Why AI is at “Tariff” risk
“The cost of servers, near term, is likely to increase materially, which may have an impact on AI deployments,” many of which are in the U.S.,” wrote New Street Research’s Pierre Ferragu in a note. “How would tariffs apply to a GPU server, selling for $400k, of which $280k is the price of GPUs alone? Will the buyer be allowed to pay for chips directly to their (mostly U.S.) manufacturers, and pay tariffs only on the cost of the server excluding silicon (assembly and other parts)?”
The current tariff structure creates a peculiar scenario where semiconductors are exempt, but other crucial server components face increased costs. This partial exemption creates confusion around complex products like GPU servers, where the chip represents a significant portion of the total cost. The logistics of separating component costs for tariff purposes remain unclear.
The exemption’s stability is questionable, and there’s a looming threat of retaliatory tariffs from other nations, particularly targeting U.S.-designed chips manufactured overseas. This could significantly impact companies like Nvidia and AMD, whose products are crucial to the AI infrastructure buildout.
The broader implications extend beyond immediate cost increases. Enterprise customers might delay or reduce AI investments due to uncertainty, while international companies could pivot toward local AI solutions. For instance, European companies might prefer Mistral over OpenAI, creating a fragmented AI technology landscape.
The ripple effect of “Tariffs” on tech is still not widely understood.