New tariffs imposed by U.S. President Donald Trump are poised to severely disrupt the construction of next-gen data centers — potentially setting back the artificial intelligence revolution the administration has championed.
The White House announced stiff import duties: 34 percent on China, 32 percent on Taiwan, and 25 percent on South Korea — some of the top sources of key data center equipment. A baseline 10 percent tariff on all imports was also slapped across the board. While semiconductors were spared, analysts warn that AI infrastructure faces a direct hit, Reuters news report said.
United States will feel the heat because electronics – which include smartphones, PCs and data-center equipment – were the second biggest imports last year at nearly $486 billion worth of goods, according to Census Bureau data.
“Data centers rely on highly specialized hardware,” said analyst Gil Luria at D.A. Davidson. “There’s no doubt the equipment going into them is about to get significantly more expensive.”
That’s bad news for tech giants like Microsoft, Amazon, and Alphabet, already under pressure to justify their enormous AI budgets. It’s worse news for projects like Stargate, the $500 billion AI-focused data center initiative backed by OpenAI, Oracle, and SoftBank. Trump announced the project earlier this year with ambitions to build 20 mega-data centers across the U.S.
“Stargate was already a long shot,” said Luria. “With these tariffs disrupting supply chains and debt financing, it’s highly unlikely it can hit anything near its intended scale.”
Data center infrastructure — from GPUs and circuit boards to power distribution units — is sourced from tariff-hit countries like Taiwan, China, and South Korea. In 2024 alone, U.S. data processing machine imports are expected to top $200 billion, according to Bernstein, with much of that spending now at risk of being reallocated or delayed.
“Capital expenditure by tech giants will get reshuffled,” said Abhishek Singh from Everest Group. “Expect spending to shift from expansion to procurement hedging and re-sourcing.”
Microsoft has already pulled back. According to TD Cowen, the company scrapped U.S. and European data center projects totaling 2 gigawatts of planned power capacity in recent months. Analysts say it’s a reaction not only to AI demand lagging expectations, but also to mounting geopolitical uncertainty.
The biggest unknown? How AI hardware will be classified under the tariffs. If components like circuit board assemblies — often bundled with semiconductors — don’t get exemptions, the total cost of building a modern data center could surge.
“It’s not just chips,” said Dylan Patel of SemiAnalysis. “If the boards that house them are tariffed, you’re looking at an across-the-board price increase for AI-capable servers.”
For cloud providers like Amazon Web Services, Google Cloud, and Microsoft Azure, those added costs will ripple outward — potentially slowing the rollout of AI services for both enterprise and consumers.
As of the end of 2024, the number of large hyperscale data centers reached 1,136, doubling over the past five years, Synergy Research Group said earlier. Hyperscale data center capacity has also doubled in under four years due to increasing average facility size.
The U.S. remains the dominant force, accounting for over half of global capacity, while Europe and China make up most of the remainder. Annual growth is expected to continue with 130–140 new data centers coming online each year, but future capacity increases will be driven primarily by larger facilities, influenced heavily by generative AI demands.
On Thursday, the market reacted sharply. Nvidia, AMD, and Broadcom shares fell 7–10 percent. Amazon and Alphabet each dropped sharply, down 9 percent and 4 percent respectively. HSBC has cut its price target for Nvidia, citing an expected slowdown in cloud spending and AI infrastructure development.
Baburajan Kizhakedath
