Trump’s long-touted Liberation Day imposed a ten percent tariff on the rest of the world and additional so-called reciprocal tariffs based on trade deficits. A week after that announcement, the president announced a 90-day pause on most countries except China—significantly escalating trade tensions between the world’s two largest economies while holding firm on 10 percent tariffs worldwide. These new costs will undermine Trump’s artificial intelligence (AI) opportunity agenda by increasing the price to build data centers and undermining AI investment.
Trump has long championed AI—and continues to do so now. During his first term, he issued an executive order to encourage its development, prioritizing research & development into AI and reducing barriers to test and develop new AI technology. Trump 2.0 was off to a good start on AI by rescinding the Biden-era executive order that imposed significant red tape and oversight on the most advanced AI models. But his tariffs will undermine all of this by increasing the costs of everything that goes into data centers built in the United States.
Data centers house equipment (Graphics Processing Units, or GPUs, networking hardware, cooling systems, and more) needed to train and run AI. Training AI algorithms requires sifting through massive amounts of data, using advanced computers housed at data centers, so the program can learn to identify patterns and create new, useful results based on the training data. This complex array of processors and the infrastructure needed to support the algorithms allow for the efficient training of the most advanced AI and tariffs are set to impact several key inputs.
For example, GPUs—critical components in data centers used to do multiple calculations simultaneously that are often imported from China—will be subject to tariffs. Optical modules, devices that facilitate the rapid transmission of high amounts of data in data centers needed for low latency data transfers, are mostly imported from Asian countries and will be subject to the global ten percent tariff generally and the tariffs on China specifically. Cooling systems needed to transfer excessive heat away from computers while they are making calculations are also mostly imported from Asian countries. Steel is often used to build the enclosures and support while aluminum is used for cable racking in data centers—both are subject to a recent increased tariff rate of 25 percent.
Fortunately, some crucial components, like semiconductors, copper, and rare minerals are exempt from at least the reciprocal tariff increase—for now. But Commerce Secretary Howard Lutnick indicated the potential for their own category of tariffs in the future. In a first step toward this goal, the Commerce Department has begun to explore options to impose tariffs on semiconductors and other electronics on national security grounds.
Outside of the computing aspect of data centers, tariffs will make meeting demand for data centers expensive in less obvious ways. For example, building out energy capacity to meet growing AI energy demands is likely to become more expensive. Parts needed for these constructions, such as large power transformers, are largely imported and would not be exempt from tariffs despite their necessity for data centers.
Even if tariffs do successfully encourage aluminum processing in the United States, the energy-intensive extraction process would put even more of a strain on the electrical grid. Aluminum production currently consumes 1.5 percent of all electricity demanded by home, commercial, and industrial sectors and is the most energy-intensive material produced in the United States. Reshoring more aluminum production to the United States will increase energy demand at the same time AI increases demand for electricity more generally.
Trump’s AI policy is promising but it will only be successful if it can be salvaged from the wreckage of his trade policy agenda. Tariffs will increase the cost of developing new AI technologies and leave less to invest in improving them. Spending more paying taxes rather than investing in data centers is likely to slow down technological development and harm the United States in the ongoing AI race.
Trey Price is a policy analyst with the American Consumer Institute, a nonprofit education and research organization. For more information about the Institute, follow us on X @ConsumerPal.