Nvidia Plans $500B AI Server Build in US

Nvidia has announced plans to spearhead the construction of AI servers valued at up to $500 billion in the U.S. over the next four years. This initiative involves key partners like TSMC and aligns with the U.S. government's ongoing efforts to bolster domestic manufacturing capabilities. The move is partly motivated by concerns over potential tariffs and the need for a more resilient supply chain.

Production plans include manufacturing Blackwell AI chips at TSMC's Arizona factory, with additional supercomputer plants slated for Texas by Foxconn and Wistron. While TSMC will handle the front-end processing of Nvidia's Blackwell chips in Arizona, the chips will be shipped to Taiwan for packaging. Nvidia CEO Jensen Huang has emphasized that U.S.-based manufacturing will be crucial in meeting the escalating demand for AI chips and supercomputers.

This announcement follows adjustments in U.S. tariff policies, including temporary exemptions for certain electronics from tariffs on China. However, these exemptions are expected to be short-lived, with new semiconductor tariffs anticipated. The broader strategy involves encouraging domestic manufacturing to reduce reliance on foreign nations and secure the supply chain for critical technologies.