Trump's New Semiconductor Tariff Plan Is a Tax on America's AI Ambitions
The Trump administration is reportedly considering another broad round of semiconductor tariffs, potentially extending beyond chips themselves to laptops, gaming consoles and, most importantly, servers used in AI data centers. The proposal could also tie tariff exemptions to companies’ commitments to invest in U.S. semiconductor manufacturing. Importantly, this is still a proposal rather than a finalized policy, but the direction is clear.
The problem is that tariffs cannot manufacture chips. They can only make imported chips and the products built around them more expensive. And right now, the United States needs enormous quantities of advanced chips to build its AI infrastructure. Nvidia, AMD and other chip designers depend heavily on Asian foundries, particularly TSMC, while the U.S. is still building the domestic capacity that Trump wants to encourage. Imposing punitive tariffs before that capacity exists is essentially putting a tollbooth in front of an unfinished bridge.
The contradiction is particularly striking. Washington wants the United States to win the AI race against China, yet it is considering making the servers, accelerators and other hardware required for that race substantially more expensive. Industry groups are already warning that higher costs and greater uncertainty could jeopardize the massive data-center investment currently underway in the United States.
There is a legitimate argument for strengthening America’s semiconductor industry. Dependence on foreign manufacturing creates genuine national-security and supply-chain risks. But a targeted industrial policy is very different from a blanket tariff. Subsidies, tax incentives, infrastructure investment, workforce development and long-term procurement commitments can encourage domestic production without immediately increasing the cost of every computer, server and AI system entering the country.
And there is another problem: tariffs are a remarkably blunt instrument for dealing with a supply chain as complicated as semiconductors. A modern chip may be designed in one country, fabricated in Taiwan or South Korea, packaged elsewhere, incorporated into a server in another country and finally shipped to an American data center. Taxing the finished product doesn’t magically relocate that entire ecosystem to Arizona or Texas.
Trump’s approach risks producing the worst of both worlds: higher costs today and uncertain domestic capacity tomorrow. U.S. companies could end up paying more for the same foreign-made technology while waiting years for new fabs and supporting industries to reach meaningful scale.
The timing is especially questionable because the AI infrastructure boom is already creating enormous demand for chips and data-center equipment. The United States should be trying to accelerate that buildout, not deliberately adding friction to it. A tariff that raises the price of an imported consumer laptop is one thing. A tariff that raises the cost of thousands of servers needed to construct America’s next generation of AI infrastructure is a very different economic proposition.
There is also a danger that tariffs become less about industrial policy and more about political leverage. If companies can obtain tariff relief by promising U.S. investment, Washington effectively gains another mechanism for deciding which companies receive favorable treatment. That creates uncertainty for businesses and encourages investment decisions based on political negotiations rather than economics.
America absolutely should rebuild strategic semiconductor capacity. But the smarter objective is to make producing chips in America competitive enough that companies actually want to do it. If the only way to make domestic chips competitive is to make foreign chips artificially expensive, the policy may protect an industry without making it genuinely competitive.
The irony is hard to miss: Washington is trying to make America the world’s AI superpower while contemplating a policy that could make the machinery powering that AI revolution more expensive. That isn’t necessarily how you win a technology race. Sometimes the best industrial policy is knowing when not to put a tax on the very technology you are trying to accelerate.