U.S. imposes 25% tariff on Nvidia’s H200 AI chips headed to China, TechCrunch reports

A new kind of chip policy: export permission plus export tariff

A new U.S. tariff aimed at advanced AI semiconductors is adding cost and complexity to an already contentious question: whether and how American chipmakers should sell high-end processors to Chinese customers. TechCrunch reported that President Donald Trump signed a proclamation imposing a 25% tariff on certain advanced AI chips produced outside the U.S. that pass through the U.S. before being exported abroad, including Nvidia’s H200 chips intended for vetted customers in China.

The announcement is notable because it does not represent a blanket chip tariff; it is targeted to a subset of semiconductors that fit the administration’s definition of advanced AI hardware and that follow a specific logistics pattern. In practice, it means a chip shipment might be authorized under Commerce Department rules yet still face a tariff charge depending on production origin and routing.

How it connects to Commerce Department approvals

TechCrunch framed the tariff as formalizing a key component of the Department of Commerce’s earlier decision to allow Nvidia to resume shipments of H200 chips to approved Chinese customers. The policy also includes other chips such as AMD’s MI325X, indicating that the administration is applying a category-based approach rather than singling out one company.

This arrangement is controversial for two reasons at once: it acknowledges that U.S. firms want access to Chinese demand, but it also treats the export pathway as a revenue opportunity for the U.S. government and as a leverage point for future restrictions.

Why the tariff matters for the AI race

For Nvidia and its peers, the immediate business question is whether customers absorb the cost, negotiate prices, or reduce orders. For Washington, the bigger question is strategic: whether allowing any advanced AI chips into China—tariffed or not—strengthens U.S. firms economically while also potentially accelerating a rival’s AI capabilities.

The policy also creates uncertainty for global supply chains. Chip exports often involve multiple jurisdictions, contract manufacturers, and transit points. A tariff triggered by routing could push companies to redesign logistics, shift assembly locations, or restructure contracts to reduce exposure.

The result is that “permission to sell” and “cost to sell” become two separate levers, giving the administration additional tools to shape the pace, profitability, and volume of advanced AI chip exports.