Trump’s Trade War Casts a Shadow on America’s AI Boom

The Washington Post

12:08 JST, April 13, 2025

SAN FRANCISCO – President Donald Trump has pledged to make the United States the “world capital” of artificial intelligence, but his aggressive trade tariffs threaten to undermine Silicon Valley’s work on the crucial technology and weaken its competition with China.

Trump’s trade levies will drive up the cost of constructing, equipping and operating the data centers that companies such as OpenAI, Google and Microsoft are racing to build to power AI development, according to executives and experts in AI and data center construction.

OpenAI CEO Sam Altman said in an interview on Thursday that the company is urgently trying to figure out how the tariffs will affect the cost of running its AI models. “We’re working around-the-clock on this,” he said.

Trump and tech industry leaders see China as America’s primary rival in AI, and the president has said the U.S. must stay ahead to preserve its economic and national security. After taking office, he quickly moved to cut AI rules introduced under the Biden administration and make it easier to build and power AI data centers.

But Trump’s aggressive tariffs on China – now at an eye-watering 145 percent – may end up helping it compete with the U.S. in artificial intelligence.

“As the administration continues to consider tariff policies, we strongly encourage efforts to provide certainty and the continued evaluation of impacts on critical data center equipment and components at this pivotal moment in the AI race,” said Josh Levi, president of the Data Center Coalition, a trade group representing companies that operate data centers.

The AI industry and data center construction is core to America’s global competitiveness and national security, Levi said.

Software is not subject to Trump’s tariffs, but they apply to the computing equipment needed to develop and deploy it and to outfit new data centers. China, which on Friday retaliated against Trump’s 145 percent tariffs on the country with a 125 percent tariff of its own on U.S. goods, is a key supplier.

The president’s tariff offensive exempts some computer chips but not the expensive and high-powered graphics processing units – or GPUs – that are the workhorses of the AI boom and enabled the creation of software like ChatGPT.

Construction materials required to build data centers, cooling equipment to prevent their computing gear from overheating and backup generators in case of power cuts are also generally sourced from complex global supply chains that now face new import taxes.

“It sounds good that chips are exempted, but on the other hand, there are so many other parts to the cost of a data center,” said Altman, who spoke in an interview after giving remarks at the Vanderbilt Summit on Modern Conflict and Emerging Threats.

The potentially steep cost increases caused by Trump’s tariffs and unpredictable and fast-changing trade policy have sent a wave of uncertainty through an industry that generally sees itself as confidently forging the future.

On Friday, Trump said in a post on his Truth Social account that the tariff policy was “moving along quickly,” without providing details on how.

“The overarching observation is one of uncertainty and confusion and inability to really plan,” said Jay Biggins, executive managing director at BLS and Co., a real estate consulting firm that helps AI data center developers find sites and plan their supply chains.

Artificial intelligence has been an obsession of the tech industry for decades, but OpenAI’s launch of ChatGPT in 2022 triggered an intense race to develop AI and find ways to profit from providing it to consumers and businesses.

The AI boom has not yet produced major profits, but Big Tech companies and venture capitalists have poured billions of dollars into developing and running AI models. They have set off a rush of data center construction across the U.S. The facilities’ huge power consumption has stressed power utilities and triggered plans to reboot mothballed coal and nuclear plants.

Google has said it plans to spend $75 billion this year on data centers for AI, while Microsoft is aiming to spend $80 billion. On the first full day of his second term, Trump hosted a public announcement in the Oval Office from OpenAI’s Altman and leaders from business software giant Oracle and Japanese mega-investor SoftBank, who said they would spend as much as $500 billion on AI data center construction during his term in a project called Stargate.

“The AI data center market is probably the largest market I’ve ever seen in my career” when it comes to construction, said Don Clark, co-CEO of Clark Pacific, a Sacramento-based prefabricated building company that has a growing line of business constructing data centers.

Some industry insiders are worried Trump’s aggressive tariffs policy could constrain that boom.

“I’m very worried,” said Andrew Ng, a former head of Google’s AI lab and the co-founder of AI education start-up DeepLearning.AI. “The U.S. has a lot of current and planned energy infrastructure and data center build-outs. Tariffs will definitely increase the cost.”

Companies at the center of the AI boom had their stock prices smashed by Trump’s initial tariff announcement but have largely recovered in recent days.

Dan Ives, a tech stock analyst with Wedbush Securities known for his optimism about the tech industry, estimated in a Thursday note to clients that at least 10 to 15 percent of the cloud and AI projects in the U.S. will be “slowed down” because of the uncertainty around tariffs.

“Thinking this tariff issue is now done is the wrong view,” he wrote – just a day before China moved its tariffs on U.S. imports higher.

Spokespeople for Microsoft, Nvidia and Google declined to comment.

The intense demand for data center components and long backlogs of orders could make it challenging for U.S. companies to rework their plans for the Trump tariff era.

Even if companies can source generators, switches and transformers from American suppliers, demand is already so intense that prices will have to increase, said Biggins, the real estate consultant.

“All of the essential hardware was already on 24-, 36-month backlogs,” Biggins said. Data center costs could rise by 15 percent or 20 percent, he said.

Some U.S. AI companies may instead decide to build data centers outside America, Biggins said. Ng, the former Google AI lab leader, who has urged the U.S. to invest in AI and avoid regulations that could limit the technology’s development, agreed.

Tech companies generally prefer to build data centers close to their customers so that they can access apps, data and emails with minimal lag, Ng said. But that constraint is less pressing for AI applications, which often require more processing time to respond to user queries.

When a service like ChatGPT takes 10 seconds or more to generate an AI image, adding the few milliseconds needed to send it from a distant data center is no big deal, he said.

Big Tech companies already have data centers all over the world and are placing new ones in locations including Malaysia and Singapore.

“When the regulations change overnight by tweet, it’s difficult to plan,” Ng said. “Unfortunately, this makes other geographies with more stable structures more attractive.”