Cracks in the Firewall: How U.S. Tech Policy is Losing Ground to China
Editor’s note: In February, The National Interest organized a symposium on the U.S.-China technology race amidst the emergence of DeepSeek and ongoing legal battles over TikTok. We asked a variety of experts the following question: “What are the three most important technology policies that the U.S. should pursue or avoid to compete adequately with China?“ The following article is one of their responses.
America’s strategy to safeguard its technological supremacy is riddled with critical weaknesses. As Washington scrambles to limit Beijing’s access to key technologies, abounding Chinese advances in AI and semiconductors expose the cracks emerging in the United States’ rapidly aging technology policy framework. Staying ahead demands more than passive containment – it necessitates decisive agility.
Three priorities stand out. First, export controls must be sharpened to close loopholes that allow Beijing to stockpile critical chips. Second, any attempt to weaken or repeal the CHIPS Act must be vanquished; the Act’s integrity must be defended to fortify domestic semiconductor production and insulate the U.S. from supply chain vulnerabilities. Third, Washington must ensure AI is not just developed but widely adopted across American industry, turning innovation into sustained economic and military advantage.
Refine U.S. Export Controls
Chinese AI firm DeepSeek’s sensational claim to have trained a GPT-4-level model for just $5.6 million has sparked spirited debate on the perceived fragility of U.S. technological predominance. Many are rushing to find the culprit. Some have declared that U.S. export controls on advanced AI chips have backfired; they have exaggerated the importance of monopolising cutting-edge compute access, compelled Chinese engineers into paradigm-shifting efficiency innovations and unsettled key U.S. allies.
This analysis blames the blueprint for the builder’s failure. The export controls, set in motion under the first Trump administration and extended under Biden, have proven vital in curbing Chinese advances in AI deployment and training, high-performance computing and next-generation chip fabrication capabilities. DeepSeek’s founders have admitted as much. While limited compute may drive efficiency improvements, it also restricts experimentation and scaling, inhibiting the pace of large-scale technological advances. This is why industry heavyweights investing hundreds of millions in chips are driving AI innovation, not compute-starved start-ups.
Rather, it is the slow and deficient nature of the export controls that imperils the American technological edge. DeepSeek leveraged pre-export control access to stockpile large quantities of Nvidia chips, accumulating a cluster of 10,000 A100 semiconductors as well as thousands of H800 chips. The U.S. Department of Commerce only banned H800 exports in October 2023, a year after it became aware of this regulatory lacuna. Had controls been implemented earlier, DeepSeek would have faced significantly greater hurdles in training an advanced model.
There is also mounting concern that DeepSeek employed cutting-edge Nvidia H100 chips that were smuggled by China through shell companies and faux data centers in countries exempted from the export-control regime. Beyond this, the U.S. has completely failed to ban exports of chips with high memory bandwidth that are crucial for deployment workloads, such as the Nvidia H20. Lastly, recent reporting reveals that hundreds of thousands of TSMC chips ended up in Huawei products, revealing vulnerabilities in the supply chain that demand stricter oversight.
Without a swift and adaptive export control regime, China will continue to exploit these loopholes and weaknesses to strengthen its technological prowess at the United States’ expense. The Trump administration must move to ban H20 exports and strengthen tracking and verification mechanisms for advanced semiconductor trade, ensuring they do not reach China through third-party countries. International data centers using U.S.-designed chips must be subject to increased security compliance measures to prevent dealings with Chinese shell companies. Further, Trump should increase the funding and powers of the Bureau of Industry and Security to improve its enforcement capabilities.
Closing the loopholes also necessitates adroit diplomacy to secure multilateral alignment. Indeed, there is little use for U.S. export controls if China can access the same technology from an alternative source. The U.S. is currently negotiating with Japan on a deal to further restrict chip exports from firms such as Tokyo Electron and Nikon to China. Yet in the face of Beijing’s threats to retaliate by cutting Japan’s access to critical minerals exports, Tokyo is reluctant to impose broader restrictions.
Trump should urge Japan, and other key allies such as the Netherlands, to fully align with U.S. export restriction. In return, the U.S. should secure alternative supply chains for critical minerals for its allies, diminishing Beijing’s leverage and cutting-off China’s access to alternative chip suppliers.
Do not repeal the CHIPS Act
Days before the presidential election, Trump condemned the CHIPS and Science Act, a groundbreaking bipartisan effort to cement U.S. leadership in semiconductor manufacturing, as “so, so bad”, threatening to replace its subsidies with tariffs. House speaker Mike Johnson later predicted the act’s imminent repeal. Washington’s abortive funding freeze threatened to cease all implementation of the Act, as the new Commerce Secretary Howard Lutnick cast doubt on its survival.
Such language should alarm all who fear a unipolar world order lead from Beijing. Not only are chips indispensable to AI development and training, they are vital components in shipping, airplanes, advanced weapons and drones – all strategic technologies. TSMC in Taiwan produces 90% of cutting-edge chips and 100% of ultra-advanced AI chips. That these critical technologies are predominantly manufactured on a vulnerable island threatened by conquest from America’s greatest adversary necessitates a wholesale reshoring and expansion of domestic U.S. semiconductor manufacturing.
The CHIPS Act is working to remedy this petrifying national security risk. Since its passage, over 70 semiconductor fabrication plants have been constructed in the U.S., driving $11 billion in private investment per month in domestic electronic manufacturing. This far outweighs the government spending involved; chip companies and supply chain partners have committed $327 billion in investments over the next decade. Samsung, TSMC, Micron and Intel are now building major new plants in the US. TSMC’s Arizona plant is operational and exceeding the chip quality and yields of its Taiwanese counterpart.
Without CHIPS, U.S. semiconductor production would have dipped from 12% to 8% by 2032, plunging ever further as Beijing strengthens its grip on frontier technologies. Instead, by 2030, the U.S. could account for 20% of global production of the most advanced chips, a rise from zero today. This would be sufficient to secure critical infrastructure like datacentres and telecoms without the need for imports, which could prove pivotal in a future East Asian security crisis.
Cancelling this unprecedented investment in domestic capacity and R&D would not only impede American technological innovation. It would also drastically escalate the risk of U.S. economic dependence on Chinese supply chains, as in the case of lithium batteries. The national security implications could be catastrophic.
AI must be diffused, not just innovated
As the superpowers lock horns in the tech race, policymakers in both Washington and Beijing focus inordinately on making sure it is their side that pioneers the next artificial intelligence breakthrough. The panic over DeepSeek is just the latest manifestation of this somewhat misguided anxiety.
Indeed, economic history reveals that true economic power lies in how well a country spreads and adopts new technology across multiple industrial sectors. Subsidies, export controls and restrictions to Chinese visas will only get you so far. The real question is not who invents new AI tools, but who applies them most effectively across the economy, thereby actually leveraging its productivity potential.
The U.S. lagged behind the U.K. in innovating sophisticated machinery during the late 19th century. However, in mere decades, America outpaced the European powers in productivity and emerged as an economic superpower by deploying machine tools across more sectors of the economy and at a greater rate. By 1907, U.S. machine intensity was more than double that of the U.K. and Germany. The U.S. also democratized access to manufacturing expertise through land-grant schools and technical institutes, spreading knowledge beyond the elite. Washington retained its technological and economic edge in the 21st century through widespread diffusion of cloud computing, industrial software and other frontier technologies.
As such, rather than focusing solely on dominating the next innovation and sophistication, technology policy in the era of AI should prioritise the diffusion of AI across as many sectors of the economy as possible. All feasible American industries should be empowered by this technology to drive the U.S. economy ahead of its rivals, not just the tech giants. This will require expanding AI workforce training beyond Silicon Valley by leveraging CHIPS Act STEM-training initiatives, government programs to help businesses adopt new technologies and partnerships with key industries.
No matter how hard America tries, it can never fully sever China from access to AI. However, Washington can ensure its emergence as victor in the race for technological predominance by drawing on its global leadership in education and historic strengths in technological diffusion.
About the Author: Uri Inspector
Uri Inspector is an Associate Researcher with the Cambridge Middle East and North Africa Forum (MENAF).
Image: Shutterstock