Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Earlier this year, an AI-piloted F-16 participated in its first full dogfight maneuver exercise. The pilot (who sat in the cockpit but did not control the aircraft) was so impressed by the computer’s performance that he stated that he would be willing to give the system fire control authority. However, America’s progress in fielding powerful AI solutions such as this may be at risk under the incoming Trump administration.
President-elect Donald Trump has made it clear that his foreign policy will be focused on promoting domestic industry through tariffs aimed at not only near-peer competitors but also allies and neighbors. With Elon Musk and crypto hawking abounding on the campaign trail, Trump appears to be a friendly ally to the technology industry. He has promised to reverse President Joseph Biden’s AI restrictions in military decision-making to promote America’s indigenous AI industry. However, proposed tariff and export control policies will have much larger effects on AI hardware supply chains. The future of AI hinges on access to advanced chips made in Taiwan, backed up by a global network of materials and manufacturing. While the first Trump administration had carved out exceptions for Taiwan in its Section 301 tariffs aimed at China, on the 2024 campaign trail, Trump made it a point to lambast not just China but took aim at Taiwan as well. He claimed that Taiwan was taking chip manufacturing away from the United States and that Taiwan should be paying the United States for defense. Moreover, aiming at traditional allies and partners that in the past have shared many of the same economic values is an even bigger challenge. Raising the price of computational hardware robs the United States of a key advantage in the global AI race — a competitive domestic market.
Hardware, Hardware, Hardware
At the Windows 2000 launch, then-Microsoft chief executive officer Steve Ballmer jumped on stage and yelled “developers, developers, developers.” At a time when Moore’s law was holding strong, the strength of software developer talent made or broke a firm. This is no longer the case. While the death knell of Moore’s law has been speculated about for years, there is a clear push to look at other areas of improvement in computing. Both the scientific literature and NVIDIA’s chief executive officer, Jensen Huang, argue that the future of computational power will come from software, algorithms, and hardware infrastructure. The inseparability of these factors is impossible to understate, and tariffs on the hardware infrastructure economy will have blanket effects across the supposed “hyper Moore’s law.” Fundamentally, future AI will hinge on access to hardware. There are two ways that would-be startups can implement AI at scale: in-house development or leveraging cloud computing systems offered by large firms. Both mechanisms will be fundamentally altered after blanket tariffs, making it harder, if not impossible, for startups to engage in the artificial intelligence market.
NVIDIA has recently emerged as the preeminent hardware design firm in the United States, and indeed globally, for AI chips. Graphics processing units from NVIDIA remain the most powerful for enterprise solutions, particularly GB200 Grace Blackwell Superchips, which represent the cutting edge for data center offerings. Already, the next generation of GB200-based systems has incurred massive orders from Microsoft and Elon Musk’s xAI. NVIDIA, however, does not produce its own chips; the major effect that tariffs will have will not be on the design end, but on production. Blackwell architecture is built on proprietary 4NP nodes at Taiwan Semiconductor Manufacturing Company plants, the only plants capable of producing NVIDIA’s chips.
Trump has indicated that he intends to implement a blanket 10 to 20 percent tariff on all imported goods, without exception. At a minimum, this would drastically raise operating costs in dealing with all foreign firms, but a 20 percent increase would hit Taiwan with a 1,000 percent increase in average tariffs. The Taiwan central bank has expressed concern over the policy, arguing that if implemented, it will be the most important trade policy of Trump’s administration. Moreover, with Trump potentially expanding export controls in strategic materials like silicon, this would massively impact Taiwan Semiconductor Manufacturing Company’s supply chain. Out of the five major suppliers of raw materials for the Taiwanese firm, two are headquartered in America — SunEdison (formerly MEMC Electronic Materials) and Short Elliott Hendrickson Incorporated.
Tariffs on neighbors and allies will also have a substantial impact. Microsoft’s data center expansion represents global ambitions of U.S. AI firms. One of Microsoft’s more notable expansions is in Quebec City, Quebec, designed specifically for AI. Trump threatened to levy a 25 percent tariff on all goods coming from Canada and Mexico, and after a meeting with Canadian Prime Minister Justin Trudeau failed to reach concession, relations have continued to cool between the two countries. As tariffs increase the price of moving goods across the Canadian border, data centers will undoubtedly be affected. Even discounting the increased transaction costs of tensions between Canada and the United States, the price of data centers will absolutely increase. Data centers will not be able to avoid a significant increase of battery prices, a market dominated by China. At the most charitable interpretation, this implicates accessibility to cloud computing systems, which certainly will become more expensive post-tariff, pricing would-be startup firms out.
Implications for the Security AI Market
The security implications for hardware constraints on new AI firms are potentially game-changing. The fundamental difference between the U.S. and Chinese models of military AI development is the democratization of the technology. Specifically, the U.S. military relies on an open and easily accessible market for AI firms, and thus security competition demands open and easy access to AI hardware. The U.S. military has made massive bets on the AI market, with billions being pumped into firms of all sizes to improve military capacity, efficiency, and efficacy. The U.S. Army has offered millions for specifically small business AI grants to enhance its capabilities. In fact, one of the first Pentagon contracts aimed at defending against generative AI threats was awarded not to large industrial players like Amazon or even OpenAI, but to Jericho Security, a small firm with a $1.8 million valuation. The U.S. Army itself has deployed its own generative AI system hosted on Microsoft’s Azure cloud services.
Thus, the hallmark of U.S. military AI development is based on competitiveness, which encourages new entrants into the market. With tariffs looming and the clear threat that computational power will become more expensive, new entrants are pushed out of the market. This only leaves large firms like Microsoft and Intel in play. Even discounting the increased cloud computing cost falling to consumers, this process introduces massive risk. While on face, a single large vendor might not be an impossible challenge to overcome (indeed, for other high-tech systems like fighter jets, this is the norm), AI is different. In the United States, AI is largely democratized already. Non-traditional defense companies like Anduril and Jericho Security represent potentially disruptive advancements in AI that democratization of both talent and hardware access was necessary to achieve. For a firm to enter the AI market requires relatively new systemic organization, but it is much cheaper than manufacturing an airframe.
Chinese Competitiveness
With the effectiveness of the future of military AI development in question, enter the challenger, China. China does not operate on market principles in its AI development or goals. The 2023 creation of China’s Central Science and Technology Commission illustrates the core strategy of Chinese innovation: to mobilize specific sectors of the Chinese state and industry to outpace the United States. The Central Science and Technology Commission was created in response to U.S. tariffs and is designed to operate in absence of access to foreign technology. This means that China has been preparing to innovate under the pressure of more U.S. tariffs. Going about AI alone is a road that will likely be tough for China to successfully traverse, but if the U.S. AI market were to be constrained, the security focus of China’s innovation system might allow it to achieve parity.
Chinese chip design firms are not as advanced as the United States’ NVIDIA, but their advantage is the development of a growing indigenous semiconductor manufacturing industry and centralized ambitions that have already weathered the tariff storm. The Center for Strategic and International Studies finds that export controls have mixed effects on technological progress in China, pointing to mixed effectiveness of protectionist policies to slow down Chinese AI development. Even under the pressure of hardware constraints, China has made major progress with older technologies. The Chinese firm Intellifusion has developed the “DeepEyes” AI box capable of competing with major players in the United States in the AI space. The remarkable aspect of DeepEyes is its development on a 14nm production node, a decade-old production standard. Thus, with or without export controls on China, there is little doubt that China is prepared to continue its AI development as part and parcel of its geopolitically competitive goals.
Notably, tariffs that target China are nothing new. Their developmental capacity has been developed under considerable tariff pressure; the proposed Trump tariffs, however, go much further than just targeting China. Playing out a scenario where hardware constraints indeed become salient for U.S. AI firms makes clear the long-term effects in a larger geopolitical context. Hardware costs increasing both data center costs and cloud computing costs will pressure potential new AI firms out of the market. With fewer vendors available to the U.S. military, there is little doubt that China will leverage its homegrown industries that it has been building for years to outpace the United States. Allies, like Canada, that have been traditionally viewed as safe investment locales will suddenly be thrown into disarray. Even though Biden’s recent export controls have attempted to carve out strategic exceptions, if the full weight of Trump’s protectionist sentiments should be believed, there is little doubt that these exceptions will stay static. Moreover, industrial watchdogs are already wary of supply chain effects due to a lack of exceptions for places like Malaysia or Taiwan. The lack of diverse options to choose from in the AI landscape will rob the Department of Defense of its current comparative advantage. Tariffs will allow the centralized mobilization of the Chinese developmental model to shine and, in the most pessimistic of views, allow the Chinese science and technology apparatus to outcompete the United States.
Reshoring Is a Fantasy
The protectionist counterargument is that tariffs forces companies like Intel to reshore manufacturing toward U.S. soil. The associated economic effects may be beneficial, overall growing the economy and encouraging high-tech employment, and allowing for decoupling. To put it bluntly, this is an impossible dream. Intel has been the poster child for America’s homegrown semiconductor manufacturing capabilities, and in recent years it has lost its competitive edge. The recent announcements are telling. Intel’s leadership restructuring promises to bring long-term gains, but there is broad skepticism that this change is on paper only.
The bigger problem, however, is that Intel is not the American competitor to Taiwan Semiconductor Manufacturing Company in chip fabrication that it used to be. Indeed, Intel relies heavily on the Taiwanese manufacturing. The newest generation of consumer-grade Intel graphics processing units, which promises robust AI improvements, is manufactured on Taiwan Semiconductor Manufacturing Company’s 5nm manufacturing node. Moreover, one of the primary points of blame placed on former Intel chief executive officer Pat Gelsinger was that he fundamentally mishandled the close relationship between Intel and the Taiwanese firm. The very ideals of a strong Intel (and thus a strong American manufacturing sector) are based in strong cooperation between the United States and overseas manufacturing.
The other barrier to American semiconductor manufacturing, which tariffs fundamentally cannot overcome, is simply an issue of labor in the United States. The semiconductor industry is well funded but faces a consistent labor shortage in the United States. This has been attributed to a wealth of federally backed job openings, fewer skilled engineers, and increased restrictions on foreign talent. Thus, relying solely on protectionist trade policy without addressing domestic shortcomings in workforce shortages, particularly an education gap, makes reshoring impossible. Finally, none of these economic considerations resolve the AI arms racing issue, and as Intel restructures, China will remain unimpeded, even under the behemoth of American tariffs.
Uncharted Territory
As the United States marches toward another Trump administration, his political motives are wracked by uncertainty. However, one thing remains certain: protectionism as a central foreign policy goal is an animating feature of his rhetoric and campaign platform. Tariffs placed on both competitors and allies are likely to be enacted, and widespread effects on not only America’s domestic economy, but also the U.S. military will be felt. The Trump administration may not let secondary and tertiary tariff effects fundamentally undermine U.S. military effectiveness. However, even if the government is willing to pay subsidies or larger contracts, the added cost is a crucially tradeoff. Tariffs are thus a lose-lose situation for high-tech applications like the U.S. military. Either the Department of Defense must offset the increased costs through other taxpayer-funded programs or risk losing ground to China. The race for advanced chips and artificial intelligence development cannot be viewed in isolation, and China is a huge potential winner poised to take advantage of constrained international markets. China’s insulated and centrally mobilized AI and security apparatus will be able to leverage its homegrown industry advantage over a U.S. market pressured by supply chain uncertainties and increased data center costs. Without the market options that it is configured to exploit, the U.S. military’s high-tech comparative advantage will be put into question.
Alex Li is a political science doctoral student at the University of Oregon. He holds a masters of international affairs from the School of Global Policy and Strategy at the University of California, San Diego, focusing on China and international politics. His work and research have focused on conflict and national security.
Image: Trump White House Archives
