Nvidia operates as a fabless semiconductor company, meaning the design of its cutting-edge chips is handled internally while the physical manufacturing is outsourced to a select group of foundry partners. This model allows the company to focus on architecture and innovation, leveraging the immense capital investment and specialized expertise of the world’s most advanced semiconductor fabrication facilities. The primary locations for this critical work are Taiwan and the United States, with TSMC serving as the cornerstone of production for the most advanced graphics and AI processors.
The Primary Foundry: TSMC and Advanced Packaging
The overwhelming majority of Nvidia’s high-performance GPUs, including those for gaming, data centers, and professional visualization, are manufactured by Taiwan Semiconductor Manufacturing Company. TSMC’s state-of-the-art facilities in Taiwan utilize leading-edge nodes such as N4P, N4, and N3 to produce the dense and complex dies that power the latest architectures. For packaging, which connects the silicon die to the printed circuit board, Nvidia relies heavily on substrates produced by companies like Unimicron and Nan Ya Printed Circuit Board, both of which have significant operations in Taiwan.
Key Assembly and Testing Sites
While the silicon is fabricated in Taiwan, the final assembly and testing of the graphics cards involve a broader network of partners across Asia. Major assembly and testing (ASE) providers include companies like ASE Technology Holding and JCET, which have massive facilities in mainland China, specifically in provinces such as Jiangsu. These sites handle the critical steps of placing the silicon die onto a substrate, encapsulating it for protection, and performing rigorous quality checks before the products are shipped globally.
Geopolitical Shifts and US Production
In response to global trade dynamics and government incentives, Nvidia is expanding its manufacturing footprint into the United States. This effort is centered around a partnership with TSMC to build and operate a new fabrication plant in Arizona. This facility, part of TSMC’s first American wafer fabrication project, is intended to produce advanced chips using N4 and future nodes, aiming to bolster domestic supply chain resilience for critical technologies.
OSAT Partners in the United States
The assembly and testing of these new US-made chips will be handled by Outsourced Semiconductor Assembly and Test providers. Companies like Amkor Technology and STATS ChipPAC operate major facilities in Arizona and other locations, where the bare dies from TSMC will be mounted onto substrates and encapsulated into the final protective packages that become the GPUs used in systems worldwide.
Design, Yield, and Quality Control
Nvidia’s engineering teams work in close collaboration with TSMC to optimize the design for manufacturability. This partnership is crucial for achieving high yields on complex dies, which directly impacts the availability and cost of the final products. The company’s rigorous validation process ensures that every batch of chips, regardless of the foundry location, meets strict performance, power, and thermal specifications before reaching the market.
The Global Supply Chain Network
Beyond the foundries and assembly houses, the production of a single Nvidia graphics card involves a vast international ecosystem. Memory chips from suppliers like Samsung and SK Hynix are sourced globally, as are the passive components and connectors that make up the printed circuit boards. This intricate web of suppliers and logistics ensures that the raw materials and finished components converge efficiently to deliver the final product to consumers and enterprises.
