Donald Trump’s global tariff regime hampers the US goal of dominating races to develop globally momentum artificial intelligence, while jeopardizing his ambition to encourage domestic chip production.
Industry officials, including high-tech executives, supply chain experts and analysts, said the US president’s escalating trade war is likely to hamper the expansion of America’s computing power. This is because measures could increase the costs of building US semiconductor manufacturing plants and AI data centers.
A concern in the technology sector is that efforts to force greater onshoring chips and electronics manufacturing will have the unintended effect of suppressing Openai, Google, Microsoft and others who are trying to beat their Chinese counterparts in the construction of advanced AI.
“The economic uncertainty caused by Trump’s tariffs could be the single biggest barrier to American AI hegemony,” said Sravan Kundojjala, consultant Semian Alicis.
Large tech groups, including Microsoft, Google, Amazon and Meta, have pledged to spend $300 million on the computing infrastructure that supports AI in 2025 alone.
Other projects, such as increasing US chip-making capabilities, such as a $100 million commitment by Taiwanese semiconductor manufacturers, will help support such ambitions
Industry figures warned that these efforts face uncertainty and disruption as tariffs hit a complex global supply chain that serves large AI computing projects.
“The USD 500 million data center project, led by Openai, Softbank and Oracle,” said someone involved in the development of Stargate:
“These are pretty complicated builds (possibly) that delay due to fan switches.”
Semiconductors and associated chip-making equipment, materials and components have been exempt from suspending current “mutual” tariffs announced on the President of the US.
However, analysts said the remaining tariff system, which includes a 145% obligation on goods from China, has increased the costs of building and financing US manufacturing plants and AI data centers.
Altana, a research group that maps global supply chains, said Chinese tariffs alone mean that US data center developers are facing an annual cost increase of over $11 billion.
The US announced this week that it will investigate the national security impact of importing semiconductors and associated chip-making equipment, materials and components as companies attempt to shift production of advanced AI-related hardware to the US.
The probe, known as the Section 232 investigation, which can take up to 270 days to complete, can create even more troublesome demands for the industry. Trump has already called Section 232 authority and has imposed a 25% tariff on the steel, aluminum and automotive sectors.
“No one is ‘off the hook,'” Trump wrote in a social media post on Sunday, adding that his administrator is “looking at the entire semiconductor and electronics supply chain.”
However, analysts said that placing a new mission on importing SMICHNUCTOR proves difficult as most chips enter the US as they have already been integrated into other products such as smartphones, laptops and graphics processing units used in AI data centers.
This includes Nvidia’s most advanced GPUs, which are used by cloud service providers such as Amazon and Microsoft to train and run large-scale language models for companies such as Openai, Google and Elon Musk’s Grok.
Mohammad Ahmad, CEO of supply chain data analytics platform Z2DATA, said most AI GPUs will enter the US in the form of servers or server racks.
GPUs contain chips generated primarily in Taiwan or Korea, but are frequently sent for packaging and testing in Southeast Asian countries such as Malaysia and the Philippines.
The chips are then sent back to Taiwan or Mexico for print circuit board assembly before integration into a server that is exported to the US for use in AI data centers.
“Even if the GPU itself is exempt from tariffs, if the tariffs still apply to components, it will be hit by massive US costs,” Ahmad said. “The number of product categories is huge and the smallest components can defeat the supply chain.”
Semianalysis’ Kundojjala said even the 32% tariff proposed by the Trump administration for chip manufacturing leader imports from Taiwan, US semiconductor production will still be expensive as tariffs push the prices of key tools and materials.
“The threat that relies on the US’ ability to rebuild its land manufacturing is real,” he said. “Building manufacturing capacity outside the US is cheaper, but the companies with the largest share of US manufacturing will lose the most.”
Recommended
An executive at Taiwan’s Chip Design House, which supplies Amazon, said that if the sector is subject to heavy tariffs, his company’s US clients will need to absorb the costs for the next few years.
“The first response from Amazon is to go to the supplier and say, ‘We’re going to produce this in Taiwan and lowering the price because it creates extra costs for me,” they said.
“(Amazon) takes years to build capacity and build products, so we’re not going to require you to have chips made in the US,” added the person. “But we won’t lower prices. If so, we’ll be screwed into the US government because it frustrating their policy of forcing people to do all the American tips.”
Jeffrey Görtz, a senior fellow at the New American Security Center in Washington, said the Trump administration still has the capacity to address risks to the AI industry following a section 232 investigation in “a much broader potential toolkit using government procurement policies, tax law changes, and other trade or non-trade policies to adjust national security risks from these imports.”
He added: “The process is whether it will close immediately with a 25% tariff on tips, or whether this will become a more creative policy process that takes into account the broader potential outcomes.”
Additional Reports by Melissa Heckilla of London
