan). In fact, if one adds up semiconductor-sector investments made by Chinese governments at all levels (national, provincial, city), Tufts University’s Chris Miller has suggested that China “has probably invested the equivalent of [America’s] CHIPS Act virtually every year since 2014.”

Uncertainty and rapid policy shifts don’t just target China—they undermine American firms that follow the rules, limit their ability to compete globally, and push international customers toward Chinese suppliers. U.S. companies such as Nvidia are being penalized for designing products that comply with the rules, only to have the rules changed after the fact. This not only damages American economic interests but also erodes trust in U.S. trade policy.

Worse still, the current approach is counterproductive. By cutting off U.S. firms from international markets, the United States is effectively accelerating China’s progress in AI chips. Chinese companies now have a stronger incentive—and greater market opportunity—to replace U.S. suppliers, spurring domestic innovation and attracting buyers who can no longer rely on American technology. Rather than slowing China's ascent, the U.S. government is hastening it.

Export controls should be precise and strategic, not sweeping and shortsighted. In that regard, one area where the Biden administration deserves credit—and which the Trump administration should preserve—is aligning key allies, notably Japan and the Netherlands, to restrict export of the most advanced semiconductor manufacturing equipment (SME), especially extreme ultraviolet (EUV) lithography systems, to Chinese entities.

While China may innovate around some of these barriers over time, these tools remain a significant chokepoint in the near term. Leveraging allied coordination in this area is a far more effective way to slow China’s AI and semiconductor progress than blanket restrictions on chip exports. They can, at a stroke, slow the development of China’s broader semiconductor and AI ecosystems. It’s much easier to control the fishing rod than the fish.

However, this strategy comes with trade-offs. Limiting SME exports imposes real costs on U.S. and allied toolmakers, underscoring the need to expand trusted markets. Policymakers should prioritize deepening information and communications technology production ecosystems and digital economies of allied nations so Western tech companies can grow their footprints in friendly countries. This is one reason why the January 2025 restrictions on AI chips were so ill-conceived. At stake is the opportunity to reshape the global economic playing field in favor of allied interests—building a world in which, a decade from now, U.S. tech companies are far less reliant on China because so much more of the world’s digital infrastructure is rooted in trusted markets across Africa, Europe, Southeast Asia, and the Americas.

If the United States wants to maintain long-term leadership in AI and semiconductors, it should establish a clear and sustainable policy that protects national security without stifling innovation or isolating U.S. companies from the global economy. This effort cannot be a U.S.-only endeavor; it should engage and deliver benefits for allies as well. Without such a course correction, the United States risks not only undermining its own technological edge for years to come—but boosting China’s in the process.