President Donald Trump says he wants the United States to lead the world in artificial intelligence. But the president’s tariffs on raw data center materials could cause American AI to fall behind.
The rapid growth in the AI industry has brought with it a data center building boom in recent years. In 2025 alone, tech giants have earmarked about $320 billion to spend on AI infrastructure, with Amazon leading the way with $100 billion in investments. High-powered data centers contain rows of computers that store and process data. As AI technology advances, larger and more sophisticated data centers are needed to handle the growing demands for capacity and performance.
But last month, the Trump administration imposed a 25 percent tariff on imports of steel and aluminum, which are crucial to data center construction. And he has threatened to impose tariffs on copper and semiconductors, which would add billions to construction budgets just as tech companies are racing to expand their AI capacity.
Higher costs may not derail the industry outright, experts say. But in an international race to build the best technology as fast as possible, further hiccups in the supply chain may put America behind competitors like China’s Deep Seek.
Even before Trump announced 25% tariffs on steel and aluminum and additional tariffs on all global imports, AI data centers were already struggling to source goods needed to construct and power these massive processing centers. A recent survey by industry group AFCOM found that 74 percent of data center professionals reported persistent supply chain issues over the past year.
In some cases, those constraints are delaying data centers from coming online. Bill Kleyman, who runs Apolo, an AI infrastructure company, said one of his clients has managed to build out his entire data center, minus one crucial piece of power equipment due to supply chain constraints. “Put simply, it’s like not having a spark plug for a car you’ve just built,” Mr. Kleyman said. With new trade barriers looming, those challenges are poised to deepen—threatening the very growth the administration says it wants to promote.
There are already over 5,000 data centers across the country, but only a small fraction of these facilities are capable of supporting the massive workloads driving today’s artificial intelligence boom. To simply keep pace with current data processing demands, at least twice the data center capacity built in the past 20 years must be constructed in less than a quarter of the time, according to an analysis by McKinsey. The AI revolution is outpacing infrastructure growth, creating a bottleneck.
Thus to lead in the global AI race, the US must build hundreds of new data centers from the ground up, beginning with the basic materials.
Data centers are made heavily with steel and concrete, as well as aluminum and copper. Steel is used in structural framework on the buildings, the support beams, and server racks while aluminum is featured in cabling infrastructure and cooling systems.
Currently this steel comes from a mix of domestic and international sources, including from China. Even if tech companies buy metals domestically, as Trump intends for them to do, they’ll see higher costs due to rising demand for American steel across industries and supply chain pressures. A year after Trump previously raised levies on international steel in 2018, steel prices rose 38% domestically. This time around, steel prices rose 4% in February alone, according to the Associated Builders and Contractors analysis of U.S. Bureau of Labor Statistics’ Producer Price Index data.
Sourcing steel is one of the first steps to constructing these mega data centers, with advanced microchips known as graphics processing units or GPUs to follow. “In this new phase of AI, steel, servers and power will replace models, compute and data as the ‘must-wins’ for anyone hoping to pull ahead,” Sequoia Capital, an investment firm, wrote in a 2024 brief.
AI investments have emerged as a critical driver of the economy–estimates show that building up AI technologies made up 16%-20% of real GDP growth in the third quarter of last year alone. Trump’s trade policy threatens to slow that headway.
Moreover, tariffs are hitting at a time when some AI insiders were already beginning to worry that the industry had built more data centers than it needed to meet demand. Last week, Microsoft scrapped plans for projects with OpenAI and cancelled data center leases in the US and Europe after walking away from other projects earlier this year.
“I start to see the beginning of some kind of bubble,” Alibaba Group Holding Ltd. Chairman Joe Tsai told investors at a recent summit. “I start to get worried when people are building data centers on spec.”
Other industry experts like Alan Howard, who studies these co-location agreements and data center construction, say the cancellation of 2 gigawatts worth of data centers is not evidence that the AI boom is about to go bust given the scale of the industry. “While it’s a colossal amount of power, at the same time, it’s peanuts,” said Howard.
For now, tech giants like Microsoft, Amazon, Google, and Meta seem prepared to absorb higher construction costs to stay ahead in the AI arms race. But down the line those costs may be passed down to consumers–meaning that the “free” AI tools many users now take for granted, like ChatGPT, might disappear behind paywalls.
