The recent wave of tariffs announced by President Trump has certainly stirred the pot in the U.S. AI industry. If you’re in the field, you might have noticed the market’s knee-jerk reaction, with stocks taking a hit as companies start to brace for increased costs. These tariffs are likely to ramp up the already high expenses tied to building datacenters, which are crucial for training AI models.
Chris Miller, the author of Chip War, explains, “The tariffs will make building AI datacenters much more expensive.” This is because there’s a heavy reliance on imported AI servers and other critical components like cooling and power systems. While chips themselves are exempt from tariffs if imported separately, they often come bundled in products like servers, which don’t get the same break.
But there’s a silver lining. Analyst Stacy Rasgon points out that many Nvidia servers might avoid these tariffs thanks to their assembly in Mexico, which benefits from a trade agreement exemption. “I think there are some workarounds to avoid massive tariffs on AI infrastructure in the U.S.,” Rasgon suggests, offering a bit of hope.
Despite these potential loopholes, the rising costs of materials and equipment could push companies to think about setting up datacenters abroad. Lucas Hansen from the Civic AI Security Program mentions that tariffs might make it more appealing to relocate where power is cheaper. This is a “real risk,” as Miller puts it, of the U.S. losing its competitive edge in the AI race against China—something the Trump administration is keen to avoid.
Adding to the challenge is the uncertainty of fluctuating tariff rates, which complicates long-term planning for companies. Even with potential tariff exemptions, the broader economic impacts of the trade war could create new hurdles. “My bigger worry is more macro now,” Rasgon mentions, highlighting concerns about a potential recession and reduced spending.
On a brighter note, advancements in algorithms and hardware are expected to make AI usage more affordable, according to research from Epoch AI. So, while datacenter expenses might climb, consumer costs should remain largely unaffected. It’s a complex landscape, but with careful navigation, the AI industry might just find its way through.
